India at the Technological Crossroads: Semiconductors, DRAM, Artificial Intelligence, and the Struggle for Strategic Autonomy in the 21st Century
A Comprehensive Essay on India's Technology Deficit, Its Historical Roots, Policy Failures, Future Pathways, and the Reshaping of the Global Order
This essay is written at a pivotal moment in human history — a moment when the sinews of national power are no longer measured solely in the tonnage of steel produced, barrels of oil extracted, or the size of standing armies. The new grammar of geopolitical dominance is written in nanometers, in petaflops, in the architecture of transformer models, and in the atomic lattices of compound semiconductors. Nations that master these domains will shape the 21st century. Nations that do not will find themselves perpetually dependent, perpetually vulnerable, and perpetually relegated to the consuming end of the global value chain.
India stands at this crossroads today — a civilization of five thousand years of intellectual heritage, the birthplace of the numeral system that undergirds all modern computation, the nation that sent a spacecraft to Mars on its first attempt for less than the cost of the Hollywood film Gravity — and yet, a country that cannot manufacture a single advanced semiconductor chip, does not possess a meaningful DRAM production capacity, has no homegrown large-scale AI foundation model that competes globally, and remains dangerously exposed to the geopolitical weaponization of technology supply chains.
This is not merely a policy failure. It is a civilizational reckoning — an accumulation of decades of mistaken priorities, bureaucratic inertia, ideological dogmatism, and a fundamentally miscalibrated understanding of where national power truly resides. To understand where India is today, we must travel back in time — not just decades, but centuries — to trace the long arc of India's relationship with technology, industrialization, and the global knowledge economy.
What follows is an attempt at a comprehensive diagnosis: historical, technical, economic, geopolitical, and deeply human. Because ultimately, the question of whether India can build a semiconductor industry, train a frontier AI model, or manufacture advanced memory chips is not just a question of policy or finance. It is a question of national will, institutional capacity, and civilizational ambition.
PART ONE: THE LONG SHADOW OF HISTORY — HOW INDIA FELL BEHIND IN THE TECHNOLOGY RACE
The Pre-Colonial Inheritance — A Civilization of Intellectual Supremacy
To fully appreciate the depth of India's current technology deficit, one must begin not with the microchip but with the mathematics. India's contribution to the foundational architecture of modern science and technology is staggering in its breadth and depth. The decimal numeral system — the very scaffolding upon which every computer, every algorithm, every line of code rests — was invented in India and transmitted to the world through Arab scholars in the medieval period. Aryabhata, writing in the 5th century CE, calculated the value of pi to four decimal places and described the rotation of the Earth on its axis. Brahmagupta, in the 7th century, gave the world the concept of zero as a number — arguably the single most consequential mathematical innovation in human history.
The Sulbasutras, composed between 800 and 200 BCE, contained sophisticated geometric principles that predated Pythagoras. The Kerala school of mathematics, flourishing between the 14th and 16th centuries, independently developed infinite series and calculus — concepts that in Europe would not be formalized until Newton and Leibniz in the 17th century. Indian metallurgists produced the famous Delhi Iron Pillar in the 4th century CE — a six-tonne column of wrought iron that has resisted corrosion for over 1,600 years, a metallurgical feat that modern science only recently explained through the formation of a unique iron hydrogen phosphate protective layer.
Wootz steel, produced in South India from as early as 300 BCE, was the finest steel in the ancient world — traded as far as Rome and Persia, forged into the legendary Damascus blades that no European sword could match. Indian textile technology, particularly in fine muslin weaving, operated at levels of precision that were not replicated industrially until the 20th century. The Dhaka muslin known as "woven air" had thread counts exceeding 1,000 per inch — a standard that modern industrial looms struggle to achieve.
This is the inheritance India brings to the technological age. It is not the inheritance of a civilization that was always behind. It is the inheritance of a civilization that was once at the frontier and was then — through a specific, identifiable, historically contingent process — dispossessed of that frontier status.
The Colonial Disruption — Deindustrialization and the Destruction of Technological Capacity
The story of how India's technological trajectory was disrupted is inseparable from the story of British colonialism. The British East India Company's conquest of Bengal in 1757 — the Battle of Plassey — marked the beginning of a systematic process of economic extraction and deindustrialization that would, over the course of nearly two centuries, reduce one of the world's most sophisticated manufacturing economies to a raw material supplier for British industry.
At the time of Plassey, India accounted for approximately 24.4% of global GDP — comparable to China's share and vastly exceeding Britain's. By the time of independence in 1947, that share had collapsed to approximately 3-4%. This was not the natural consequence of India's failure to industrialize. It was the deliberate consequence of British policies specifically designed to prevent Indian industrialization.
The cotton textile industry tells the story most vividly. Indian weavers of Dhaka, Surat, and Murshidabad produced cotton fabrics of extraordinary quality that dominated Asian and European markets. British merchants, unable to compete, pressured the British Parliament to impose prohibitive tariffs on Indian textiles — duties of up to 70-80% — while flooding India with machine-made British cloth tariff-free. The result was the systematic destruction of India's textile industry. The weavers of Dhaka — who had numbered in the hundreds of thousands — were reduced to penury within decades. Some accounts, though perhaps apocryphal, speak of British administrators cutting off the thumbs of Dhaka weavers to prevent them from competing. Whether literally true or not, it captures the essential violence of the process.
The same pattern repeated across sectors. Indian shipbuilding, which had produced vessels that British merchants themselves preferred for their superior quality, was suppressed by Acts of Parliament that prohibited Indian-built ships from carrying goods between British ports. Indian iron smelting — which had produced high-quality iron from local ores using sophisticated charcoal-based processes — was undermined by cheap British iron made possible by coke smelting and the scale economies of the Industrial Revolution.
What is crucial to understand is that this was not simply a matter of India failing to adopt new technologies. India did not have the opportunity to fail on its own terms. The structural conditions under which technological adoption and industrial development occur — capital accumulation, tariff protection, domestic market development, state investment in infrastructure — were systematically denied to India by colonial policy. Britain's own Industrial Revolution had been enabled by precisely these conditions: high tariffs protecting nascent industries, state support for infrastructure, and aggressive acquisition of export markets. India was denied all of these.
The colonial period also had profound effects on India's educational and scientific institutions. While the British did establish universities and colleges in India, the explicit purpose of these institutions — as stated famously by Thomas Babington Macaulay in his 1835 Minute on Indian Education — was to produce "a class of persons Indian in blood and colour, but English in tastes, in opinions, in morals and in intellect" who could serve as intermediaries in the colonial administration. The emphasis was on classical humanities, law, and administration — not on technical education, engineering, or applied science. The Indian Institutes of Technology that would later become world-famous had no colonial antecedents — the first IIT was established in Kharagpur in 1951, four years after independence.
The railway system that the British built in India — often cited as a colonial benefit — was designed primarily to extract resources from the interior to the ports, not to build an integrated industrial economy. The railways did not connect industrial centers to facilitate manufacturing; they connected mines and agricultural regions to coastal ports. The steel for the railways was imported from Britain. The locomotives were built in Britain. The engineering expertise was British. India served as the market, not as the producer.
By 1947, India inherited an economy that had been structurally distorted by 190 years of deliberate colonial underdevelopment. It had a small, underfunded scientific establishment, a negligible industrial base, almost no technological manufacturing capacity, and a population of 340 million people of whom 84% were illiterate. The technological deficit with the advanced industrial nations was not decades but centuries in the making.
The Nehruvian Vision — Temples of Modern India and the Mixed Record of Import Substitution
Jawaharlal Nehru, India's first Prime Minister, understood with clarity that independence without economic and technological self-sufficiency was incomplete. His vision for India was explicitly technological: he famously called the new steel plants, hydroelectric dams, and research laboratories the "temples of modern India." He established the Indian Institutes of Technology, the Indian Institutes of Management, the Council of Scientific and Industrial Research, the Atomic Energy Commission, the Defence Research and Development Organisation, and the Indian Space Research Organisation. He invested heavily in scientific infrastructure at a time when India could barely afford to feed its population.
Nehru's economic framework, however, was shaped by the dominant economic thinking of his era — a synthesis of Soviet planning, Fabian socialism, and Keynesian demand management. The result was the famous "License Raj" — a system of centralized planning in which virtually every significant economic decision required government approval. Firms needed licenses to expand capacity, to import technology, to enter new product lines, to price their products. The Planning Commission allocated investment across sectors through Five-Year Plans modeled on Soviet precedent.
This system had a specific, identifiable logic: India, deeply scarred by colonial economic exploitation, was determined to build domestic industrial capacity behind high protective tariff walls. The strategy was "Import Substitution Industrialization" (ISI) — rather than importing finished goods, India would build the capacity to produce them domestically. Tariff walls would protect nascent industries until they could compete globally.
The ISI strategy had genuine successes. India built a substantial steel industry, a heavy engineering sector, a petrochemical complex, a pharmaceutical manufacturing base, and a growing defense industrial complex. By the 1970s, India was producing locomotives, heavy machine tools, power generation equipment, and a range of capital goods that it had previously imported entirely. The Bhilai Steel Plant, the Bokaro Steel Plant, the Fertilizer Corporation of India — these were real industrial achievements, built against enormous odds.
But ISI also had catastrophic failures, and the failures were particularly stark in exactly the domains that would come to matter most: electronics, computing, and advanced manufacturing technology.
The transistor was invented in 1947 — the same year as Indian independence. The integrated circuit followed in 1958. The first commercial silicon chips appeared in the early 1960s. By the mid-1960s, Silicon Valley was taking shape as the global center of semiconductor development, and American firms like Intel, Texas Instruments, Fairchild Semiconductor, and Motorola were building the foundations of the digital age.
India was not entirely absent from this early technological moment. The Indian statistician P.C. Mahalanobis had established IISCO (Indian Iron and Steel Company) and was a globally recognized figure. Homi Bhabha, the physicist who founded India's nuclear program, was deeply connected to the international scientific community and understood the importance of computing. The Tata Institute of Fundamental Research (TIFR) built India's first digital computer — TIFRAC (Tata Institute of Fundamental Research Automatic Calculator) — in 1960, making India one of the first Asian countries to have a domestically built digital computer.
But these bright spots could not compensate for the structural failure of the ISI framework to nurture a genuinely competitive electronics industry. The electronics industry was treated as a strategic sector requiring tight controls, which in practice meant that Indian companies could not access the latest foreign technology, could not partner freely with global firms, and operated in a protected market that insulated them from competitive pressure. The result was that Indian electronics firms fell further and further behind the global frontier.
The IBM episode is particularly instructive. IBM had operated in India since 1951, providing mainframe computers to Indian businesses and government agencies. In 1973, the Indian government, under the Foreign Exchange Regulation Act (FERA), demanded that IBM reduce its equity stake to 40% and transfer technology to Indian partners. IBM refused and exited India in 1977. This was, on one level, a principled assertion of sovereignty against a powerful multinational. On another level, it severed India's connection to the leading edge of global computing technology at a critical moment. The domestic firms that attempted to fill the gap — companies like Hindustan Computers Limited (HCL), formed in 1976 — did remarkable things with limited resources, but they could not match the technological dynamism of global leaders.
The License Raj imposed specific, quantifiable costs on the Indian electronics and technology sector. To import a single computer, a firm needed to navigate a bureaucratic process involving multiple ministries, could wait years for approval, and would typically find that by the time the approval came through, the machine they had applied to import was already technologically obsolete. This created a systematic lag — Indian firms always operated one or two technology generations behind the global frontier.
The Lost Decades — The 1970s and 1980s in Global Semiconductor Context
While India wrestled with the License Raj, the global semiconductor industry underwent the most consequential technological transformation in economic history. Understanding this period is essential to understanding the depth of India's current deficit.
The year 1971 marked the beginning of the modern semiconductor era. Intel introduced the 4004 — the world's first commercially available microprocessor — a chip containing 2,300 transistors on a 10-micron process node, capable of executing 92,000 instructions per second. The integrated circuit had taken the leap from specialized component to general-purpose computing engine. Gordon Moore's observation — that the number of transistors on a chip would double approximately every two years — became the organizing principle of an entire industry.
What followed over the next twenty years was a technological revolution of almost incomprehensible velocity. By 1974, Intel's 8080 contained 6,000 transistors. By 1982, the 286 contained 134,000. By 1989, the 486 contained 1.2 million. By 1993, the Pentium contained 3.1 million. Each generation brought not just more computing power but fundamentally new capabilities — capabilities that transformed what computers could do and who could use them.
The industry that drove this revolution was built around a specific geographic cluster — Silicon Valley in Northern California — and a specific set of institutions: venture capital firms willing to fund high-risk technological bets, research universities (Stanford, Berkeley, Caltech, MIT) that produced a continuous stream of engineers and generated fundamental research, and a defense establishment (DARPA, the Department of Defense) that provided both funding and guaranteed procurement for early-stage technologies.
The DRAM — Dynamic Random Access Memory — story is central to this period and directly relevant to India's current situation. DRAM was invented by Robert Dennard at IBM in 1966. By the early 1970s, it had become the dominant form of computer memory, replacing earlier magnetic core memory. The DRAM market became fiercely competitive, with American firms like Intel, Texas Instruments, and Motorola initially dominating.
But the 1970s and 1980s saw the dramatic entry of Japanese firms into the DRAM market — a competitive battle that would reshape the global semiconductor industry and offer profound lessons for anyone thinking about how to build semiconductor capability from scratch. Japanese firms — NEC, Hitachi, Toshiba, Fujitsu — backed by coordinated government support through MITI (the Ministry of International Trade and Industry) and the Japan Development Bank, invested massively in DRAM manufacturing. They achieved superior quality, higher yields, and lower costs through relentless process optimization. By the mid-1980s, Japanese firms had captured over 80% of the global DRAM market, driving most American DRAM manufacturers out of the business entirely.
This Japanese challenge provoked an extraordinary American political and industrial response — including the US-Japan Semiconductor Trade Agreement of 1986, which is often cited as one of the first instances of technology trade conflict being used as a geopolitical tool. The battle for semiconductor supremacy between the United States and Japan in the 1980s was the direct predecessor of the US-China technology conflict of the 2020s — a template for how nations weaponize technological advantage.
Where was India in all of this? Essentially absent. The Indian electronics industry of the 1970s and 1980s was focused primarily on reverse-engineering consumer electronics — radios, televisions, calculators — for the protected domestic market. The government's Computer Policy of 1978 and 1984 showed some awareness of the importance of the computer industry, but the measures taken were insufficient to build genuine competitive capacity.
Rajiv Gandhi's tenure as Prime Minister (1984-1989) brought a partial opening — import duties on computers were reduced, and there was greater recognition of the importance of the IT sector. Sam Pitroda's initiative to modernize India's telecommunications infrastructure — introducing digital telephone exchanges — was genuinely transformative for India's connectivity. But the semiconductor and hardware manufacturing dimensions of the technology revolution were largely neglected.
The critical point is this: the 1970s and 1980s were the window during which the global semiconductor industry's foundational architecture was established. The firms that dominated then — Intel in processors, Samsung and Micron in memory, TSMC (founded 1987) in foundry services — became the global leaders, building technological leads that compounded over decades. Countries that participated in this foundational phase — the United States, Japan, South Korea, Taiwan — built institutions, skills, supply chains, and intellectual property portfolios that gave them structural advantages that are almost impossible to replicate from scratch. Countries that did not participate — including India — face a gap that is not merely financial but deeply structural.
The 1991 Reforms and the IT Services Paradox
In 1991, India faced a balance of payments crisis so severe that it had to pledge its gold reserves as collateral for emergency IMF loans. The crisis forced a fundamental rethinking of economic policy. Finance Minister Manmohan Singh, working under Prime Minister P.V. Narasimha Rao, introduced a program of economic liberalization that dismantled much of the License Raj, reduced tariffs, opened India to foreign investment, and set the economy on a higher growth trajectory.
The liberalization of 1991 had profound effects on India's technology sector, but its effects were paradoxical. On one hand, it enabled the extraordinary rise of India's IT services industry — the story of Infosys, Wipro, TCS, and HCL that is now globally famous. On the other hand, it may have actually deepened India's deficit in hardware and semiconductor manufacturing by substituting cheap software services for the harder work of building domestic technological production capacity.
The IT services story is genuinely remarkable. Indian engineers, educated at the IITs and regional engineering colleges, proved extraordinarily capable at software development, systems integration, and IT services delivery. The combination of English language proficiency, strong mathematical training, and a massive wage differential with the United States created a compelling proposition: Indian firms could deliver high-quality software services at a fraction of the cost of comparable American work.
The Y2K crisis of the late 1990s served as a forcing function — the desperate need to fix date-related bugs in millions of lines of legacy code created a massive demand for programmers that American firms could not meet domestically. Indian IT firms stepped in, demonstrating their capability at scale, and established relationships that persisted long after Y2K. The dot-com boom of 1999-2000 further accelerated demand for Indian IT services, as American startups needed to build technology infrastructure quickly and cheaply.
By the mid-2000s, India had established itself as the global capital of IT outsourcing. Bangalore, Hyderabad, Pune, Chennai, and Gurgaon had developed into thriving technology hubs. Infosys and Wipro were Fortune 500-scale companies. Indian engineers at Microsoft, Google, Intel, and other Silicon Valley firms were rising to positions of prominence.
But here is the paradox: India's success in IT services was built on a business model that was fundamentally about consuming and deploying technology developed elsewhere, not about creating new technology. Indian IT firms wrote code that ran on Intel chips, stored on Seagate hard drives, transmitted over Qualcomm modems, processed by Microsoft operating systems, and managed through Oracle databases. Not one of these critical technology components was Indian. The entire IT services edifice was built on a foundation of imported technology.
This created what might be called the IT Services Trap: India's most technically skilled people — its IIT graduates, its top engineers — were absorbed into an industry optimized for services delivery rather than technology creation. The incentive structure of IT services — high salaries, foreign exchange earnings, relatively predictable work — drew talent away from the harder, riskier, more uncertain work of building new hardware technologies, developing novel algorithms, or creating breakthrough products.
More fundamentally, the IT services success created a policy illusion: India appeared to be a technology powerhouse when in reality it was a technology services powerhouse — a crucial distinction. The country that writes code for Wall Street banks is not the same as the country that designs the chips those banks' computers run on. The former can be replicated with sufficient English-language education and mathematics training. The latter requires decades of accumulated engineering knowledge, specialized capital equipment, and complex supply chains that cannot be assembled quickly.
This illusion — that India's IT services success represented genuine technological capability — may have been the single most important factor in delaying India's recognition of its semiconductor deficit. If Indian companies were managing the IT systems of every major American bank and corporation, surely India was a technology leader? The answer, of course, is no — and the distinction between technology consumption and technology creation is the core of India's current challenge.
PART TWO: THE TECHNICAL LANDSCAPE — UNDERSTANDING SEMICONDUCTORS, DRAM, AND AI
The Semiconductor — The Atom of the Digital Age
To understand what India lacks, one must first understand what semiconductors are, why they matter, and why making them is so extraordinarily difficult. The semiconductor is the foundational component of the modern economy — the physical substrate upon which the digital world is built. Every smartphone, every laptop, every data center server, every electric vehicle, every medical device, every modern weapons system runs on semiconductors. The global semiconductor market was valued at approximately $574 billion in 2022 and is projected to exceed $1 trillion by 2030.
A semiconductor is, at its most basic, a material — typically silicon — whose electrical conductivity lies between that of a conductor (like copper) and an insulator (like rubber). By doping silicon with specific impurities in precisely controlled quantities and patterns, engineers can create regions of different electrical behavior that can function as switches, amplifiers, and logic gates. The integrated circuit packs millions, billions, and now trillions of these switches onto a single chip of silicon the size of a fingernail.
The journey from sand (silicon dioxide, SiO₂) to a finished semiconductor chip involves hundreds of distinct manufacturing steps, each requiring extraordinary precision. The process begins with the production of ultra-pure silicon — so pure that impurities must be measured in parts per trillion. Silicon ingots are grown through the Czochralski process, sliced into wafers, polished to atomic-scale flatness, and then subjected to photolithography — the process by which circuit patterns are etched onto the silicon surface using light.
Modern photolithography uses Extreme Ultraviolet (EUV) light — light with a wavelength of 13.5 nanometers — to print circuit features smaller than the wavelength of visible light. This requires the most sophisticated optical systems ever built — mirrors polished to tolerances measured in fractions of an atom, plasma-generated light sources, and vacuum chambers where even a single dust particle would be catastrophic. The EUV lithography machines made by ASML of the Netherlands are arguably the most complex manufactured objects in human history, containing over 100,000 components, weighing 180 tonnes, and costing over $150 million each.
The most advanced chips today — produced by Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung — are manufactured on 3-nanometer and 2-nanometer process nodes, meaning the smallest features are approximately 3-2 times the width of a DNA strand. A single chip may contain over 100 billion transistors. The yield management — ensuring that chips function correctly — requires statistical process control of hundreds of variables simultaneously.
This manufacturing capability has taken decades to build and represents the accumulated learning of thousands of engineers over fifty years. The knowledge embedded in a modern semiconductor fab is not in any manual or textbook — it is in the tacit knowledge of experienced engineers, in the institutional memory of manufacturing teams, in the calibration data of specific machines, in the understanding of subtle interactions between process variables that only reveals itself through years of practice.
This is why semiconductor manufacturing is often described as the hardest thing humans have ever built at scale. And it is why the barriers to entry are so formidable that, despite semiconductors being enormously profitable, fewer than a dozen companies in the world are capable of manufacturing them at the frontier, and only three — TSMC, Samsung, and Intel — are at the absolute leading edge.
This essay is written at a pivotal moment in human history — a moment when the sinews of national power are no longer measured solely in the tonnage of steel produced, barrels of oil extracted, or the size of standing armies. The new grammar of geopolitical dominance is written in nanometers, in petaflops, in the architecture of transformer models, and in the atomic lattices of compound semiconductors. Nations that master these domains will shape the 21st century. Nations that do not will find themselves perpetually dependent, perpetually vulnerable, and perpetually relegated to the consuming end of the global value chain.
India stands at this crossroads today — a civilization of five thousand years of intellectual heritage, the birthplace of the numeral system that undergirds all modern computation, the nation that sent a spacecraft to Mars on its first attempt for less than the cost of the Hollywood film Gravity — and yet, a country that cannot manufacture a single advanced semiconductor chip, does not possess a meaningful DRAM production capacity, has no homegrown large-scale AI foundation model that competes globally, and remains dangerously exposed to the geopolitical weaponization of technology supply chains.
This is not merely a policy failure. It is a civilizational reckoning — an accumulation of decades of mistaken priorities, bureaucratic inertia, ideological dogmatism, and a fundamentally miscalibrated understanding of where national power truly resides. To understand where India is today, we must travel back in time — not just decades, but centuries — to trace the long arc of India's relationship with technology, industrialization, and the global knowledge economy.
What follows is an attempt at a comprehensive diagnosis: historical, technical, economic, geopolitical, and deeply human. Because ultimately, the question of whether India can build a semiconductor industry, train a frontier AI model, or manufacture advanced memory chips is not just a question of policy or finance. It is a question of national will, institutional capacity, and civilizational ambition.
PART ONE: THE LONG SHADOW OF HISTORY — HOW INDIA FELL BEHIND IN THE TECHNOLOGY RACE
The Pre-Colonial Inheritance — A Civilization of Intellectual Supremacy
To fully appreciate the depth of India's current technology deficit, one must begin not with the microchip but with the mathematics. India's contribution to the foundational architecture of modern science and technology is staggering in its breadth and depth. The decimal numeral system — the very scaffolding upon which every computer, every algorithm, every line of code rests — was invented in India and transmitted to the world through Arab scholars in the medieval period. Aryabhata, writing in the 5th century CE, calculated the value of pi to four decimal places and described the rotation of the Earth on its axis. Brahmagupta, in the 7th century, gave the world the concept of zero as a number — arguably the single most consequential mathematical innovation in human history.
The Sulbasutras, composed between 800 and 200 BCE, contained sophisticated geometric principles that predated Pythagoras. The Kerala school of mathematics, flourishing between the 14th and 16th centuries, independently developed infinite series and calculus — concepts that in Europe would not be formalized until Newton and Leibniz in the 17th century. Indian metallurgists produced the famous Delhi Iron Pillar in the 4th century CE — a six-tonne column of wrought iron that has resisted corrosion for over 1,600 years, a metallurgical feat that modern science only recently explained through the formation of a unique iron hydrogen phosphate protective layer.
Wootz steel, produced in South India from as early as 300 BCE, was the finest steel in the ancient world — traded as far as Rome and Persia, forged into the legendary Damascus blades that no European sword could match. Indian textile technology, particularly in fine muslin weaving, operated at levels of precision that were not replicated industrially until the 20th century. The Dhaka muslin known as "woven air" had thread counts exceeding 1,000 per inch — a standard that modern industrial looms struggle to achieve.
This is the inheritance India brings to the technological age. It is not the inheritance of a civilization that was always behind. It is the inheritance of a civilization that was once at the frontier and was then — through a specific, identifiable, historically contingent process — dispossessed of that frontier status.
The Colonial Disruption — Deindustrialization and the Destruction of Technological Capacity
The story of how India's technological trajectory was disrupted is inseparable from the story of British colonialism. The British East India Company's conquest of Bengal in 1757 — the Battle of Plassey — marked the beginning of a systematic process of economic extraction and deindustrialization that would, over the course of nearly two centuries, reduce one of the world's most sophisticated manufacturing economies to a raw material supplier for British industry.
At the time of Plassey, India accounted for approximately 24.4% of global GDP — comparable to China's share and vastly exceeding Britain's. By the time of independence in 1947, that share had collapsed to approximately 3-4%. This was not the natural consequence of India's failure to industrialize. It was the deliberate consequence of British policies specifically designed to prevent Indian industrialization.
The cotton textile industry tells the story most vividly. Indian weavers of Dhaka, Surat, and Murshidabad produced cotton fabrics of extraordinary quality that dominated Asian and European markets. British merchants, unable to compete, pressured the British Parliament to impose prohibitive tariffs on Indian textiles — duties of up to 70-80% — while flooding India with machine-made British cloth tariff-free. The result was the systematic destruction of India's textile industry. The weavers of Dhaka — who had numbered in the hundreds of thousands — were reduced to penury within decades. Some accounts, though perhaps apocryphal, speak of British administrators cutting off the thumbs of Dhaka weavers to prevent them from competing. Whether literally true or not, it captures the essential violence of the process.
The same pattern repeated across sectors. Indian shipbuilding, which had produced vessels that British merchants themselves preferred for their superior quality, was suppressed by Acts of Parliament that prohibited Indian-built ships from carrying goods between British ports. Indian iron smelting — which had produced high-quality iron from local ores using sophisticated charcoal-based processes — was undermined by cheap British iron made possible by coke smelting and the scale economies of the Industrial Revolution.
What is crucial to understand is that this was not simply a matter of India failing to adopt new technologies. India did not have the opportunity to fail on its own terms. The structural conditions under which technological adoption and industrial development occur — capital accumulation, tariff protection, domestic market development, state investment in infrastructure — were systematically denied to India by colonial policy. Britain's own Industrial Revolution had been enabled by precisely these conditions: high tariffs protecting nascent industries, state support for infrastructure, and aggressive acquisition of export markets. India was denied all of these.
The colonial period also had profound effects on India's educational and scientific institutions. While the British did establish universities and colleges in India, the explicit purpose of these institutions — as stated famously by Thomas Babington Macaulay in his 1835 Minute on Indian Education — was to produce "a class of persons Indian in blood and colour, but English in tastes, in opinions, in morals and in intellect" who could serve as intermediaries in the colonial administration. The emphasis was on classical humanities, law, and administration — not on technical education, engineering, or applied science. The Indian Institutes of Technology that would later become world-famous had no colonial antecedents — the first IIT was established in Kharagpur in 1951, four years after independence.
The railway system that the British built in India — often cited as a colonial benefit — was designed primarily to extract resources from the interior to the ports, not to build an integrated industrial economy. The railways did not connect industrial centers to facilitate manufacturing; they connected mines and agricultural regions to coastal ports. The steel for the railways was imported from Britain. The locomotives were built in Britain. The engineering expertise was British. India served as the market, not as the producer.
By 1947, India inherited an economy that had been structurally distorted by 190 years of deliberate colonial underdevelopment. It had a small, underfunded scientific establishment, a negligible industrial base, almost no technological manufacturing capacity, and a population of 340 million people of whom 84% were illiterate. The technological deficit with the advanced industrial nations was not decades but centuries in the making.
The Nehruvian Vision — Temples of Modern India and the Mixed Record of Import Substitution
Jawaharlal Nehru, India's first Prime Minister, understood with clarity that independence without economic and technological self-sufficiency was incomplete. His vision for India was explicitly technological: he famously called the new steel plants, hydroelectric dams, and research laboratories the "temples of modern India." He established the Indian Institutes of Technology, the Indian Institutes of Management, the Council of Scientific and Industrial Research, the Atomic Energy Commission, the Defence Research and Development Organisation, and the Indian Space Research Organisation. He invested heavily in scientific infrastructure at a time when India could barely afford to feed its population.
Nehru's economic framework, however, was shaped by the dominant economic thinking of his era — a synthesis of Soviet planning, Fabian socialism, and Keynesian demand management. The result was the famous "License Raj" — a system of centralized planning in which virtually every significant economic decision required government approval. Firms needed licenses to expand capacity, to import technology, to enter new product lines, to price their products. The Planning Commission allocated investment across sectors through Five-Year Plans modeled on Soviet precedent.
This system had a specific, identifiable logic: India, deeply scarred by colonial economic exploitation, was determined to build domestic industrial capacity behind high protective tariff walls. The strategy was "Import Substitution Industrialization" (ISI) — rather than importing finished goods, India would build the capacity to produce them domestically. Tariff walls would protect nascent industries until they could compete globally.
The ISI strategy had genuine successes. India built a substantial steel industry, a heavy engineering sector, a petrochemical complex, a pharmaceutical manufacturing base, and a growing defense industrial complex. By the 1970s, India was producing locomotives, heavy machine tools, power generation equipment, and a range of capital goods that it had previously imported entirely. The Bhilai Steel Plant, the Bokaro Steel Plant, the Fertilizer Corporation of India — these were real industrial achievements, built against enormous odds.
But ISI also had catastrophic failures, and the failures were particularly stark in exactly the domains that would come to matter most: electronics, computing, and advanced manufacturing technology.
The transistor was invented in 1947 — the same year as Indian independence. The integrated circuit followed in 1958. The first commercial silicon chips appeared in the early 1960s. By the mid-1960s, Silicon Valley was taking shape as the global center of semiconductor development, and American firms like Intel, Texas Instruments, Fairchild Semiconductor, and Motorola were building the foundations of the digital age.
India was not entirely absent from this early technological moment. The Indian statistician P.C. Mahalanobis had established IISCO (Indian Iron and Steel Company) and was a globally recognized figure. Homi Bhabha, the physicist who founded India's nuclear program, was deeply connected to the international scientific community and understood the importance of computing. The Tata Institute of Fundamental Research (TIFR) built India's first digital computer — TIFRAC (Tata Institute of Fundamental Research Automatic Calculator) — in 1960, making India one of the first Asian countries to have a domestically built digital computer.
But these bright spots could not compensate for the structural failure of the ISI framework to nurture a genuinely competitive electronics industry. The electronics industry was treated as a strategic sector requiring tight controls, which in practice meant that Indian companies could not access the latest foreign technology, could not partner freely with global firms, and operated in a protected market that insulated them from competitive pressure. The result was that Indian electronics firms fell further and further behind the global frontier.
The IBM episode is particularly instructive. IBM had operated in India since 1951, providing mainframe computers to Indian businesses and government agencies. In 1973, the Indian government, under the Foreign Exchange Regulation Act (FERA), demanded that IBM reduce its equity stake to 40% and transfer technology to Indian partners. IBM refused and exited India in 1977. This was, on one level, a principled assertion of sovereignty against a powerful multinational. On another level, it severed India's connection to the leading edge of global computing technology at a critical moment. The domestic firms that attempted to fill the gap — companies like Hindustan Computers Limited (HCL), formed in 1976 — did remarkable things with limited resources, but they could not match the technological dynamism of global leaders.
The License Raj imposed specific, quantifiable costs on the Indian electronics and technology sector. To import a single computer, a firm needed to navigate a bureaucratic process involving multiple ministries, could wait years for approval, and would typically find that by the time the approval came through, the machine they had applied to import was already technologically obsolete. This created a systematic lag — Indian firms always operated one or two technology generations behind the global frontier.
The Lost Decades — The 1970s and 1980s in Global Semiconductor Context
While India wrestled with the License Raj, the global semiconductor industry underwent the most consequential technological transformation in economic history. Understanding this period is essential to understanding the depth of India's current deficit.
The year 1971 marked the beginning of the modern semiconductor era. Intel introduced the 4004 — the world's first commercially available microprocessor — a chip containing 2,300 transistors on a 10-micron process node, capable of executing 92,000 instructions per second. The integrated circuit had taken the leap from specialized component to general-purpose computing engine. Gordon Moore's observation — that the number of transistors on a chip would double approximately every two years — became the organizing principle of an entire industry.
What followed over the next twenty years was a technological revolution of almost incomprehensible velocity. By 1974, Intel's 8080 contained 6,000 transistors. By 1982, the 286 contained 134,000. By 1989, the 486 contained 1.2 million. By 1993, the Pentium contained 3.1 million. Each generation brought not just more computing power but fundamentally new capabilities — capabilities that transformed what computers could do and who could use them.
The industry that drove this revolution was built around a specific geographic cluster — Silicon Valley in Northern California — and a specific set of institutions: venture capital firms willing to fund high-risk technological bets, research universities (Stanford, Berkeley, Caltech, MIT) that produced a continuous stream of engineers and generated fundamental research, and a defense establishment (DARPA, the Department of Defense) that provided both funding and guaranteed procurement for early-stage technologies.
The DRAM — Dynamic Random Access Memory — story is central to this period and directly relevant to India's current situation. DRAM was invented by Robert Dennard at IBM in 1966. By the early 1970s, it had become the dominant form of computer memory, replacing earlier magnetic core memory. The DRAM market became fiercely competitive, with American firms like Intel, Texas Instruments, and Motorola initially dominating.
But the 1970s and 1980s saw the dramatic entry of Japanese firms into the DRAM market — a competitive battle that would reshape the global semiconductor industry and offer profound lessons for anyone thinking about how to build semiconductor capability from scratch. Japanese firms — NEC, Hitachi, Toshiba, Fujitsu — backed by coordinated government support through MITI (the Ministry of International Trade and Industry) and the Japan Development Bank, invested massively in DRAM manufacturing. They achieved superior quality, higher yields, and lower costs through relentless process optimization. By the mid-1980s, Japanese firms had captured over 80% of the global DRAM market, driving most American DRAM manufacturers out of the business entirely.
This Japanese challenge provoked an extraordinary American political and industrial response — including the US-Japan Semiconductor Trade Agreement of 1986, which is often cited as one of the first instances of technology trade conflict being used as a geopolitical tool. The battle for semiconductor supremacy between the United States and Japan in the 1980s was the direct predecessor of the US-China technology conflict of the 2020s — a template for how nations weaponize technological advantage.
Where was India in all of this? Essentially absent. The Indian electronics industry of the 1970s and 1980s was focused primarily on reverse-engineering consumer electronics — radios, televisions, calculators — for the protected domestic market. The government's Computer Policy of 1978 and 1984 showed some awareness of the importance of the computer industry, but the measures taken were insufficient to build genuine competitive capacity.
Rajiv Gandhi's tenure as Prime Minister (1984-1989) brought a partial opening — import duties on computers were reduced, and there was greater recognition of the importance of the IT sector. Sam Pitroda's initiative to modernize India's telecommunications infrastructure — introducing digital telephone exchanges — was genuinely transformative for India's connectivity. But the semiconductor and hardware manufacturing dimensions of the technology revolution were largely neglected.
The critical point is this: the 1970s and 1980s were the window during which the global semiconductor industry's foundational architecture was established. The firms that dominated then — Intel in processors, Samsung and Micron in memory, TSMC (founded 1987) in foundry services — became the global leaders, building technological leads that compounded over decades. Countries that participated in this foundational phase — the United States, Japan, South Korea, Taiwan — built institutions, skills, supply chains, and intellectual property portfolios that gave them structural advantages that are almost impossible to replicate from scratch. Countries that did not participate — including India — face a gap that is not merely financial but deeply structural.
The 1991 Reforms and the IT Services Paradox
In 1991, India faced a balance of payments crisis so severe that it had to pledge its gold reserves as collateral for emergency IMF loans. The crisis forced a fundamental rethinking of economic policy. Finance Minister Manmohan Singh, working under Prime Minister P.V. Narasimha Rao, introduced a program of economic liberalization that dismantled much of the License Raj, reduced tariffs, opened India to foreign investment, and set the economy on a higher growth trajectory.
The liberalization of 1991 had profound effects on India's technology sector, but its effects were paradoxical. On one hand, it enabled the extraordinary rise of India's IT services industry — the story of Infosys, Wipro, TCS, and HCL that is now globally famous. On the other hand, it may have actually deepened India's deficit in hardware and semiconductor manufacturing by substituting cheap software services for the harder work of building domestic technological production capacity.
The IT services story is genuinely remarkable. Indian engineers, educated at the IITs and regional engineering colleges, proved extraordinarily capable at software development, systems integration, and IT services delivery. The combination of English language proficiency, strong mathematical training, and a massive wage differential with the United States created a compelling proposition: Indian firms could deliver high-quality software services at a fraction of the cost of comparable American work.
The Y2K crisis of the late 1990s served as a forcing function — the desperate need to fix date-related bugs in millions of lines of legacy code created a massive demand for programmers that American firms could not meet domestically. Indian IT firms stepped in, demonstrating their capability at scale, and established relationships that persisted long after Y2K. The dot-com boom of 1999-2000 further accelerated demand for Indian IT services, as American startups needed to build technology infrastructure quickly and cheaply.
By the mid-2000s, India had established itself as the global capital of IT outsourcing. Bangalore, Hyderabad, Pune, Chennai, and Gurgaon had developed into thriving technology hubs. Infosys and Wipro were Fortune 500-scale companies. Indian engineers at Microsoft, Google, Intel, and other Silicon Valley firms were rising to positions of prominence.
But here is the paradox: India's success in IT services was built on a business model that was fundamentally about consuming and deploying technology developed elsewhere, not about creating new technology. Indian IT firms wrote code that ran on Intel chips, stored on Seagate hard drives, transmitted over Qualcomm modems, processed by Microsoft operating systems, and managed through Oracle databases. Not one of these critical technology components was Indian. The entire IT services edifice was built on a foundation of imported technology.
This created what might be called the IT Services Trap: India's most technically skilled people — its IIT graduates, its top engineers — were absorbed into an industry optimized for services delivery rather than technology creation. The incentive structure of IT services — high salaries, foreign exchange earnings, relatively predictable work — drew talent away from the harder, riskier, more uncertain work of building new hardware technologies, developing novel algorithms, or creating breakthrough products.
More fundamentally, the IT services success created a policy illusion: India appeared to be a technology powerhouse when in reality it was a technology services powerhouse — a crucial distinction. The country that writes code for Wall Street banks is not the same as the country that designs the chips those banks' computers run on. The former can be replicated with sufficient English-language education and mathematics training. The latter requires decades of accumulated engineering knowledge, specialized capital equipment, and complex supply chains that cannot be assembled quickly.
This illusion — that India's IT services success represented genuine technological capability — may have been the single most important factor in delaying India's recognition of its semiconductor deficit. If Indian companies were managing the IT systems of every major American bank and corporation, surely India was a technology leader? The answer, of course, is no — and the distinction between technology consumption and technology creation is the core of India's current challenge.
PART TWO: THE TECHNICAL LANDSCAPE — UNDERSTANDING SEMICONDUCTORS, DRAM, AND AI
The Semiconductor — The Atom of the Digital Age
To understand what India lacks, one must first understand what semiconductors are, why they matter, and why making them is so extraordinarily difficult. The semiconductor is the foundational component of the modern economy — the physical substrate upon which the digital world is built. Every smartphone, every laptop, every data center server, every electric vehicle, every medical device, every modern weapons system runs on semiconductors. The global semiconductor market was valued at approximately $574 billion in 2022 and is projected to exceed $1 trillion by 2030.
A semiconductor is, at its most basic, a material — typically silicon — whose electrical conductivity lies between that of a conductor (like copper) and an insulator (like rubber). By doping silicon with specific impurities in precisely controlled quantities and patterns, engineers can create regions of different electrical behavior that can function as switches, amplifiers, and logic gates. The integrated circuit packs millions, billions, and now trillions of these switches onto a single chip of silicon the size of a fingernail.
The journey from sand (silicon dioxide, SiO₂) to a finished semiconductor chip involves hundreds of distinct manufacturing steps, each requiring extraordinary precision. The process begins with the production of ultra-pure silicon — so pure that impurities must be measured in parts per trillion. Silicon ingots are grown through the Czochralski process, sliced into wafers, polished to atomic-scale flatness, and then subjected to photolithography — the process by which circuit patterns are etched onto the silicon surface using light.
Modern photolithography uses Extreme Ultraviolet (EUV) light — light with a wavelength of 13.5 nanometers — to print circuit features smaller than the wavelength of visible light. This requires the most sophisticated optical systems ever built — mirrors polished to tolerances measured in fractions of an atom, plasma-generated light sources, and vacuum chambers where even a single dust particle would be catastrophic. The EUV lithography machines made by ASML of the Netherlands are arguably the most complex manufactured objects in human history, containing over 100,000 components, weighing 180 tonnes, and costing over $150 million each.
The most advanced chips today — produced by Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung — are manufactured on 3-nanometer and 2-nanometer process nodes, meaning the smallest features are approximately 3-2 times the width of a DNA strand. A single chip may contain over 100 billion transistors. The yield management — ensuring that chips function correctly — requires statistical process control of hundreds of variables simultaneously.
This manufacturing capability has taken decades to build and represents the accumulated learning of thousands of engineers over fifty years. The knowledge embedded in a modern semiconductor fab is not in any manual or textbook — it is in the tacit knowledge of experienced engineers, in the institutional memory of manufacturing teams, in the calibration data of specific machines, in the understanding of subtle interactions between process variables that only reveals itself through years of practice.
This is why semiconductor manufacturing is often described as the hardest thing humans have ever built at scale. And it is why the barriers to entry are so formidable that, despite semiconductors being enormously profitable, fewer than a dozen companies in the world are capable of manufacturing them at the frontier, and only three — TSMC, Samsung, and Intel — are at the absolute leading edge.
The Semiconductor Value Chain — Where India Is and Is Not
The semiconductor industry is organized into a complex, globally distributed value chain, with different countries dominating different segments. Understanding this value chain is essential to understanding where India fits (and does not fit) today.
Electronic Design Automation (EDA) Tools: Before a chip can be manufactured, it must be designed — and modern chip designs are impossibly complex for human engineers to handle without sophisticated software tools. EDA tools — software that helps engineers design, simulate, verify, and test chip designs — are dominated by three American companies: Synopsys, Cadence, and Mentor Graphics (owned by Siemens). These companies have near-monopolistic positions in the software that enables semiconductor design. Without access to EDA tools, chip design is effectively impossible.
Intellectual Property (IP) Cores: Chip designers frequently incorporate pre-designed functional blocks — called IP cores — into their chips rather than designing every component from scratch. The most important IP cores include processor architectures (ARM Holdings of the UK dominates mobile processor design, with its architecture licensed by virtually every smartphone chip), memory interfaces, USB controllers, and GPU compute blocks. ARM's architecture is the foundation of virtually every mobile processor on the planet — and ARM was acquired by SoftBank in 2016, then Nvidia attempted to acquire it (blocked by regulators), and it subsequently went public. India has essentially no significant IP core providers.
Chip Design (Fabless): Companies that design chips but do not manufacture them — "fabless" companies — include the world's most valuable semiconductor firms: Nvidia (graphics processing units and AI accelerators), Qualcomm (mobile processor chips), AMD (processors and GPUs), Apple (custom chips for iPhones and Macs), Broadcom, MediaTek, and many others. The United States and Taiwan dominate fabless chip design. India has a growing chip design community, primarily through the captive design centers of global companies (Intel, Qualcomm, Texas Instruments, and others have large design teams in Hyderabad, Bangalore, and Pune), but essentially no significant Indian-owned fabless chip companies of global relevance.
Foundry Manufacturing: Companies that manufacture chips on behalf of fabless designers — "pure play foundries" — are dominated almost entirely by TSMC (Taiwan, ~55% global market share), Samsung Foundry (South Korea, ~17%), and GlobalFoundries (United States, ~7%). Intel is now trying to re-enter the foundry business. India has essentially zero presence in this segment.
Integrated Device Manufacturers (IDMs): Companies that both design and manufacture their own chips — IDMs — include Intel, Samsung, Texas Instruments, STMicroelectronics, and Infineon. Again, India has essentially no presence.
DRAM and NAND Flash Memory: Memory chips — the DRAM used in computer RAM and the NAND flash used in SSDs and USB drives — are an entirely separate segment of the semiconductor industry, dominated by three companies: Samsung (South Korea), SK Hynix (South Korea), and Micron (United States) in DRAM; and Samsung, SK Hynix, Micron, Kioxia (Japan), and Western Digital in NAND flash. India has essentially zero presence in semiconductor memory manufacturing.
Semiconductor Equipment: The equipment used to manufacture chips — lithography machines, etch systems, deposition systems, inspection tools — represents a separate, crucial segment dominated by a handful of Western and Japanese companies. ASML (Netherlands) makes virtually all advanced lithography machines. Applied Materials, Lam Research, and KLA (all American) dominate deposition, etch, and inspection equipment. Tokyo Electron (Japan) is another major player. India has essentially no semiconductor equipment manufacturing capacity.
Semiconductor Materials: The materials used in chip manufacturing — ultra-pure silicon, photoresists, specialty gases, chemicals — represent another critical segment. Key suppliers include Shin-Etsu Chemical and Sumco (Japan) for silicon wafers, and JSR and TOK (Japan) for photoresists. Japan's dominance in semiconductor materials — representing roughly 50-60% of global supply in some critical materials — is a strategic factor that has come into sharp focus during recent geopolitical tensions. India has very limited presence in semiconductor materials supply.
Advanced Packaging: After chips are manufactured, they must be packaged — placed in housing that protects the chip and provides electrical connections to the circuit board. Advanced packaging has become increasingly important as traditional scaling approaches the physical limits of Moore's Law, with techniques like 2.5D and 3D packaging (chiplets) becoming critical for high-performance computing. Advanced packaging is dominated by TSMC's CoWoS technology, Samsung, and specialized companies in Taiwan, South Korea, and Malaysia. India has limited but growing capacity in this segment.
Outsourced Semiconductor Assembly and Test (OSAT): The lower-cost end of semiconductor packaging and testing is performed by OSAT companies, largely based in Malaysia, Thailand, Vietnam, and China, with some in the Philippines. India has minimal OSAT capacity but is beginning to attract some investment in this segment.
The picture is stark: India is absent or marginally present in virtually every segment of the semiconductor value chain. This is not a matter of a few missing capabilities; it is a comprehensive absence from an industry that is central to the global economy.
Chapter 8: Dynamic Random Access Memory — The Working Memory of Civilization
DRAM — Dynamic Random Access Memory — deserves special attention because it represents a specific, critical technology where India's absence is particularly acute and particularly costly.
DRAM is the short-term memory of computing systems — the space where actively used data and instructions are stored for immediate access by the processor. Every time you open a browser tab, run a program, or edit a document, that activity takes place in DRAM. Every server in every data center uses DRAM. Every smartphone has DRAM. Every automotive system controller has DRAM. The global DRAM market was approximately $90 billion in 2022, representing about 16% of the total semiconductor market.
The physics of DRAM is elegant: each memory cell consists of a single transistor and a single capacitor. The capacitor stores a charge (representing a binary "1") or no charge (representing "0"). The transistor acts as a switch to read or write the charge. Because capacitors leak charge over time, DRAM cells must be "refreshed" periodically — typically thousands of times per second — hence the term "dynamic." This constant refreshing requires power, making DRAM one of the largest consumers of power in data centers.
The manufacturing of DRAM is extraordinarily demanding, for several reasons that are specific to this technology. DRAM cells must be packed as densely as possible to maximize capacity while minimizing cost, which requires some of the smallest feature sizes in semiconductor manufacturing. But unlike logic chips (processors and GPUs), where transistors can be arranged in complex two-dimensional patterns with many different sizes, DRAM cells must be arranged in regular arrays with extremely tight uniformity requirements. A single DRAM chip may contain billions of cells, each of which must function correctly — a single defective cell can render the entire chip non-functional or require complex error correction.
DRAM manufacturing involves several unique process steps, including buried word line (bWL) technology and high-aspect-ratio capacitor etching. Modern DRAM capacitors must store charge while taking up almost no horizontal space, so they are built vertically — etching deep, narrow holes in the silicon and filling them with dielectric and electrode materials. Achieving the right capacitance in these structures while maintaining yield requires extraordinary process control.
The DRAM industry is characterized by extreme capital intensity and very few players. Building a new DRAM fab requires investment of $10-15 billion or more. The technology requires decades of accumulated expertise. And the market is dominated by Samsung, SK Hynix, and Micron in an oligopoly that has survived multiple industry cycles precisely because the barriers to entry are so high that no new competitor has successfully entered the DRAM market in over two decades.
The most recent entrant attempting to crack the DRAM oligopoly is China's ChangXin Memory Technologies (CXMT), which is producing LPDDR4 DRAM for the domestic market, though still significantly behind the frontier. Even with enormous state subsidies and aggressive talent recruitment, China has found DRAM manufacturing extraordinarily difficult to catch up in. India has not even attempted to build DRAM manufacturing capacity.
Why does this matter for India specifically? Because the demand for DRAM in India is substantial and growing exponentially. Every smartphone sold in India contains DRAM. India is now the world's second-largest smartphone market, with over 600 million smartphone users. Every data center being built to support India's growing digital economy uses DRAM. The Indian government's data center buildout, driven by cloud computing demand, UPI transaction processing, and AI applications, requires massive quantities of DRAM. India imports essentially all of this DRAM — spending billions of dollars annually on what is, in the global supply chain, a commodity controlled by three companies in three countries (South Korea, South Korea, and the United States).
The geopolitical vulnerability this creates is severe. If relations with South Korea or the United States were to deteriorate, or if a conflict in Taiwan were to disrupt global semiconductor supply chains, India's entire digital economy could be starved of the memory chips it requires to function. This is not a theoretical risk — the COVID-19 pandemic demonstrated exactly how semiconductor supply chain disruptions can bring entire industries (automotive, consumer electronics, industrial equipment) to a halt.
The semiconductor industry is organized into a complex, globally distributed value chain, with different countries dominating different segments. Understanding this value chain is essential to understanding where India fits (and does not fit) today.
Electronic Design Automation (EDA) Tools: Before a chip can be manufactured, it must be designed — and modern chip designs are impossibly complex for human engineers to handle without sophisticated software tools. EDA tools — software that helps engineers design, simulate, verify, and test chip designs — are dominated by three American companies: Synopsys, Cadence, and Mentor Graphics (owned by Siemens). These companies have near-monopolistic positions in the software that enables semiconductor design. Without access to EDA tools, chip design is effectively impossible.
Intellectual Property (IP) Cores: Chip designers frequently incorporate pre-designed functional blocks — called IP cores — into their chips rather than designing every component from scratch. The most important IP cores include processor architectures (ARM Holdings of the UK dominates mobile processor design, with its architecture licensed by virtually every smartphone chip), memory interfaces, USB controllers, and GPU compute blocks. ARM's architecture is the foundation of virtually every mobile processor on the planet — and ARM was acquired by SoftBank in 2016, then Nvidia attempted to acquire it (blocked by regulators), and it subsequently went public. India has essentially no significant IP core providers.
Chip Design (Fabless): Companies that design chips but do not manufacture them — "fabless" companies — include the world's most valuable semiconductor firms: Nvidia (graphics processing units and AI accelerators), Qualcomm (mobile processor chips), AMD (processors and GPUs), Apple (custom chips for iPhones and Macs), Broadcom, MediaTek, and many others. The United States and Taiwan dominate fabless chip design. India has a growing chip design community, primarily through the captive design centers of global companies (Intel, Qualcomm, Texas Instruments, and others have large design teams in Hyderabad, Bangalore, and Pune), but essentially no significant Indian-owned fabless chip companies of global relevance.
Foundry Manufacturing: Companies that manufacture chips on behalf of fabless designers — "pure play foundries" — are dominated almost entirely by TSMC (Taiwan, ~55% global market share), Samsung Foundry (South Korea, ~17%), and GlobalFoundries (United States, ~7%). Intel is now trying to re-enter the foundry business. India has essentially zero presence in this segment.
Integrated Device Manufacturers (IDMs): Companies that both design and manufacture their own chips — IDMs — include Intel, Samsung, Texas Instruments, STMicroelectronics, and Infineon. Again, India has essentially no presence.
DRAM and NAND Flash Memory: Memory chips — the DRAM used in computer RAM and the NAND flash used in SSDs and USB drives — are an entirely separate segment of the semiconductor industry, dominated by three companies: Samsung (South Korea), SK Hynix (South Korea), and Micron (United States) in DRAM; and Samsung, SK Hynix, Micron, Kioxia (Japan), and Western Digital in NAND flash. India has essentially zero presence in semiconductor memory manufacturing.
Semiconductor Equipment: The equipment used to manufacture chips — lithography machines, etch systems, deposition systems, inspection tools — represents a separate, crucial segment dominated by a handful of Western and Japanese companies. ASML (Netherlands) makes virtually all advanced lithography machines. Applied Materials, Lam Research, and KLA (all American) dominate deposition, etch, and inspection equipment. Tokyo Electron (Japan) is another major player. India has essentially no semiconductor equipment manufacturing capacity.
Semiconductor Materials: The materials used in chip manufacturing — ultra-pure silicon, photoresists, specialty gases, chemicals — represent another critical segment. Key suppliers include Shin-Etsu Chemical and Sumco (Japan) for silicon wafers, and JSR and TOK (Japan) for photoresists. Japan's dominance in semiconductor materials — representing roughly 50-60% of global supply in some critical materials — is a strategic factor that has come into sharp focus during recent geopolitical tensions. India has very limited presence in semiconductor materials supply.
Advanced Packaging: After chips are manufactured, they must be packaged — placed in housing that protects the chip and provides electrical connections to the circuit board. Advanced packaging has become increasingly important as traditional scaling approaches the physical limits of Moore's Law, with techniques like 2.5D and 3D packaging (chiplets) becoming critical for high-performance computing. Advanced packaging is dominated by TSMC's CoWoS technology, Samsung, and specialized companies in Taiwan, South Korea, and Malaysia. India has limited but growing capacity in this segment.
Outsourced Semiconductor Assembly and Test (OSAT): The lower-cost end of semiconductor packaging and testing is performed by OSAT companies, largely based in Malaysia, Thailand, Vietnam, and China, with some in the Philippines. India has minimal OSAT capacity but is beginning to attract some investment in this segment.
The picture is stark: India is absent or marginally present in virtually every segment of the semiconductor value chain. This is not a matter of a few missing capabilities; it is a comprehensive absence from an industry that is central to the global economy.
Chapter 8: Dynamic Random Access Memory — The Working Memory of Civilization
DRAM — Dynamic Random Access Memory — deserves special attention because it represents a specific, critical technology where India's absence is particularly acute and particularly costly.
DRAM is the short-term memory of computing systems — the space where actively used data and instructions are stored for immediate access by the processor. Every time you open a browser tab, run a program, or edit a document, that activity takes place in DRAM. Every server in every data center uses DRAM. Every smartphone has DRAM. Every automotive system controller has DRAM. The global DRAM market was approximately $90 billion in 2022, representing about 16% of the total semiconductor market.
The physics of DRAM is elegant: each memory cell consists of a single transistor and a single capacitor. The capacitor stores a charge (representing a binary "1") or no charge (representing "0"). The transistor acts as a switch to read or write the charge. Because capacitors leak charge over time, DRAM cells must be "refreshed" periodically — typically thousands of times per second — hence the term "dynamic." This constant refreshing requires power, making DRAM one of the largest consumers of power in data centers.
The manufacturing of DRAM is extraordinarily demanding, for several reasons that are specific to this technology. DRAM cells must be packed as densely as possible to maximize capacity while minimizing cost, which requires some of the smallest feature sizes in semiconductor manufacturing. But unlike logic chips (processors and GPUs), where transistors can be arranged in complex two-dimensional patterns with many different sizes, DRAM cells must be arranged in regular arrays with extremely tight uniformity requirements. A single DRAM chip may contain billions of cells, each of which must function correctly — a single defective cell can render the entire chip non-functional or require complex error correction.
DRAM manufacturing involves several unique process steps, including buried word line (bWL) technology and high-aspect-ratio capacitor etching. Modern DRAM capacitors must store charge while taking up almost no horizontal space, so they are built vertically — etching deep, narrow holes in the silicon and filling them with dielectric and electrode materials. Achieving the right capacitance in these structures while maintaining yield requires extraordinary process control.
The DRAM industry is characterized by extreme capital intensity and very few players. Building a new DRAM fab requires investment of $10-15 billion or more. The technology requires decades of accumulated expertise. And the market is dominated by Samsung, SK Hynix, and Micron in an oligopoly that has survived multiple industry cycles precisely because the barriers to entry are so high that no new competitor has successfully entered the DRAM market in over two decades.
The most recent entrant attempting to crack the DRAM oligopoly is China's ChangXin Memory Technologies (CXMT), which is producing LPDDR4 DRAM for the domestic market, though still significantly behind the frontier. Even with enormous state subsidies and aggressive talent recruitment, China has found DRAM manufacturing extraordinarily difficult to catch up in. India has not even attempted to build DRAM manufacturing capacity.
Why does this matter for India specifically? Because the demand for DRAM in India is substantial and growing exponentially. Every smartphone sold in India contains DRAM. India is now the world's second-largest smartphone market, with over 600 million smartphone users. Every data center being built to support India's growing digital economy uses DRAM. The Indian government's data center buildout, driven by cloud computing demand, UPI transaction processing, and AI applications, requires massive quantities of DRAM. India imports essentially all of this DRAM — spending billions of dollars annually on what is, in the global supply chain, a commodity controlled by three companies in three countries (South Korea, South Korea, and the United States).
The geopolitical vulnerability this creates is severe. If relations with South Korea or the United States were to deteriorate, or if a conflict in Taiwan were to disrupt global semiconductor supply chains, India's entire digital economy could be starved of the memory chips it requires to function. This is not a theoretical risk — the COVID-19 pandemic demonstrated exactly how semiconductor supply chain disruptions can bring entire industries (automotive, consumer electronics, industrial equipment) to a halt.
Artificial Intelligence — The Apex Technology of the 21st Century
If semiconductors are the physical substrate of the digital age, artificial intelligence — and specifically the large language models and foundation models of the current era — represent the apex application layer. Understanding why AI matters, why it is so resource-intensive, and why India's position in AI is so precarious requires a brief technical history.
The history of artificial intelligence as a field goes back to the 1950s, with Alan Turing's famous question "Can machines think?" and the Dartmouth Conference of 1956, which is generally considered the founding moment of AI as an academic discipline. The subsequent decades saw alternating periods of optimism and "AI winters" — periods when progress stalled and funding dried up, as the difficulty of replicating human intelligence became apparent.
The modern AI era — characterized by deep learning, neural networks, and massive-scale foundation models — began in earnest with a paper published in 2012: "ImageNet Classification with Deep Convolutional Neural Networks" by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton at the University of Toronto. Their deep convolutional neural network — AlexNet — achieved a dramatic reduction in error rate on the ImageNet image recognition benchmark, demonstrating that neural networks trained on large amounts of data using Graphics Processing Units (GPUs) could surpass hand-engineered feature-based approaches.
The key insight — that raw computational power applied to large datasets, using neural network architectures, could achieve performance that no explicit programming approach could match — unleashed a technological revolution. Over the next decade, GPUs were repurposed from gaming to AI training, datasets grew to internet scale, and neural network architectures became increasingly sophisticated.
The transformer architecture, introduced in Google's 2017 paper "Attention Is All You Need," became the foundational architecture for the current generation of large language models. Unlike previous sequence modeling approaches, transformers use a "self-attention" mechanism that allows each element of a sequence to attend to every other element simultaneously, enabling parallel processing and capturing long-range dependencies in text. This made transformers both more powerful and more efficiently trainable than predecessors.
The scaling laws that govern transformer training — established empirically through research at OpenAI and other labs — showed that model performance improved predictably as a function of three variables: the number of model parameters, the quantity of training data, and the compute budget. Larger models, trained on more data with more compute, were systematically better. This created a dynamic where the firms that could afford the most compute — running training runs that cost tens or hundreds of millions of dollars — achieved the most capable models.
OpenAI's GPT series traced this scaling trajectory dramatically: GPT-2 (2019, 1.5 billion parameters) → GPT-3 (2020, 175 billion parameters) → GPT-4 (2023, estimated 1.7 trillion parameters in a mixture-of-experts architecture). Google developed the BERT, T5, PaLM, and Gemini families. Meta released the LLaMA family (open weights). Anthropic developed the Claude series. Each generation brought qualitatively new capabilities — from coherent paragraph generation to multi-step reasoning, code generation, mathematical problem-solving, and multimodal understanding of images, audio, and video.
The training of these frontier models requires extraordinary computational resources. GPT-4 training is estimated to have required approximately $50-100 million of compute, running on tens of thousands of Nvidia A100 GPUs. The inference (serving model outputs to users) requires comparable scale. Nvidia's H100 GPU, released in 2022, became the critical resource for AI training, with prices exceeding $30,000 per unit and demand dramatically outstripping supply. The AI data center buildout — driven by Microsoft, Google, Amazon, Meta, and a constellation of AI startups — has become one of the largest infrastructure investments in human history.
The geopolitical dimension of AI is acute. The United States and China are engaged in a fierce competition for AI leadership, with enormous national security implications. AI systems are being applied to autonomous weapons, signals intelligence, strategic decision-making, and cyber operations. The nation with the most capable AI systems will have profound military advantages. The United States has used export controls — specifically the restrictions on Nvidia's advanced GPU exports to China — as a tool to slow China's AI development. This represents the explicit weaponization of technology supply chains in the service of geopolitical competition.
Where does India stand in this AI landscape? The picture is complex but ultimately concerning. India has enormous talent — hundreds of thousands of software engineers, a growing ML research community, and several world-class AI researchers (many of them at American universities and companies). But India lacks the foundational infrastructure required to compete at the frontier of AI development:
India has no large-scale domestic AI training infrastructure — no significant cluster of advanced AI GPUs operating at the scale required to train frontier models. The National Supercomputing Mission is building capacity, but at a level that is orders of magnitude below what is required for frontier model training.
India has no homegrown frontier AI foundation model. The models developed by Indian startups and institutions — including those developed by companies like Krutrim (founded by Ola's Bhavish Aggarwal), Sarvam AI, and various academic efforts — are significant and promising, but they are not competitive with GPT-4, Claude, or Gemini in general capability. They are typically based on fine-tuning or adaptation of open-source base models (LLaMA, Mistral) rather than training from scratch on proprietary data at frontier scale.
India lacks the domestic chip design and manufacturing capacity to produce the AI accelerators required for frontier training or inference at scale. India's AI ambitions are therefore entirely dependent on Nvidia GPUs, AMD GPUs, or cloud AI services — all of which are controlled by American companies and subject to US export control regulations.
PART THREE: UNDERSTANDING THE CONTEMPORARY DEFICIT — A COMPREHENSIVE DIAGNOSIS
If semiconductors are the physical substrate of the digital age, artificial intelligence — and specifically the large language models and foundation models of the current era — represent the apex application layer. Understanding why AI matters, why it is so resource-intensive, and why India's position in AI is so precarious requires a brief technical history.
The history of artificial intelligence as a field goes back to the 1950s, with Alan Turing's famous question "Can machines think?" and the Dartmouth Conference of 1956, which is generally considered the founding moment of AI as an academic discipline. The subsequent decades saw alternating periods of optimism and "AI winters" — periods when progress stalled and funding dried up, as the difficulty of replicating human intelligence became apparent.
The modern AI era — characterized by deep learning, neural networks, and massive-scale foundation models — began in earnest with a paper published in 2012: "ImageNet Classification with Deep Convolutional Neural Networks" by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton at the University of Toronto. Their deep convolutional neural network — AlexNet — achieved a dramatic reduction in error rate on the ImageNet image recognition benchmark, demonstrating that neural networks trained on large amounts of data using Graphics Processing Units (GPUs) could surpass hand-engineered feature-based approaches.
The key insight — that raw computational power applied to large datasets, using neural network architectures, could achieve performance that no explicit programming approach could match — unleashed a technological revolution. Over the next decade, GPUs were repurposed from gaming to AI training, datasets grew to internet scale, and neural network architectures became increasingly sophisticated.
The transformer architecture, introduced in Google's 2017 paper "Attention Is All You Need," became the foundational architecture for the current generation of large language models. Unlike previous sequence modeling approaches, transformers use a "self-attention" mechanism that allows each element of a sequence to attend to every other element simultaneously, enabling parallel processing and capturing long-range dependencies in text. This made transformers both more powerful and more efficiently trainable than predecessors.
The scaling laws that govern transformer training — established empirically through research at OpenAI and other labs — showed that model performance improved predictably as a function of three variables: the number of model parameters, the quantity of training data, and the compute budget. Larger models, trained on more data with more compute, were systematically better. This created a dynamic where the firms that could afford the most compute — running training runs that cost tens or hundreds of millions of dollars — achieved the most capable models.
OpenAI's GPT series traced this scaling trajectory dramatically: GPT-2 (2019, 1.5 billion parameters) → GPT-3 (2020, 175 billion parameters) → GPT-4 (2023, estimated 1.7 trillion parameters in a mixture-of-experts architecture). Google developed the BERT, T5, PaLM, and Gemini families. Meta released the LLaMA family (open weights). Anthropic developed the Claude series. Each generation brought qualitatively new capabilities — from coherent paragraph generation to multi-step reasoning, code generation, mathematical problem-solving, and multimodal understanding of images, audio, and video.
The training of these frontier models requires extraordinary computational resources. GPT-4 training is estimated to have required approximately $50-100 million of compute, running on tens of thousands of Nvidia A100 GPUs. The inference (serving model outputs to users) requires comparable scale. Nvidia's H100 GPU, released in 2022, became the critical resource for AI training, with prices exceeding $30,000 per unit and demand dramatically outstripping supply. The AI data center buildout — driven by Microsoft, Google, Amazon, Meta, and a constellation of AI startups — has become one of the largest infrastructure investments in human history.
The geopolitical dimension of AI is acute. The United States and China are engaged in a fierce competition for AI leadership, with enormous national security implications. AI systems are being applied to autonomous weapons, signals intelligence, strategic decision-making, and cyber operations. The nation with the most capable AI systems will have profound military advantages. The United States has used export controls — specifically the restrictions on Nvidia's advanced GPU exports to China — as a tool to slow China's AI development. This represents the explicit weaponization of technology supply chains in the service of geopolitical competition.
Where does India stand in this AI landscape? The picture is complex but ultimately concerning. India has enormous talent — hundreds of thousands of software engineers, a growing ML research community, and several world-class AI researchers (many of them at American universities and companies). But India lacks the foundational infrastructure required to compete at the frontier of AI development:
India has no large-scale domestic AI training infrastructure — no significant cluster of advanced AI GPUs operating at the scale required to train frontier models. The National Supercomputing Mission is building capacity, but at a level that is orders of magnitude below what is required for frontier model training.
India has no homegrown frontier AI foundation model. The models developed by Indian startups and institutions — including those developed by companies like Krutrim (founded by Ola's Bhavish Aggarwal), Sarvam AI, and various academic efforts — are significant and promising, but they are not competitive with GPT-4, Claude, or Gemini in general capability. They are typically based on fine-tuning or adaptation of open-source base models (LLaMA, Mistral) rather than training from scratch on proprietary data at frontier scale.
India lacks the domestic chip design and manufacturing capacity to produce the AI accelerators required for frontier training or inference at scale. India's AI ambitions are therefore entirely dependent on Nvidia GPUs, AMD GPUs, or cloud AI services — all of which are controlled by American companies and subject to US export control regulations.
PART THREE: UNDERSTANDING THE CONTEMPORARY DEFICIT — A COMPREHENSIVE DIAGNOSIS
The Semiconductor Policy Failures — Specific and Identifiable
India's semiconductor deficit is not a mystery or an inevitability. It is the result of specific, identifiable policy choices — and, equally, policy omissions — over a period of several decades. Tracing these specific failures is important not merely for historical accountability but because understanding what went wrong is the necessary first step to understanding how to go right.
The 1988 Semiconductor Policy That Wasn't: In 1988, the Indian government's Department of Electronics formulated an ambitious semiconductor policy that called for the establishment of wafer fabrication facilities in India, with significant state investment and technology transfer from foreign partners. The Texas Instruments deal to set up a design center in Bangalore in 1985 had demonstrated that foreign technology companies would invest in India. But the semiconductor manufacturing ambitions of 1988 foundered on a combination of factors: insufficient budgetary allocation, the complexity of technology transfer negotiations with potential foreign partners, the bureaucratic inertia of the License Raj, and a fundamental underestimation of the capital intensity and technical difficulty of semiconductor manufacturing. The policy was announced, partially implemented in terms of design infrastructure, and then effectively abandoned.
The 1994-2000 Missed Window: In the 1990s, as India was liberalizing its economy and the global semiconductor industry was expanding dramatically, there was a genuine window to establish a semiconductor manufacturing base. South Korea's Samsung was still a mid-tier player; TSMC had been established only since 1987; the technology gaps between frontier and slightly-behind-the-frontier processes were smaller than they are today. India had just undergone economic liberalization, foreign direct investment was being welcomed, and there was genuine enthusiasm for technology investment.
Instead, the 1990s saw India's technology sector pivot entirely toward IT services. The incentives were irresistible: software services required relatively low capital investment, could be scaled quickly by hiring more engineers, generated immediate export revenues, and created highly visible success stories that validated the IT services path. The harder, longer, more capital-intensive work of building semiconductor manufacturing was not attempted.
The 2006-2014 Period — Plans Without Execution: In 2006, the UPA government's Department of Information Technology formulated a "Semiconductor Wafer Fabrication" policy that identified semiconductor manufacturing as a strategic priority and outlined incentives to attract investment. The policy offered capital subsidy of 20% for units established outside of special economic zones and recognized the strategic importance of domestic semiconductor capability. Proposals were received from several parties, including a consortium involving ST Microelectronics and the Hindustan Semiconductors Limited project. But none of these materialized. The proposals were beset by disagreements over subsidy levels, concerns about technology transfer, skepticism about commercial viability, and bureaucratic delays that stretched for years.
The 2007-2008 global financial crisis further dampened investment sentiment. By the time economic conditions recovered, the political environment had changed, and momentum was lost. A second semiconductor policy in 2012 attempted to revive the ambition, but again without decisive executive commitment or adequate funding, it produced little result.
The Land and Infrastructure Gap: Unlike the advanced semiconductor nations — Taiwan, South Korea, the United States — India did not build the industrial infrastructure that semiconductor manufacturing requires: guaranteed, uninterrupted power supply (semiconductor fabs require 24/7 power without even momentary interruptions); ultra-pure water supply in large volumes (a semiconductor fab uses 2-5 million gallons of ultra-pure water per day); sophisticated waste treatment infrastructure; efficient logistics connectivity to import materials and export chips; and proximity to a deep, multi-generational ecosystem of materials suppliers, equipment service providers, and specialized engineering firms. Establishing a semiconductor fab in India means building all of this from scratch — at enormous additional cost compared to locations where this infrastructure already exists.
The Education-Industry Misalignment: India's engineering education system, while producing large numbers of graduates, has not produced graduates with the specific skills required for semiconductor process engineering, VLSI design at advanced nodes, or semiconductor equipment engineering in sufficient numbers. The IITs have excellent electrical engineering and computer science programs, but specialized semiconductor process engineering — the skills needed to actually run a fab — requires a very specific curriculum and hands-on laboratory experience that very few Indian institutions provide.
The talent pipeline for semiconductor manufacturing is quite different from the talent pipeline for software development. Semiconductor process engineers need deep knowledge of physics, chemistry, materials science, and manufacturing process technology. They need to understand plasma physics (for plasma etch and deposition processes), photochemistry (for photolithography), and the thermodynamics of thin film deposition. This is a different and more specialized skill set than general software engineering, and India has not systematically developed it.
India's semiconductor deficit is not a mystery or an inevitability. It is the result of specific, identifiable policy choices — and, equally, policy omissions — over a period of several decades. Tracing these specific failures is important not merely for historical accountability but because understanding what went wrong is the necessary first step to understanding how to go right.
The 1988 Semiconductor Policy That Wasn't: In 1988, the Indian government's Department of Electronics formulated an ambitious semiconductor policy that called for the establishment of wafer fabrication facilities in India, with significant state investment and technology transfer from foreign partners. The Texas Instruments deal to set up a design center in Bangalore in 1985 had demonstrated that foreign technology companies would invest in India. But the semiconductor manufacturing ambitions of 1988 foundered on a combination of factors: insufficient budgetary allocation, the complexity of technology transfer negotiations with potential foreign partners, the bureaucratic inertia of the License Raj, and a fundamental underestimation of the capital intensity and technical difficulty of semiconductor manufacturing. The policy was announced, partially implemented in terms of design infrastructure, and then effectively abandoned.
The 1994-2000 Missed Window: In the 1990s, as India was liberalizing its economy and the global semiconductor industry was expanding dramatically, there was a genuine window to establish a semiconductor manufacturing base. South Korea's Samsung was still a mid-tier player; TSMC had been established only since 1987; the technology gaps between frontier and slightly-behind-the-frontier processes were smaller than they are today. India had just undergone economic liberalization, foreign direct investment was being welcomed, and there was genuine enthusiasm for technology investment.
Instead, the 1990s saw India's technology sector pivot entirely toward IT services. The incentives were irresistible: software services required relatively low capital investment, could be scaled quickly by hiring more engineers, generated immediate export revenues, and created highly visible success stories that validated the IT services path. The harder, longer, more capital-intensive work of building semiconductor manufacturing was not attempted.
The 2006-2014 Period — Plans Without Execution: In 2006, the UPA government's Department of Information Technology formulated a "Semiconductor Wafer Fabrication" policy that identified semiconductor manufacturing as a strategic priority and outlined incentives to attract investment. The policy offered capital subsidy of 20% for units established outside of special economic zones and recognized the strategic importance of domestic semiconductor capability. Proposals were received from several parties, including a consortium involving ST Microelectronics and the Hindustan Semiconductors Limited project. But none of these materialized. The proposals were beset by disagreements over subsidy levels, concerns about technology transfer, skepticism about commercial viability, and bureaucratic delays that stretched for years.
The 2007-2008 global financial crisis further dampened investment sentiment. By the time economic conditions recovered, the political environment had changed, and momentum was lost. A second semiconductor policy in 2012 attempted to revive the ambition, but again without decisive executive commitment or adequate funding, it produced little result.
The Land and Infrastructure Gap: Unlike the advanced semiconductor nations — Taiwan, South Korea, the United States — India did not build the industrial infrastructure that semiconductor manufacturing requires: guaranteed, uninterrupted power supply (semiconductor fabs require 24/7 power without even momentary interruptions); ultra-pure water supply in large volumes (a semiconductor fab uses 2-5 million gallons of ultra-pure water per day); sophisticated waste treatment infrastructure; efficient logistics connectivity to import materials and export chips; and proximity to a deep, multi-generational ecosystem of materials suppliers, equipment service providers, and specialized engineering firms. Establishing a semiconductor fab in India means building all of this from scratch — at enormous additional cost compared to locations where this infrastructure already exists.
The Education-Industry Misalignment: India's engineering education system, while producing large numbers of graduates, has not produced graduates with the specific skills required for semiconductor process engineering, VLSI design at advanced nodes, or semiconductor equipment engineering in sufficient numbers. The IITs have excellent electrical engineering and computer science programs, but specialized semiconductor process engineering — the skills needed to actually run a fab — requires a very specific curriculum and hands-on laboratory experience that very few Indian institutions provide.
The talent pipeline for semiconductor manufacturing is quite different from the talent pipeline for software development. Semiconductor process engineers need deep knowledge of physics, chemistry, materials science, and manufacturing process technology. They need to understand plasma physics (for plasma etch and deposition processes), photochemistry (for photolithography), and the thermodynamics of thin film deposition. This is a different and more specialized skill set than general software engineering, and India has not systematically developed it.
The DRAM Deficit — Specific Consequences
The absence of any domestic DRAM manufacturing capacity has specific, measurable consequences for India across multiple dimensions.
Import Dependency and Balance of Payments: India's electronics imports — which include a large component of semiconductor and memory chips — were approximately $70 billion in fiscal year 2022-23, making electronics the largest import category after oil and gold. A significant portion of this is DRAM and NAND flash memory. This represents a continuous drain on foreign exchange reserves — money leaving the Indian economy to pay for components that could, in principle, be manufactured domestically.
Supply Chain Vulnerability: India's dependence on DRAM from Samsung and SK Hynix in South Korea (which together control approximately 70% of the global DRAM market) creates a specific geopolitical vulnerability. South Korea, while generally allied with the US and friendly to India, is not an unconstrained supplier — it operates under its own foreign policy constraints, and its supply capacity can be disrupted by factors including natural disasters (a fire at an SK Hynix fab in 2013 caused global DRAM prices to spike), manufacturing accidents, or labor disputes.
More acutely, the dominant position of TSMC and Samsung in advanced semiconductor manufacturing creates Taiwan Strait risk for India. If conflict in the Taiwan Strait were to disrupt TSMC's production, the consequences for India's electronics sector would be severe. Every mobile phone, laptop, server, and smart device sold in India uses TSMC-manufactured chips — Apple's A-series and M-series chips, Qualcomm's Snapdragon series, MediaTek's Dimensity series are all TSMC-fabricated.
Smartphone Market Exposure: India is the world's second-largest smartphone market, with over 1.4 billion mobile connections and approximately 600-700 million smartphone users. The smartphone market alone represents a massive annual dependency on imported DRAM. An entry-level smartphone today contains 3-4GB of DRAM; a flagship device may have 12-16GB. India is expected to become the world's second-largest smartphone manufacturer in terms of assembly volume — with Apple, Samsung, and Chinese brands all significantly expanding Indian assembly operations — but the DRAM inside every phone assembled in India is imported, representing a fundamental limit on how much value India captures from its growing role in electronics manufacturing.
Data Center and Cloud Infrastructure: India's burgeoning digital economy — including the extraordinary success of the Unified Payments Interface (UPI), the Aadhaar digital identity system, the DigiLocker platform, and the growing e-commerce and streaming sectors — is built on data center infrastructure that requires massive quantities of DRAM. India has become one of the world's fastest-growing data center markets, with significant new capacity being built by global hyperscalers (Amazon AWS, Microsoft Azure, Google Cloud) as well as domestic providers (NxtGen, CtrlS, Yotta, Nxtra).
Every server in every one of these data centers uses DRAM — typically 256GB to 768GB or more per server, in the form of DDR5 DIMMs. A large hyperscale data center may house tens of thousands of servers, representing enormous quantities of DRAM. All of this DRAM is imported.
AI Infrastructure Dependency: The specific DRAM variant used in AI training — High Bandwidth Memory (HBM) — is even more concentrated in supply. HBM is manufactured by only two companies in the world: SK Hynix (which has approximately 50% market share in HBM) and Samsung, with Micron recently entering the market. HBM is stacked in 3D configurations and placed directly adjacent to or on top of AI accelerator chips to provide the enormous memory bandwidth that AI training requires. Nvidia's H100 and H200 GPUs — the dominant AI training chips — use SK Hynix's HBM3 and HBM3E respectively. India's aspiration to build domestic AI capability thus runs directly into the wall of HBM dependency — and HBM is even more concentrated in supply than standard DRAM.
The absence of any domestic DRAM manufacturing capacity has specific, measurable consequences for India across multiple dimensions.
Import Dependency and Balance of Payments: India's electronics imports — which include a large component of semiconductor and memory chips — were approximately $70 billion in fiscal year 2022-23, making electronics the largest import category after oil and gold. A significant portion of this is DRAM and NAND flash memory. This represents a continuous drain on foreign exchange reserves — money leaving the Indian economy to pay for components that could, in principle, be manufactured domestically.
Supply Chain Vulnerability: India's dependence on DRAM from Samsung and SK Hynix in South Korea (which together control approximately 70% of the global DRAM market) creates a specific geopolitical vulnerability. South Korea, while generally allied with the US and friendly to India, is not an unconstrained supplier — it operates under its own foreign policy constraints, and its supply capacity can be disrupted by factors including natural disasters (a fire at an SK Hynix fab in 2013 caused global DRAM prices to spike), manufacturing accidents, or labor disputes.
More acutely, the dominant position of TSMC and Samsung in advanced semiconductor manufacturing creates Taiwan Strait risk for India. If conflict in the Taiwan Strait were to disrupt TSMC's production, the consequences for India's electronics sector would be severe. Every mobile phone, laptop, server, and smart device sold in India uses TSMC-manufactured chips — Apple's A-series and M-series chips, Qualcomm's Snapdragon series, MediaTek's Dimensity series are all TSMC-fabricated.
Smartphone Market Exposure: India is the world's second-largest smartphone market, with over 1.4 billion mobile connections and approximately 600-700 million smartphone users. The smartphone market alone represents a massive annual dependency on imported DRAM. An entry-level smartphone today contains 3-4GB of DRAM; a flagship device may have 12-16GB. India is expected to become the world's second-largest smartphone manufacturer in terms of assembly volume — with Apple, Samsung, and Chinese brands all significantly expanding Indian assembly operations — but the DRAM inside every phone assembled in India is imported, representing a fundamental limit on how much value India captures from its growing role in electronics manufacturing.
Data Center and Cloud Infrastructure: India's burgeoning digital economy — including the extraordinary success of the Unified Payments Interface (UPI), the Aadhaar digital identity system, the DigiLocker platform, and the growing e-commerce and streaming sectors — is built on data center infrastructure that requires massive quantities of DRAM. India has become one of the world's fastest-growing data center markets, with significant new capacity being built by global hyperscalers (Amazon AWS, Microsoft Azure, Google Cloud) as well as domestic providers (NxtGen, CtrlS, Yotta, Nxtra).
Every server in every one of these data centers uses DRAM — typically 256GB to 768GB or more per server, in the form of DDR5 DIMMs. A large hyperscale data center may house tens of thousands of servers, representing enormous quantities of DRAM. All of this DRAM is imported.
AI Infrastructure Dependency: The specific DRAM variant used in AI training — High Bandwidth Memory (HBM) — is even more concentrated in supply. HBM is manufactured by only two companies in the world: SK Hynix (which has approximately 50% market share in HBM) and Samsung, with Micron recently entering the market. HBM is stacked in 3D configurations and placed directly adjacent to or on top of AI accelerator chips to provide the enormous memory bandwidth that AI training requires. Nvidia's H100 and H200 GPUs — the dominant AI training chips — use SK Hynix's HBM3 and HBM3E respectively. India's aspiration to build domestic AI capability thus runs directly into the wall of HBM dependency — and HBM is even more concentrated in supply than standard DRAM.
The AI Model Gap — India's Specific Situation
India's position in the global AI landscape is characterized by a genuine paradox: the country has abundant AI talent (India is the second-largest country of origin for AI researchers at top global institutions and companies after China), growing domestic AI demand (India's digital economy generates enormous quantities of data), and stated government ambition for AI leadership — but lacks the foundational infrastructure to train frontier models.
The specific dimensions of India's AI gap deserve careful examination.
Compute Deficit: Training a frontier AI model — comparable to GPT-4, Claude, or Gemini — requires a cluster of thousands to tens of thousands of Nvidia H100 or H200 GPUs, operating continuously for weeks or months. The cost is measured in the tens of millions of dollars for a single training run. The key bottleneck is not money alone but access to the chips themselves. Nvidia's advanced AI GPUs are in extraordinarily tight supply globally — hyperscalers (Amazon, Microsoft, Google, Meta) have taken a large proportion of production, and export controls restrict sales to certain countries. India, while not explicitly restricted in the same way as China, competes for limited GPU supply in a global market where American cloud companies and AI startups have first priority.
The National Supercomputing Mission (NSM), launched in 2015 and updated in 2022, aims to build a distributed supercomputing infrastructure for India. The AI component of this involves AI-focused GPU clusters at several institutions. But the scale is inadequate for frontier model training: the combined AI compute infrastructure available to Indian academic institutions and public research bodies is orders of magnitude below what is required to train a frontier LLM. Private companies in India can purchase or lease GPU compute from global cloud providers (AWS, Azure, GCP) but at costs that make training frontier models economically very challenging for Indian startups.
Data Ecosystem: India's data advantage is real but underutilized. India is one of the world's largest internet markets, with over 800 million internet users generating enormous quantities of data in English, Hindi, and 20+ other Indian languages. High-quality Indian language data — the linguistic substrate required to train capable multilingual models — is a genuine asset. But collecting, cleaning, and curating this data at the scale required for frontier training requires significant investment and infrastructure that most Indian AI companies cannot afford.
Moreover, India's data governance framework — currently under development with the Digital Personal Data Protection Act of 2023 — creates uncertainty about how AI training data can be collected and used, which complicates the development of large-scale training datasets.
Research Ecosystem: India's AI research output — measured by publications in top venues like NeurIPS, ICML, ICLR, and ACL — has been growing rapidly, but remains well below the United States and China in absolute terms. More significantly, the research being done in India tends to be at the application layer — using existing models and methods for specific tasks — rather than at the foundational layer of developing new architectures, new training methods, or new theoretical understanding. The fundamental research required to push the frontier of AI capability is concentrated in a small number of US, UK, and Chinese institutions and companies with the resources to pursue it.
The IITs have been developing AI research programs, and several Indian companies run AI research labs. But the scale and ambition of Indian AI research — in terms of the scope of problems being attacked and the compute resources available — is not comparable to what DeepMind, OpenAI, Anthropic, Meta AI Research, or Google Brain are doing.
Indian Language AI: One area where India has specific opportunities is in the development of AI systems for Indian languages. India's linguistic diversity — 22 scheduled languages, hundreds of dialects, and enormous variation in script, grammar, and lexical structure — creates both a challenge and an opportunity. English-dominant models perform poorly on Indian languages, creating a genuine domestic need for better Indian language AI. Startups like Sarvam AI (working on Indic language models), AI4Bharat (an IIT Madras initiative developing open datasets and models for Indian languages), and several others are working in this space. But even the best Indian language models today rely on fine-tuning of globally developed base models rather than training from scratch.
PART FOUR: THE GLOBAL CONTEXT — GEOPOLITICS OF SEMICONDUCTORS AND AI
India's position in the global AI landscape is characterized by a genuine paradox: the country has abundant AI talent (India is the second-largest country of origin for AI researchers at top global institutions and companies after China), growing domestic AI demand (India's digital economy generates enormous quantities of data), and stated government ambition for AI leadership — but lacks the foundational infrastructure to train frontier models.
The specific dimensions of India's AI gap deserve careful examination.
Compute Deficit: Training a frontier AI model — comparable to GPT-4, Claude, or Gemini — requires a cluster of thousands to tens of thousands of Nvidia H100 or H200 GPUs, operating continuously for weeks or months. The cost is measured in the tens of millions of dollars for a single training run. The key bottleneck is not money alone but access to the chips themselves. Nvidia's advanced AI GPUs are in extraordinarily tight supply globally — hyperscalers (Amazon, Microsoft, Google, Meta) have taken a large proportion of production, and export controls restrict sales to certain countries. India, while not explicitly restricted in the same way as China, competes for limited GPU supply in a global market where American cloud companies and AI startups have first priority.
The National Supercomputing Mission (NSM), launched in 2015 and updated in 2022, aims to build a distributed supercomputing infrastructure for India. The AI component of this involves AI-focused GPU clusters at several institutions. But the scale is inadequate for frontier model training: the combined AI compute infrastructure available to Indian academic institutions and public research bodies is orders of magnitude below what is required to train a frontier LLM. Private companies in India can purchase or lease GPU compute from global cloud providers (AWS, Azure, GCP) but at costs that make training frontier models economically very challenging for Indian startups.
Data Ecosystem: India's data advantage is real but underutilized. India is one of the world's largest internet markets, with over 800 million internet users generating enormous quantities of data in English, Hindi, and 20+ other Indian languages. High-quality Indian language data — the linguistic substrate required to train capable multilingual models — is a genuine asset. But collecting, cleaning, and curating this data at the scale required for frontier training requires significant investment and infrastructure that most Indian AI companies cannot afford.
Moreover, India's data governance framework — currently under development with the Digital Personal Data Protection Act of 2023 — creates uncertainty about how AI training data can be collected and used, which complicates the development of large-scale training datasets.
Research Ecosystem: India's AI research output — measured by publications in top venues like NeurIPS, ICML, ICLR, and ACL — has been growing rapidly, but remains well below the United States and China in absolute terms. More significantly, the research being done in India tends to be at the application layer — using existing models and methods for specific tasks — rather than at the foundational layer of developing new architectures, new training methods, or new theoretical understanding. The fundamental research required to push the frontier of AI capability is concentrated in a small number of US, UK, and Chinese institutions and companies with the resources to pursue it.
The IITs have been developing AI research programs, and several Indian companies run AI research labs. But the scale and ambition of Indian AI research — in terms of the scope of problems being attacked and the compute resources available — is not comparable to what DeepMind, OpenAI, Anthropic, Meta AI Research, or Google Brain are doing.
Indian Language AI: One area where India has specific opportunities is in the development of AI systems for Indian languages. India's linguistic diversity — 22 scheduled languages, hundreds of dialects, and enormous variation in script, grammar, and lexical structure — creates both a challenge and an opportunity. English-dominant models perform poorly on Indian languages, creating a genuine domestic need for better Indian language AI. Startups like Sarvam AI (working on Indic language models), AI4Bharat (an IIT Madras initiative developing open datasets and models for Indian languages), and several others are working in this space. But even the best Indian language models today rely on fine-tuning of globally developed base models rather than training from scratch.
PART FOUR: THE GLOBAL CONTEXT — GEOPOLITICS OF SEMICONDUCTORS AND AI
The US-China Technology War and Its Implications for India
The escalating technology conflict between the United States and China is the most important geopolitical context for understanding India's semiconductor and AI challenge. This conflict — which began under the Trump administration and has dramatically intensified under Biden and continues to evolve — is reshaping the global technology supply chain in ways that create both risks and opportunities for India.
The US-China technology war in semiconductors began in earnest with the Trump administration's actions against Huawei in 2019. Huawei — at the time the world's largest telecommunications equipment company and a significant smartphone brand — was placed on the US Entity List, restricting American companies from supplying Huawei without government authorization. This was intended to prevent Huawei from accessing American semiconductor technology required to build its products. More importantly, it demonstrated the extraordinary power of US control over the semiconductor supply chain as a geopolitical weapon.
The Entity List action against Huawei had devastating consequences for the company's smartphone business. Huawei's HiSilicon division designed advanced chips for Huawei smartphones — the Kirin series — that were fabricated by TSMC using advanced process nodes. Once US export controls prevented TSMC from manufacturing chips for Huawei (because EDA tools and technology used by TSMC are subject to US jurisdiction under the "foreign direct product rule"), Huawei lost access to advanced chips for its smartphones. Its market share in smartphones collapsed from 20% globally to under 4%.
The Biden administration went further. In October 2022, the Commerce Department issued sweeping new export controls on advanced semiconductors and semiconductor manufacturing equipment going to China. The controls targeted:
• Advanced AI chips (A100, H100 GPUs and equivalents) — directly limiting China's AI training capability
• Advanced chip manufacturing equipment — specifically EUV lithography, but also advanced etch, deposition, and inspection tools
• Chips manufactured anywhere in the world using American technology (foreign direct product rule) going to specific Chinese entities
These controls were subsequently tightened in October 2023 and again in 2024, closing loopholes and expanding the list of affected technologies and entities. The United States also successfully pressured the Netherlands (which produces ASML's EUV machines) and Japan (which produces critical semiconductor equipment and materials) to join the export control regime.
The consequences for China have been severe but not terminal. China's domestic semiconductor industry — SMIC (Semiconductor Manufacturing International Corporation), CXMT (DRAM), Yangtze Memory Technologies (NAND flash) — has been forced to accelerate its development of domestic alternatives to American and Dutch equipment. China's SMIC surprised the world in 2023 by producing Huawei's Mate 60 Pro smartphone using a domestically developed 7nm process — a significant achievement that demonstrated China's ability to work around some aspects of export controls, even if behind the frontier.
For India, the US-China technology conflict creates both opportunities and risks. The opportunities are significant: as the United States and its allies work to reduce dependence on Chinese manufacturing and build more resilient supply chains, they are actively seeking alternative locations for semiconductor manufacturing, assembly, and testing. India — with its large, English-speaking technical workforce, democratic governance, and strategic alignment with the United States through frameworks like the Quad — is an obvious candidate for supply chain diversification.
The risks are also significant: the technology conflict is causing a fragmentation of the global technology ecosystem into competing blocs, which increases the cost and complexity of India's technology development. India, which has historically pursued strategic autonomy and non-alignment, faces pressure to choose sides in the technology conflict — a choice with major implications for its relationships with China (a neighbor with whom India has a fraught relationship but significant economic ties) and the United States (its most important technology partner).
The escalating technology conflict between the United States and China is the most important geopolitical context for understanding India's semiconductor and AI challenge. This conflict — which began under the Trump administration and has dramatically intensified under Biden and continues to evolve — is reshaping the global technology supply chain in ways that create both risks and opportunities for India.
The US-China technology war in semiconductors began in earnest with the Trump administration's actions against Huawei in 2019. Huawei — at the time the world's largest telecommunications equipment company and a significant smartphone brand — was placed on the US Entity List, restricting American companies from supplying Huawei without government authorization. This was intended to prevent Huawei from accessing American semiconductor technology required to build its products. More importantly, it demonstrated the extraordinary power of US control over the semiconductor supply chain as a geopolitical weapon.
The Entity List action against Huawei had devastating consequences for the company's smartphone business. Huawei's HiSilicon division designed advanced chips for Huawei smartphones — the Kirin series — that were fabricated by TSMC using advanced process nodes. Once US export controls prevented TSMC from manufacturing chips for Huawei (because EDA tools and technology used by TSMC are subject to US jurisdiction under the "foreign direct product rule"), Huawei lost access to advanced chips for its smartphones. Its market share in smartphones collapsed from 20% globally to under 4%.
The Biden administration went further. In October 2022, the Commerce Department issued sweeping new export controls on advanced semiconductors and semiconductor manufacturing equipment going to China. The controls targeted:
• Advanced AI chips (A100, H100 GPUs and equivalents) — directly limiting China's AI training capability
• Advanced chip manufacturing equipment — specifically EUV lithography, but also advanced etch, deposition, and inspection tools
• Chips manufactured anywhere in the world using American technology (foreign direct product rule) going to specific Chinese entities
These controls were subsequently tightened in October 2023 and again in 2024, closing loopholes and expanding the list of affected technologies and entities. The United States also successfully pressured the Netherlands (which produces ASML's EUV machines) and Japan (which produces critical semiconductor equipment and materials) to join the export control regime.
The consequences for China have been severe but not terminal. China's domestic semiconductor industry — SMIC (Semiconductor Manufacturing International Corporation), CXMT (DRAM), Yangtze Memory Technologies (NAND flash) — has been forced to accelerate its development of domestic alternatives to American and Dutch equipment. China's SMIC surprised the world in 2023 by producing Huawei's Mate 60 Pro smartphone using a domestically developed 7nm process — a significant achievement that demonstrated China's ability to work around some aspects of export controls, even if behind the frontier.
For India, the US-China technology conflict creates both opportunities and risks. The opportunities are significant: as the United States and its allies work to reduce dependence on Chinese manufacturing and build more resilient supply chains, they are actively seeking alternative locations for semiconductor manufacturing, assembly, and testing. India — with its large, English-speaking technical workforce, democratic governance, and strategic alignment with the United States through frameworks like the Quad — is an obvious candidate for supply chain diversification.
The risks are also significant: the technology conflict is causing a fragmentation of the global technology ecosystem into competing blocs, which increases the cost and complexity of India's technology development. India, which has historically pursued strategic autonomy and non-alignment, faces pressure to choose sides in the technology conflict — a choice with major implications for its relationships with China (a neighbor with whom India has a fraught relationship but significant economic ties) and the United States (its most important technology partner).
Taiwan — The Epicenter of Semiconductor Geopolitics
Taiwan's position in the global semiconductor supply chain is so extreme that it represents a unique geopolitical vulnerability for the entire world — and for India specifically. Understanding the "Taiwan Risk" is essential to understanding why India's lack of domestic semiconductor capability is so strategically dangerous.
TSMC — Taiwan Semiconductor Manufacturing Company — manufactures approximately 55-60% of all chips globally by value, and an even larger share of advanced chips. At the 5nm and 3nm nodes — the most advanced manufacturing processes — TSMC's market share is approximately 90%+. The company's fabs in Hsinchu, Taichung, and Tainan represent a concentration of irreplaceable manufacturing capability that has no parallel in human economic history. No other factory complex in the world is as economically critical.
The political status of Taiwan is, of course, deeply contested. The People's Republic of China claims Taiwan as part of Chinese territory and has never renounced the use of force to achieve unification. The United States maintains strategic ambiguity about whether it would militarily defend Taiwan against Chinese attack, while providing Taiwan with defensive weapons under the Taiwan Relations Act. The risk of military conflict across the Taiwan Strait — whether through Chinese invasion, a naval blockade, or escalating incidents — is a genuine and growing concern that has moved from the periphery of geopolitical risk analysis to its center.
Were a military conflict to erupt in the Taiwan Strait — even a limited one, let alone a full-scale invasion — the consequences for the global semiconductor supply chain would be catastrophic. TSMC's fabs would be unable to operate during an active conflict. Even if the physical infrastructure survived (which is uncertain), the evacuation or detention of skilled engineers, the disruption of ultra-pure materials supply chains, and the uncertainty about the political outcome would effectively halt production.
The consequences for the global economy would be severe: within months, the supply of advanced chips — required for everything from smartphones to servers to automotive electronics — would run out. Every data center would eventually degrade as servers could not be replaced. Every carmaker would halt production. Every smartphone manufacturer would have no chips for new devices. The global economy would face a shock comparable to or greater than the 2008 financial crisis, but with physical rather than financial origins.
For India, this scenario is particularly concerning. India's growing digital economy is entirely dependent on imported chips. A Taiwan Strait conflict could sever the supply of chips required to maintain Indian digital infrastructure — UPI transaction processing, mobile networks, data centers — within months. India has no domestic alternative supply.
This is not merely theoretical. The COVID-19 pandemic created a much smaller supply chain disruption — a partial, temporary disruption — that still resulted in a global semiconductor shortage that cost the automotive industry alone an estimated $200 billion in lost production. A Taiwan Strait conflict would be orders of magnitude more severe.
Chapter 15: The South Korean and Japanese Nodes — Other Vulnerabilities
While Taiwan represents the most acute single-point vulnerability in India's semiconductor supply chain dependency, it is not the only one. South Korea and Japan are equally critical nodes in the global semiconductor value chain, and India's exposure to disruptions in these countries is substantial.
South Korea is home to Samsung Electronics — the world's largest semiconductor company by revenue, the dominant player in DRAM (approximately 40% market share) and NAND flash (approximately 33% market share), and a significant foundry. SK Hynix — the second-largest DRAM manufacturer globally — is also Korean. Between them, Samsung and SK Hynix control approximately 70% of the global DRAM market. South Korea is also home to a significant portion of global NAND flash manufacturing capacity.
South Korea faces its own geopolitical vulnerabilities: North Korea's missile and nuclear program poses an existential threat; the country has complex relationships with both the United States and China; and its semiconductor industry would be directly threatened by any military escalation on the Korean Peninsula. A North Korean attack on South Korea — or even a credible nuclear threat — would disrupt Samsung and SK Hynix production with immediate global consequences.
Japan is critical for semiconductor materials and equipment. Japan supplies approximately:
• 56% of global silicon wafers (Shin-Etsu, Sumco)
• ~90% of some specific photoresist chemicals used in semiconductor lithography
• A large share of specialty gases and process chemicals
• Significant semiconductor equipment (Tokyo Electron, Nikon for lithography, Lasertec for inspection)
Japan's position in semiconductor materials means that any major disruption to Japanese supply — whether from a natural disaster (Japan's location on major seismic fault lines makes this a real risk), a political crisis, or deliberate weaponization — would have severe consequences for the global semiconductor industry. When a 2011 earthquake and tsunami devastated parts of Japan, the semiconductor industry (and particularly the automotive industry) felt supply shocks from disruption to Japanese materials suppliers.
For India, these concentrated vulnerabilities highlight the fundamental strategic risk of the current situation: India's digital economy, its growing electronics manufacturing sector, and its AI ambitions are all built on a foundation of completely external semiconductor supply — supply that is concentrated in three countries (Taiwan, South Korea, Japan) all located in the Indo-Pacific theater of geopolitical competition.
Taiwan's position in the global semiconductor supply chain is so extreme that it represents a unique geopolitical vulnerability for the entire world — and for India specifically. Understanding the "Taiwan Risk" is essential to understanding why India's lack of domestic semiconductor capability is so strategically dangerous.
TSMC — Taiwan Semiconductor Manufacturing Company — manufactures approximately 55-60% of all chips globally by value, and an even larger share of advanced chips. At the 5nm and 3nm nodes — the most advanced manufacturing processes — TSMC's market share is approximately 90%+. The company's fabs in Hsinchu, Taichung, and Tainan represent a concentration of irreplaceable manufacturing capability that has no parallel in human economic history. No other factory complex in the world is as economically critical.
The political status of Taiwan is, of course, deeply contested. The People's Republic of China claims Taiwan as part of Chinese territory and has never renounced the use of force to achieve unification. The United States maintains strategic ambiguity about whether it would militarily defend Taiwan against Chinese attack, while providing Taiwan with defensive weapons under the Taiwan Relations Act. The risk of military conflict across the Taiwan Strait — whether through Chinese invasion, a naval blockade, or escalating incidents — is a genuine and growing concern that has moved from the periphery of geopolitical risk analysis to its center.
Were a military conflict to erupt in the Taiwan Strait — even a limited one, let alone a full-scale invasion — the consequences for the global semiconductor supply chain would be catastrophic. TSMC's fabs would be unable to operate during an active conflict. Even if the physical infrastructure survived (which is uncertain), the evacuation or detention of skilled engineers, the disruption of ultra-pure materials supply chains, and the uncertainty about the political outcome would effectively halt production.
The consequences for the global economy would be severe: within months, the supply of advanced chips — required for everything from smartphones to servers to automotive electronics — would run out. Every data center would eventually degrade as servers could not be replaced. Every carmaker would halt production. Every smartphone manufacturer would have no chips for new devices. The global economy would face a shock comparable to or greater than the 2008 financial crisis, but with physical rather than financial origins.
For India, this scenario is particularly concerning. India's growing digital economy is entirely dependent on imported chips. A Taiwan Strait conflict could sever the supply of chips required to maintain Indian digital infrastructure — UPI transaction processing, mobile networks, data centers — within months. India has no domestic alternative supply.
This is not merely theoretical. The COVID-19 pandemic created a much smaller supply chain disruption — a partial, temporary disruption — that still resulted in a global semiconductor shortage that cost the automotive industry alone an estimated $200 billion in lost production. A Taiwan Strait conflict would be orders of magnitude more severe.
Chapter 15: The South Korean and Japanese Nodes — Other Vulnerabilities
While Taiwan represents the most acute single-point vulnerability in India's semiconductor supply chain dependency, it is not the only one. South Korea and Japan are equally critical nodes in the global semiconductor value chain, and India's exposure to disruptions in these countries is substantial.
South Korea is home to Samsung Electronics — the world's largest semiconductor company by revenue, the dominant player in DRAM (approximately 40% market share) and NAND flash (approximately 33% market share), and a significant foundry. SK Hynix — the second-largest DRAM manufacturer globally — is also Korean. Between them, Samsung and SK Hynix control approximately 70% of the global DRAM market. South Korea is also home to a significant portion of global NAND flash manufacturing capacity.
South Korea faces its own geopolitical vulnerabilities: North Korea's missile and nuclear program poses an existential threat; the country has complex relationships with both the United States and China; and its semiconductor industry would be directly threatened by any military escalation on the Korean Peninsula. A North Korean attack on South Korea — or even a credible nuclear threat — would disrupt Samsung and SK Hynix production with immediate global consequences.
Japan is critical for semiconductor materials and equipment. Japan supplies approximately:
• 56% of global silicon wafers (Shin-Etsu, Sumco)
• ~90% of some specific photoresist chemicals used in semiconductor lithography
• A large share of specialty gases and process chemicals
• Significant semiconductor equipment (Tokyo Electron, Nikon for lithography, Lasertec for inspection)
Japan's position in semiconductor materials means that any major disruption to Japanese supply — whether from a natural disaster (Japan's location on major seismic fault lines makes this a real risk), a political crisis, or deliberate weaponization — would have severe consequences for the global semiconductor industry. When a 2011 earthquake and tsunami devastated parts of Japan, the semiconductor industry (and particularly the automotive industry) felt supply shocks from disruption to Japanese materials suppliers.
For India, these concentrated vulnerabilities highlight the fundamental strategic risk of the current situation: India's digital economy, its growing electronics manufacturing sector, and its AI ambitions are all built on a foundation of completely external semiconductor supply — supply that is concentrated in three countries (Taiwan, South Korea, Japan) all located in the Indo-Pacific theater of geopolitical competition.
China's Semiconductor Ambition — The Competitive Pressure on India
China's aggressive drive to develop domestic semiconductor capability — accelerated by US export controls — creates a competitive pressure on India that is often underappreciated. China is investing at a scale that dwarfs India's efforts, and understanding the Chinese approach provides both a template and a competitive challenge.
China's semiconductor strategy has been in development for decades, but received its most powerful articulation in the "Made in China 2025" plan announced in 2015, which identified semiconductors as a strategic priority for domestic development. The plan called for China to produce 40% of the semiconductors it consumes domestically by 2020 and 70% by 2025 — ambitions that have not been fully achieved but have driven significant investment.
The primary vehicle for Chinese semiconductor investment has been the National Integrated Circuit Industry Investment Fund — colloquially known as the "Big Fund" — which has committed approximately ¥200 billion (about $27 billion) across two tranches (2014 and 2019) to semiconductor companies, with additional provincial government funds adding comparable amounts. Total Chinese government investment in the semiconductor industry since 2014 is estimated to exceed $50 billion.
This investment has produced results. SMIC — China's leading pure-play foundry — has achieved 7nm process capability despite restrictions on advanced equipment access, though its yields and production volumes at advanced nodes remain below TSMC's. CXMT has become a domestic supplier of LPDDR4 DRAM. YMTC (Yangtze Memory Technologies) has developed 128-layer and 232-layer 3D NAND flash chips. Loongson and Zhaoxin have developed domestic general-purpose processor chips. Biren Technology and Cambricon have developed domestic AI accelerator chips.
The quality gap between Chinese chips and the global frontier is still significant — particularly in AI accelerators, where Chinese domestic chips are estimated to be 1-2 generations behind Nvidia's latest offerings. But the trajectory is one of rapid progress, and China's domestic semiconductor industry will likely be substantially more capable in five years than it is today.
For India, China's semiconductor progress is relevant in several ways. First, China and India compete for the same pool of global semiconductor investment — as multinationals look to diversify away from Taiwan and China, they are evaluating both India and China (and Southeast Asian countries) as investment destinations. China's improved domestic capability reduces the incentive for foreign semiconductor companies to invest in China for access to the Chinese market, which may redirect some investment toward India. Conversely, Chinese semiconductor companies with cost-competitive products will compete with potential Indian producers in third markets.
Second, China's semiconductor progress — particularly in AI chips — means that if Indian AI development falls further behind the global frontier, Indian AI applications may become dependent on Chinese AI chips (already sold through third-country resellers despite US export controls). This would create a different but equally concerning geopolitical dependency.
PART FIVE: INDIA'S POLICY RESPONSE — CURRENT INITIATIVES AND THEIR ADEQUACY
China's aggressive drive to develop domestic semiconductor capability — accelerated by US export controls — creates a competitive pressure on India that is often underappreciated. China is investing at a scale that dwarfs India's efforts, and understanding the Chinese approach provides both a template and a competitive challenge.
China's semiconductor strategy has been in development for decades, but received its most powerful articulation in the "Made in China 2025" plan announced in 2015, which identified semiconductors as a strategic priority for domestic development. The plan called for China to produce 40% of the semiconductors it consumes domestically by 2020 and 70% by 2025 — ambitions that have not been fully achieved but have driven significant investment.
The primary vehicle for Chinese semiconductor investment has been the National Integrated Circuit Industry Investment Fund — colloquially known as the "Big Fund" — which has committed approximately ¥200 billion (about $27 billion) across two tranches (2014 and 2019) to semiconductor companies, with additional provincial government funds adding comparable amounts. Total Chinese government investment in the semiconductor industry since 2014 is estimated to exceed $50 billion.
This investment has produced results. SMIC — China's leading pure-play foundry — has achieved 7nm process capability despite restrictions on advanced equipment access, though its yields and production volumes at advanced nodes remain below TSMC's. CXMT has become a domestic supplier of LPDDR4 DRAM. YMTC (Yangtze Memory Technologies) has developed 128-layer and 232-layer 3D NAND flash chips. Loongson and Zhaoxin have developed domestic general-purpose processor chips. Biren Technology and Cambricon have developed domestic AI accelerator chips.
The quality gap between Chinese chips and the global frontier is still significant — particularly in AI accelerators, where Chinese domestic chips are estimated to be 1-2 generations behind Nvidia's latest offerings. But the trajectory is one of rapid progress, and China's domestic semiconductor industry will likely be substantially more capable in five years than it is today.
For India, China's semiconductor progress is relevant in several ways. First, China and India compete for the same pool of global semiconductor investment — as multinationals look to diversify away from Taiwan and China, they are evaluating both India and China (and Southeast Asian countries) as investment destinations. China's improved domestic capability reduces the incentive for foreign semiconductor companies to invest in China for access to the Chinese market, which may redirect some investment toward India. Conversely, Chinese semiconductor companies with cost-competitive products will compete with potential Indian producers in third markets.
Second, China's semiconductor progress — particularly in AI chips — means that if Indian AI development falls further behind the global frontier, Indian AI applications may become dependent on Chinese AI chips (already sold through third-country resellers despite US export controls). This would create a different but equally concerning geopolitical dependency.
PART FIVE: INDIA'S POLICY RESPONSE — CURRENT INITIATIVES AND THEIR ADEQUACY
The India Semiconductor Mission — Ambition Meets Reality
The Indian government's most significant recent response to the semiconductor challenge has been the Semicon India program, launched in December 2021 and significantly expanded in 2023 and 2024. Understanding this program — its scale, its structure, its achievements, and its shortcomings — is essential for assessing whether India is on a credible path to building domestic semiconductor capability.
The program's centerpiece is the Production Linked Incentive (PLI) scheme for semiconductors, which commits government support of approximately ₹76,000 crore (approximately $10 billion) for semiconductor and display manufacturing. The scheme is structured as a fiscal support mechanism: the central government will fund 50% of the project cost for semiconductor fabs, display fabs, and packaging facilities. State governments are expected to contribute additional support through land, power, water, and additional fiscal incentives.
The India Semiconductor Mission (ISM), established as a specialized body under the Ministry of Electronics and Information Technology (MeitY), has been designated as the nodal agency for semiconductor policy implementation. ISM is tasked with attracting investment, coordinating with state governments, facilitating technology transfer, and building the broader semiconductor ecosystem.
The first significant announcement under the Semicon India program came in June 2023, when the government approved three proposals:
1. Micron Technology — the American DRAM and NAND flash manufacturer — announced an investment of approximately $825 million (with government support bringing total investment to $2.75 billion) to build a semiconductor assembly and test (OSAT) facility in Sanand, Gujarat. This was the first major American semiconductor company to commit to manufacturing investment in India.
2. CG Power and Industrial Solutions in collaboration with Renesas Electronics (Japan) and Stars Microelectronics (Thailand) — announced a semiconductor OSAT facility with a planned investment of ₹7,600 crore in Sanand, Gujarat.
3. Tata Electronics in collaboration with Power Semiconductor Limited (PSMC) of Taiwan — announced a semiconductor fab investment in Gujarat with an estimated investment of ₹91,000 crore, focusing on 28nm process technology.
These announcements were significant — they represented real commitments from serious companies, not just letters of intent. But it is important to assess them in context.
The Micron facility is an OSAT — assembly and test — not a wafer fab. OSAT represents the lowest-value, least technology-intensive segment of the semiconductor supply chain. The value added in OSAT — packaging and testing chips made elsewhere — is a fraction of the value added in wafer fabrication. Micron's investment in India will create manufacturing jobs and build some semiconductor ecosystem capabilities, but it does not address India's dependency on imported chips at the wafer level.
The Tata-PSMC joint venture — focused on 28nm process technology — is more significant, representing India's first serious attempt to establish wafer fabrication capability. But 28nm is a process node that was at the frontier in approximately 2011-2012. Today's frontier process nodes are 3nm (TSMC) and 2nm (being developed). 28nm chips are used in mature, less performance-critical applications — automotive chips, industrial microcontrollers, display drivers, power management chips, and some wireless connectivity chips. They are not suitable for the AI accelerators, smartphone processors, or advanced memory chips that represent the highest-value and most strategically critical semiconductors.
This "lagging edge" approach — building 28nm (or slightly more advanced) capability while the global frontier moves to 2nm — is not inherently wrong. In fact, there is a strong argument that the mature node market is where India can realistically compete in the near term, and that establishing any domestic wafer fabrication capability is better than none. But it must be understood for what it is: a beginning, not an arrival. India building 28nm wafer fabrication capability in 2025-2028 is equivalent to building a steel plant in the era of advanced composites — useful and necessary, but not sufficient for the highest-value applications.
The Indian government's most significant recent response to the semiconductor challenge has been the Semicon India program, launched in December 2021 and significantly expanded in 2023 and 2024. Understanding this program — its scale, its structure, its achievements, and its shortcomings — is essential for assessing whether India is on a credible path to building domestic semiconductor capability.
The program's centerpiece is the Production Linked Incentive (PLI) scheme for semiconductors, which commits government support of approximately ₹76,000 crore (approximately $10 billion) for semiconductor and display manufacturing. The scheme is structured as a fiscal support mechanism: the central government will fund 50% of the project cost for semiconductor fabs, display fabs, and packaging facilities. State governments are expected to contribute additional support through land, power, water, and additional fiscal incentives.
The India Semiconductor Mission (ISM), established as a specialized body under the Ministry of Electronics and Information Technology (MeitY), has been designated as the nodal agency for semiconductor policy implementation. ISM is tasked with attracting investment, coordinating with state governments, facilitating technology transfer, and building the broader semiconductor ecosystem.
The first significant announcement under the Semicon India program came in June 2023, when the government approved three proposals:
1. Micron Technology — the American DRAM and NAND flash manufacturer — announced an investment of approximately $825 million (with government support bringing total investment to $2.75 billion) to build a semiconductor assembly and test (OSAT) facility in Sanand, Gujarat. This was the first major American semiconductor company to commit to manufacturing investment in India.
2. CG Power and Industrial Solutions in collaboration with Renesas Electronics (Japan) and Stars Microelectronics (Thailand) — announced a semiconductor OSAT facility with a planned investment of ₹7,600 crore in Sanand, Gujarat.
3. Tata Electronics in collaboration with Power Semiconductor Limited (PSMC) of Taiwan — announced a semiconductor fab investment in Gujarat with an estimated investment of ₹91,000 crore, focusing on 28nm process technology.
These announcements were significant — they represented real commitments from serious companies, not just letters of intent. But it is important to assess them in context.
The Micron facility is an OSAT — assembly and test — not a wafer fab. OSAT represents the lowest-value, least technology-intensive segment of the semiconductor supply chain. The value added in OSAT — packaging and testing chips made elsewhere — is a fraction of the value added in wafer fabrication. Micron's investment in India will create manufacturing jobs and build some semiconductor ecosystem capabilities, but it does not address India's dependency on imported chips at the wafer level.
The Tata-PSMC joint venture — focused on 28nm process technology — is more significant, representing India's first serious attempt to establish wafer fabrication capability. But 28nm is a process node that was at the frontier in approximately 2011-2012. Today's frontier process nodes are 3nm (TSMC) and 2nm (being developed). 28nm chips are used in mature, less performance-critical applications — automotive chips, industrial microcontrollers, display drivers, power management chips, and some wireless connectivity chips. They are not suitable for the AI accelerators, smartphone processors, or advanced memory chips that represent the highest-value and most strategically critical semiconductors.
This "lagging edge" approach — building 28nm (or slightly more advanced) capability while the global frontier moves to 2nm — is not inherently wrong. In fact, there is a strong argument that the mature node market is where India can realistically compete in the near term, and that establishing any domestic wafer fabrication capability is better than none. But it must be understood for what it is: a beginning, not an arrival. India building 28nm wafer fabrication capability in 2025-2028 is equivalent to building a steel plant in the era of advanced composites — useful and necessary, but not sufficient for the highest-value applications.
The CHIPS Act Context — How India's Policy Compares Internationally
To assess India's Semicon India program, it is instructive to compare it to the semiconductor industrial policies being implemented by other major economies.
The United States CHIPS and Science Act (2022): The US government committed $52.7 billion for semiconductor manufacturing, research, and workforce development — approximately five times India's Semicon India commitment. The US CHIPS Act includes $39 billion in direct subsidies for semiconductor manufacturing, $13 billion for research and development, and a 25% investment tax credit for semiconductor manufacturing equipment. The US effort is targeted explicitly at attracting TSMC, Samsung, and Intel to build leading-edge fabrication capacity on American soil — TSMC's fab in Arizona (targeting 3nm and 2nm production by 2026-2028) received approximately $6.6 billion in direct grants and up to $5 billion in loans from the US government.
The European Chips Act (2023): The European Union committed €43 billion (approximately $46 billion) to semiconductor manufacturing and research, with the goal of doubling Europe's share of global semiconductor production to 20% by 2030. Intel's planned fab in Germany (announced but subsequently delayed) would have received significant European and German government support.
Japan's Semiconductor Revitalization: Japan committed approximately ¥4 trillion (approximately $30 billion) to semiconductor manufacturing, with a focus on attracting TSMC to build a fab in Japan — TSMC Kumamoto Phase 1 (16nm/22nm processes) opened in early 2024 with ¥476 billion ($3.3 billion) in Japanese government support. Phase 2, targeting more advanced processes, is also planned.
South Korea's K-Semiconductor Strategy: South Korea announced a ₩510 trillion (approximately $450 billion) private-sector semiconductor investment plan through 2030, with government support through tax incentives, R&D subsidies, and infrastructure investment. This is Samsung and SK Hynix investing to maintain their dominant global positions — but backed by powerful government support.
China's Investment: As discussed previously, China's government has committed approximately $50 billion in direct investment through the Big Fund and provincial governments, with total public-private semiconductor investment estimated in the hundreds of billions of dollars.
Against this backdrop, India's $10 billion commitment — much of which is structured as matching grants requiring equal private investment — appears modest. The United States, Europe, Japan, and South Korea are all committing substantially more in absolute terms to semiconductor capability development, even though India's per-capita income is far lower and the fiscal headroom is more constrained.
More significantly, the international programs are explicitly competing for the same global semiconductor investment — particularly TSMC's fab expansion plans, advanced memory manufacturing investment, and equipment purchases. India is competing for semiconductor investment with countries that are offering more subsidies, better infrastructure, larger domestic markets (US, Europe), established semiconductor ecosystems (Japan, South Korea), or all of the above.
To assess India's Semicon India program, it is instructive to compare it to the semiconductor industrial policies being implemented by other major economies.
The United States CHIPS and Science Act (2022): The US government committed $52.7 billion for semiconductor manufacturing, research, and workforce development — approximately five times India's Semicon India commitment. The US CHIPS Act includes $39 billion in direct subsidies for semiconductor manufacturing, $13 billion for research and development, and a 25% investment tax credit for semiconductor manufacturing equipment. The US effort is targeted explicitly at attracting TSMC, Samsung, and Intel to build leading-edge fabrication capacity on American soil — TSMC's fab in Arizona (targeting 3nm and 2nm production by 2026-2028) received approximately $6.6 billion in direct grants and up to $5 billion in loans from the US government.
The European Chips Act (2023): The European Union committed €43 billion (approximately $46 billion) to semiconductor manufacturing and research, with the goal of doubling Europe's share of global semiconductor production to 20% by 2030. Intel's planned fab in Germany (announced but subsequently delayed) would have received significant European and German government support.
Japan's Semiconductor Revitalization: Japan committed approximately ¥4 trillion (approximately $30 billion) to semiconductor manufacturing, with a focus on attracting TSMC to build a fab in Japan — TSMC Kumamoto Phase 1 (16nm/22nm processes) opened in early 2024 with ¥476 billion ($3.3 billion) in Japanese government support. Phase 2, targeting more advanced processes, is also planned.
South Korea's K-Semiconductor Strategy: South Korea announced a ₩510 trillion (approximately $450 billion) private-sector semiconductor investment plan through 2030, with government support through tax incentives, R&D subsidies, and infrastructure investment. This is Samsung and SK Hynix investing to maintain their dominant global positions — but backed by powerful government support.
China's Investment: As discussed previously, China's government has committed approximately $50 billion in direct investment through the Big Fund and provincial governments, with total public-private semiconductor investment estimated in the hundreds of billions of dollars.
Against this backdrop, India's $10 billion commitment — much of which is structured as matching grants requiring equal private investment — appears modest. The United States, Europe, Japan, and South Korea are all committing substantially more in absolute terms to semiconductor capability development, even though India's per-capita income is far lower and the fiscal headroom is more constrained.
More significantly, the international programs are explicitly competing for the same global semiconductor investment — particularly TSMC's fab expansion plans, advanced memory manufacturing investment, and equipment purchases. India is competing for semiconductor investment with countries that are offering more subsidies, better infrastructure, larger domestic markets (US, Europe), established semiconductor ecosystems (Japan, South Korea), or all of the above.
The AI Policy Response — India's AI Mission
India's policy response to the AI challenge has been more recent and is still taking shape. The government's approach has several components:
IndiaAI Mission: Launched in 2024 with a budget of ₹10,371 crore (approximately $1.25 billion) over five years, the IndiaAI Mission aims to build AI infrastructure, develop AI for public good applications, create AI datasets, support AI startups, and develop AI skills. The centerpiece is the creation of a shared AI compute infrastructure — a national AI supercomputing cluster — that would make GPU compute available to Indian startups, researchers, and government agencies at subsidized rates.
The compute commitment under IndiaAI — targeting approximately 10,000 GPUs — is a meaningful start but far below what is required for frontier model training. Training a single frontier LLM comparable to GPT-4 would require the equivalent of thousands of H100 GPUs running for months. A cluster of 10,000 GPUs, shared across all of India's AI researchers and companies, will be constantly oversubscribed and insufficient for the most ambitious training runs.
National AI Strategy: The NITI Aayog has published a series of National Strategy for AI documents (first in 2018, updated subsequently) that lay out India's vision for AI development — emphasizing AI for social good, healthcare, education, smart mobility, and agriculture. These documents are thoughtful in identifying the domains where AI can make the greatest difference for India's population, but they are light on the specific interventions required to build foundational AI capability.
Krutrim and the Private Sector: The most ambitious Indian private sector AI initiative is Krutrim, founded by Bhavish Aggarwal (founder of Ola). Krutrim became India's first AI unicorn in January 2024. The company has announced plans to develop Indian language AI models, AI-optimized chips, and a dedicated AI cloud infrastructure for India. Krutrim's chip initiative — developing India's first AI chip — would be a significant achievement if it succeeds, but chip design (even if successful) would still require offshore manufacturing, likely at TSMC or Samsung, maintaining the fabrication dependency even if the design is Indian.
Sarvam AI — founded by former Google and Meta researchers — is another significant initiative, focusing specifically on Indian language AI models. Sarvam's SarvamGPT models have demonstrated competitive performance on Indian language tasks. AI4Bharat, an academic initiative from IIT Madras, has developed open-source tools and datasets for Indian language AI.
The Compute Gap as the Binding Constraint: Despite these initiatives, the binding constraint on Indian AI capability remains compute. The cost of training frontier models has been growing rapidly — estimates suggest that training GPT-4 equivalent models now requires $50-100 million in compute costs alone, and the next generation will require more. Indian AI companies and research institutions cannot afford this level of compute investment, and the national AI compute infrastructure being built is insufficient to bridge the gap.
The solution requires either (a) massive increase in AI compute investment — orders of magnitude more than currently planned, or (b) a different strategic approach that focuses on fine-tuning and adapting global frontier models for Indian use cases rather than training frontier models from scratch, or (c) international cooperation — accessing compute resources from partner countries (potentially the US, through Quad frameworks, or other partners) on preferential terms.
India's policy response to the AI challenge has been more recent and is still taking shape. The government's approach has several components:
IndiaAI Mission: Launched in 2024 with a budget of ₹10,371 crore (approximately $1.25 billion) over five years, the IndiaAI Mission aims to build AI infrastructure, develop AI for public good applications, create AI datasets, support AI startups, and develop AI skills. The centerpiece is the creation of a shared AI compute infrastructure — a national AI supercomputing cluster — that would make GPU compute available to Indian startups, researchers, and government agencies at subsidized rates.
The compute commitment under IndiaAI — targeting approximately 10,000 GPUs — is a meaningful start but far below what is required for frontier model training. Training a single frontier LLM comparable to GPT-4 would require the equivalent of thousands of H100 GPUs running for months. A cluster of 10,000 GPUs, shared across all of India's AI researchers and companies, will be constantly oversubscribed and insufficient for the most ambitious training runs.
National AI Strategy: The NITI Aayog has published a series of National Strategy for AI documents (first in 2018, updated subsequently) that lay out India's vision for AI development — emphasizing AI for social good, healthcare, education, smart mobility, and agriculture. These documents are thoughtful in identifying the domains where AI can make the greatest difference for India's population, but they are light on the specific interventions required to build foundational AI capability.
Krutrim and the Private Sector: The most ambitious Indian private sector AI initiative is Krutrim, founded by Bhavish Aggarwal (founder of Ola). Krutrim became India's first AI unicorn in January 2024. The company has announced plans to develop Indian language AI models, AI-optimized chips, and a dedicated AI cloud infrastructure for India. Krutrim's chip initiative — developing India's first AI chip — would be a significant achievement if it succeeds, but chip design (even if successful) would still require offshore manufacturing, likely at TSMC or Samsung, maintaining the fabrication dependency even if the design is Indian.
Sarvam AI — founded by former Google and Meta researchers — is another significant initiative, focusing specifically on Indian language AI models. Sarvam's SarvamGPT models have demonstrated competitive performance on Indian language tasks. AI4Bharat, an academic initiative from IIT Madras, has developed open-source tools and datasets for Indian language AI.
The Compute Gap as the Binding Constraint: Despite these initiatives, the binding constraint on Indian AI capability remains compute. The cost of training frontier models has been growing rapidly — estimates suggest that training GPT-4 equivalent models now requires $50-100 million in compute costs alone, and the next generation will require more. Indian AI companies and research institutions cannot afford this level of compute investment, and the national AI compute infrastructure being built is insufficient to bridge the gap.
The solution requires either (a) massive increase in AI compute investment — orders of magnitude more than currently planned, or (b) a different strategic approach that focuses on fine-tuning and adapting global frontier models for Indian use cases rather than training frontier models from scratch, or (c) international cooperation — accessing compute resources from partner countries (potentially the US, through Quad frameworks, or other partners) on preferential terms.
The Electronics Manufacturing Push — From iPhone Assembly to Something More?
India's most visible recent success in electronics is the growing role of Indian manufacturing in the smartphone supply chain — particularly Apple's rapid shift toward India manufacturing. Understanding this development is important for assessing India's trajectory in electronics and its semiconductor implications.
Apple began manufacturing iPhones in India in 2017 through its contract manufacturers Foxconn and Wistron. Initially, production was limited to older models for the domestic market. But the pace of Apple's India manufacturing expansion has accelerated dramatically, driven by supply chain diversification from China, India's Production Linked Incentive (PLI) scheme for smartphones, and India's growing market importance.
By 2024, India was assembling approximately 14% of global iPhone production — up from essentially zero in 2020. Apple's target is to manufacture approximately 25% of its global iPhone production in India by 2025-2026. Foxconn has made massive investments in its Tamil Nadu facilities; Tata Electronics acquired Wistron's Indian operations in 2023 and has become Apple's first Indian contract manufacturer. The PLI scheme for smartphones has delivered strong results, with mobile phone exports from India growing from essentially zero to over $15 billion.
But here is the critical limitation: the iPhone assembled in India is assembled from components that are almost entirely imported. The display is made in South Korea or China (Samsung, LG, BOE). The processor (Apple's A-series chips) is manufactured by TSMC in Taiwan. The DRAM (LPDDR5) is supplied by SK Hynix. The NAND flash (UFS 3.1) is supplied by SK Hynix or Samsung. The battery is manufactured by Chinese suppliers. The cameras are supplied by Japanese (Sony, Sharp) or Chinese companies.
India's role in the iPhone supply chain is primarily final assembly and testing — the least value-intensive step, accounting for perhaps 5-10% of the total value of the device. The remaining 90-95% of the value is created outside India, primarily in Taiwan, South Korea, Japan, and China. This is an improvement over zero, and the ecosystem effects — training a manufacturing workforce, building logistics infrastructure, attracting component suppliers — are real and important. But it is not domestic semiconductor manufacturing, and it does not address India's fundamental technology deficit.
The PLI scheme has also attracted investment in several other electronics categories: display manufacturing, wearables, laptops and tablets, and telecommunications equipment. These are meaningful steps but still largely assembly-oriented rather than component-manufacturing-oriented.
The critical question for India's electronics strategy is whether the current phase — assembly of imported components — can be leveraged into a deeper integration in the supply chain, moving progressively from assembly to component manufacturing to semiconductor manufacturing. South Korea's trajectory from consumer electronics assembly in the 1970s to Samsung's global semiconductor dominance in the 2000s shows that this progression is possible but requires deliberate industrial policy, massive long-term investment, and decades of patient effort.
India's most visible recent success in electronics is the growing role of Indian manufacturing in the smartphone supply chain — particularly Apple's rapid shift toward India manufacturing. Understanding this development is important for assessing India's trajectory in electronics and its semiconductor implications.
Apple began manufacturing iPhones in India in 2017 through its contract manufacturers Foxconn and Wistron. Initially, production was limited to older models for the domestic market. But the pace of Apple's India manufacturing expansion has accelerated dramatically, driven by supply chain diversification from China, India's Production Linked Incentive (PLI) scheme for smartphones, and India's growing market importance.
By 2024, India was assembling approximately 14% of global iPhone production — up from essentially zero in 2020. Apple's target is to manufacture approximately 25% of its global iPhone production in India by 2025-2026. Foxconn has made massive investments in its Tamil Nadu facilities; Tata Electronics acquired Wistron's Indian operations in 2023 and has become Apple's first Indian contract manufacturer. The PLI scheme for smartphones has delivered strong results, with mobile phone exports from India growing from essentially zero to over $15 billion.
But here is the critical limitation: the iPhone assembled in India is assembled from components that are almost entirely imported. The display is made in South Korea or China (Samsung, LG, BOE). The processor (Apple's A-series chips) is manufactured by TSMC in Taiwan. The DRAM (LPDDR5) is supplied by SK Hynix. The NAND flash (UFS 3.1) is supplied by SK Hynix or Samsung. The battery is manufactured by Chinese suppliers. The cameras are supplied by Japanese (Sony, Sharp) or Chinese companies.
India's role in the iPhone supply chain is primarily final assembly and testing — the least value-intensive step, accounting for perhaps 5-10% of the total value of the device. The remaining 90-95% of the value is created outside India, primarily in Taiwan, South Korea, Japan, and China. This is an improvement over zero, and the ecosystem effects — training a manufacturing workforce, building logistics infrastructure, attracting component suppliers — are real and important. But it is not domestic semiconductor manufacturing, and it does not address India's fundamental technology deficit.
The PLI scheme has also attracted investment in several other electronics categories: display manufacturing, wearables, laptops and tablets, and telecommunications equipment. These are meaningful steps but still largely assembly-oriented rather than component-manufacturing-oriented.
The critical question for India's electronics strategy is whether the current phase — assembly of imported components — can be leveraged into a deeper integration in the supply chain, moving progressively from assembly to component manufacturing to semiconductor manufacturing. South Korea's trajectory from consumer electronics assembly in the 1970s to Samsung's global semiconductor dominance in the 2000s shows that this progression is possible but requires deliberate industrial policy, massive long-term investment, and decades of patient effort.
PART SIX: HOW TO TACKLE IT — A STRATEGIC FRAMEWORK FOR INDIA'S TECHNOLOGY FUTURE
The Strategic Vision — What India Must Achieve
Before discussing specific policies, it is important to establish a clear strategic vision. What precisely should India aim to achieve in semiconductors, DRAM, and AI, over what timeframe, and at what cost? Without a clear vision — one that is specific, measurable, and time-bound — policy will remain diffuse and insufficient.
India's strategic objectives in technology should be organized around three interrelated goals:
Strategic Autonomy: The ability to function as a sovereign nation in the digital age without being held hostage to supplier decisions, export controls, or geopolitical pressures of foreign powers. This requires domestic capability in at least some critical segments of the semiconductor value chain — not necessarily self-sufficiency (which is economically impossible and undesirable) but enough domestic capacity to ensure continued operation of critical systems even if imports are disrupted.
Economic Value Creation: The ability to capture a larger share of the enormous value created in the global technology sector. India currently captures a large share of IT services value but a minimal share of semiconductor, AI, and hardware value. Shifting this balance would dramatically improve India's trade position, create high-quality jobs, and accelerate GDP growth.
Technological Leadership: The ability to be at or near the frontier in at least some technology domains — not as a follower or adaptor but as an originator of new technologies, architectures, and approaches. This is the most ambitious goal and the longest-term, but without it, India will always be dependent on others' innovations.
Given these goals, India's specific near-term (5-10 year), medium-term (10-20 year), and long-term (20-30 year) objectives might be:
Near-term (by 2030):
• Establish functioning OSAT capabilities at competitive scale (underway with Micron)
• Commission India's first 28nm wafer fabrication facility (Tata-PSMC project)
• Build national AI compute infrastructure sufficient for research and model fine-tuning
• Develop competitive Indian language AI models fine-tuned from global frontier models
• Create a semiconductor design ecosystem with a growing number of domestic fabless companies
• Establish chip design partnerships and technology licensing agreements with leading global firms
Medium-term (2030-2040):
• Upgrade wafer fabrication to sub-10nm process nodes (requires technology transfer from TSMC, Samsung, or Intel, or domestic development)
• Establish domestic DRAM manufacturing at older nodes (LPDDR4 or DDR4 equivalent)
• Develop domestic AI chip designs fabricated at TSMC or Samsung
• Build a competitive domestic AI foundation model ecosystem
• Capture 5-10% of global electronics component value added (vs. near-zero today)
• Develop a domestic semiconductor equipment manufacturing ecosystem for selected equipment categories
Long-term (2040-2050):
• Participate in frontier semiconductor process development (sub-3nm equivalent)
• Be a significant global player in at least one semiconductor memory technology
• Have at least one globally competitive AI model frontier company
• Achieve domestic semiconductor supply sufficient to insulate critical systems from global supply disruptions
• Be a net exporter of semiconductor technology, rather than purely a net importer
This is an ambitious but achievable trajectory — if the right policies are implemented with sufficient commitment and resources.
Before discussing specific policies, it is important to establish a clear strategic vision. What precisely should India aim to achieve in semiconductors, DRAM, and AI, over what timeframe, and at what cost? Without a clear vision — one that is specific, measurable, and time-bound — policy will remain diffuse and insufficient.
India's strategic objectives in technology should be organized around three interrelated goals:
Strategic Autonomy: The ability to function as a sovereign nation in the digital age without being held hostage to supplier decisions, export controls, or geopolitical pressures of foreign powers. This requires domestic capability in at least some critical segments of the semiconductor value chain — not necessarily self-sufficiency (which is economically impossible and undesirable) but enough domestic capacity to ensure continued operation of critical systems even if imports are disrupted.
Economic Value Creation: The ability to capture a larger share of the enormous value created in the global technology sector. India currently captures a large share of IT services value but a minimal share of semiconductor, AI, and hardware value. Shifting this balance would dramatically improve India's trade position, create high-quality jobs, and accelerate GDP growth.
Technological Leadership: The ability to be at or near the frontier in at least some technology domains — not as a follower or adaptor but as an originator of new technologies, architectures, and approaches. This is the most ambitious goal and the longest-term, but without it, India will always be dependent on others' innovations.
Given these goals, India's specific near-term (5-10 year), medium-term (10-20 year), and long-term (20-30 year) objectives might be:
Near-term (by 2030):
• Establish functioning OSAT capabilities at competitive scale (underway with Micron)
• Commission India's first 28nm wafer fabrication facility (Tata-PSMC project)
• Build national AI compute infrastructure sufficient for research and model fine-tuning
• Develop competitive Indian language AI models fine-tuned from global frontier models
• Create a semiconductor design ecosystem with a growing number of domestic fabless companies
• Establish chip design partnerships and technology licensing agreements with leading global firms
Medium-term (2030-2040):
• Upgrade wafer fabrication to sub-10nm process nodes (requires technology transfer from TSMC, Samsung, or Intel, or domestic development)
• Establish domestic DRAM manufacturing at older nodes (LPDDR4 or DDR4 equivalent)
• Develop domestic AI chip designs fabricated at TSMC or Samsung
• Build a competitive domestic AI foundation model ecosystem
• Capture 5-10% of global electronics component value added (vs. near-zero today)
• Develop a domestic semiconductor equipment manufacturing ecosystem for selected equipment categories
Long-term (2040-2050):
• Participate in frontier semiconductor process development (sub-3nm equivalent)
• Be a significant global player in at least one semiconductor memory technology
• Have at least one globally competitive AI model frontier company
• Achieve domestic semiconductor supply sufficient to insulate critical systems from global supply disruptions
• Be a net exporter of semiconductor technology, rather than purely a net importer
This is an ambitious but achievable trajectory — if the right policies are implemented with sufficient commitment and resources.
The Investment Imperative — Scale of Funding Required
The single most important message about India's semiconductor and AI ambitions is that they require resources at a scale that India has not yet committed. Semiconductor manufacturing is the most capital-intensive industry in human history. Getting this right requires confronting the numbers honestly.
Building a competitive 28nm wafer fabrication facility costs approximately $3-5 billion. Building a competitive advanced node fab (7nm or below) costs $15-20 billion. Building a DRAM fab costs $10-15 billion. Building a frontier AI training compute cluster requires $1-5 billion in GPU hardware alone, plus data center infrastructure.
India's total Semicon India commitment of approximately $10 billion — spread over multiple projects, requiring 1:1 private matching — is insufficient for more than a beginning. The IndiaAI Mission budget of $1.25 billion over five years is insufficient for frontier AI development.
The realistic investment required for India to become a significant participant in the global semiconductor and AI ecosystem — not a leader, but a credible participant — over the next decade is approximately $50-100 billion in public and private investment combined. This is a large number in absolute terms, but it must be placed in perspective:
India's nominal GDP in 2024 was approximately $3.7 trillion. The semiconductor and AI investment required — $50-100 billion over a decade — represents approximately 1.5-3% of a single year's GDP, spread over ten years. South Korea committed comparable resources to its semiconductor development in the 1980s and 1990s, relative to its GDP at that time. Taiwan built TSMC — the most valuable company in Asia — with coordinated government-industry effort over three decades. The question is not whether India can afford to make this investment. The question is whether India's leadership understands that it cannot afford not to.
The investment case is also compelling from a return-on-investment perspective. A $50 billion investment in semiconductor manufacturing that produces $20 billion per year in export revenues after ten years — comparable to what South Korea achieves from Samsung's semiconductor exports — would provide an extraordinary return and would transform India's trade position.
The specific investment priorities should include:
Wafer Fab Capital Subsidy Enhancement: The current 50% capital subsidy for semiconductor fabs is appropriate but may need to be increased to 60-70% for the most advanced technology investments, reflecting the higher risk and strategic value. The subsidy should also be complemented by assured power supply, infrastructure provision, and regulatory fast-tracking.
National Semiconductor Research Fund: A dedicated fund of ₹10,000 crore ($1.2 billion) over five years for basic and applied research in semiconductor science and technology — materials research, process technology development, device physics — at Indian research institutions in partnership with global semiconductor companies.
DRAM Initiative: A specific, dedicated initiative to establish domestic DRAM manufacturing capability — beginning with LPDDR4 (mature technology) but with a clear roadmap to more advanced specifications. This requires finding a strategic partner (Micron or SK Hynix, if US-India or India-Korea semiconductor cooperation agreements can be established) and committing 60-70% capital subsidy for the first fab.
AI Compute Infrastructure: Massively scaling the IndiaAI compute infrastructure — from the planned 10,000 GPUs to at least 50,000-100,000 advanced AI GPUs — through a combination of direct government procurement, public-private partnership, and international cooperation. This would require approximately $5-10 billion in investment but would transform India's AI capability.
The single most important message about India's semiconductor and AI ambitions is that they require resources at a scale that India has not yet committed. Semiconductor manufacturing is the most capital-intensive industry in human history. Getting this right requires confronting the numbers honestly.
Building a competitive 28nm wafer fabrication facility costs approximately $3-5 billion. Building a competitive advanced node fab (7nm or below) costs $15-20 billion. Building a DRAM fab costs $10-15 billion. Building a frontier AI training compute cluster requires $1-5 billion in GPU hardware alone, plus data center infrastructure.
India's total Semicon India commitment of approximately $10 billion — spread over multiple projects, requiring 1:1 private matching — is insufficient for more than a beginning. The IndiaAI Mission budget of $1.25 billion over five years is insufficient for frontier AI development.
The realistic investment required for India to become a significant participant in the global semiconductor and AI ecosystem — not a leader, but a credible participant — over the next decade is approximately $50-100 billion in public and private investment combined. This is a large number in absolute terms, but it must be placed in perspective:
India's nominal GDP in 2024 was approximately $3.7 trillion. The semiconductor and AI investment required — $50-100 billion over a decade — represents approximately 1.5-3% of a single year's GDP, spread over ten years. South Korea committed comparable resources to its semiconductor development in the 1980s and 1990s, relative to its GDP at that time. Taiwan built TSMC — the most valuable company in Asia — with coordinated government-industry effort over three decades. The question is not whether India can afford to make this investment. The question is whether India's leadership understands that it cannot afford not to.
The investment case is also compelling from a return-on-investment perspective. A $50 billion investment in semiconductor manufacturing that produces $20 billion per year in export revenues after ten years — comparable to what South Korea achieves from Samsung's semiconductor exports — would provide an extraordinary return and would transform India's trade position.
The specific investment priorities should include:
Wafer Fab Capital Subsidy Enhancement: The current 50% capital subsidy for semiconductor fabs is appropriate but may need to be increased to 60-70% for the most advanced technology investments, reflecting the higher risk and strategic value. The subsidy should also be complemented by assured power supply, infrastructure provision, and regulatory fast-tracking.
National Semiconductor Research Fund: A dedicated fund of ₹10,000 crore ($1.2 billion) over five years for basic and applied research in semiconductor science and technology — materials research, process technology development, device physics — at Indian research institutions in partnership with global semiconductor companies.
DRAM Initiative: A specific, dedicated initiative to establish domestic DRAM manufacturing capability — beginning with LPDDR4 (mature technology) but with a clear roadmap to more advanced specifications. This requires finding a strategic partner (Micron or SK Hynix, if US-India or India-Korea semiconductor cooperation agreements can be established) and committing 60-70% capital subsidy for the first fab.
AI Compute Infrastructure: Massively scaling the IndiaAI compute infrastructure — from the planned 10,000 GPUs to at least 50,000-100,000 advanced AI GPUs — through a combination of direct government procurement, public-private partnership, and international cooperation. This would require approximately $5-10 billion in investment but would transform India's AI capability.
The Education and Talent Strategy — Building the Human Capital for Technology Leadership
No amount of financial investment will produce a viable semiconductor and AI industry if India lacks the human capital to support it. The talent gap in semiconductor manufacturing and advanced AI is as serious as the financial gap, and it will take longer to fill. Education and workforce development must be front and center in India's technology strategy.
Semiconductor Process Engineering Education: India needs to dramatically expand education in semiconductor process engineering — the specific skills needed to design, operate, and improve semiconductor manufacturing processes. Currently, only a handful of Indian universities have programs that even come close to this specialization. A national effort to establish Semiconductor Process Engineering Centers of Excellence — ideally co-located with or near actual semiconductor facilities — is required. These centers should have hands-on cleanroom facilities, advanced characterization equipment, and industry partnerships that provide students with practical training.
The IITs are obvious anchors for this effort, but the scale required goes far beyond what the IITs alone can deliver. Regional technical universities — NITs, state engineering colleges — need to be upgraded to offer relevant programs. A national scholarship program specifically for semiconductor engineering would attract top talent to this field, which currently has a glamour deficit compared to software and data science.
AI and ML Research Capacity: India needs to dramatically expand the number of PhD-level researchers working on AI fundamentals — architecture development, training methodology, reasoning, safety, and interpretability. Currently, the number of such researchers in India is a small fraction of what exists in the United States or China. Building PhD programs, providing competitive fellowships, and establishing AI research institutes that can retain top researchers in India (rather than losing them to Silicon Valley) is essential.
The brain drain problem is real: India produces many of the world's best AI researchers, but most of them build their careers at American or European companies and universities. Creating conditions that make it attractive for them to work in India — competitive salaries (through industry partnerships), world-class research environments, and meaningful research problems — is essential.
Apprenticeship and Technical Training: The semiconductor industry requires not just PhDs and engineers but a large workforce of skilled technicians — people who can operate lithography machines, maintain etch chambers, perform precision metrology, and manage clean room protocols. This is a different skill level from research or design, but it is critical for manufacturing at scale. South Korea and Taiwan have extensive vocational and technical training programs specifically designed for semiconductor manufacturing. India needs to build equivalent programs — ideally in partnership with companies like Micron and Tata Semiconductors who are investing in Indian facilities and will need this workforce.
Semiconductor Design Training: Chip design using modern EDA tools is a skill that can be taught at scale and built on India's existing software engineering talent base. Expanding VLSI design education — the curriculum that teaches how to design chips using hardware description languages (VHDL, Verilog, SystemVerilog) and EDA tools — across the IITs, NITs, and private engineering colleges would dramatically expand the pool of chip design talent. Companies like Synopsys and Cadence have academic programs that make their tools available to universities; India should maximize utilization of these programs.
No amount of financial investment will produce a viable semiconductor and AI industry if India lacks the human capital to support it. The talent gap in semiconductor manufacturing and advanced AI is as serious as the financial gap, and it will take longer to fill. Education and workforce development must be front and center in India's technology strategy.
Semiconductor Process Engineering Education: India needs to dramatically expand education in semiconductor process engineering — the specific skills needed to design, operate, and improve semiconductor manufacturing processes. Currently, only a handful of Indian universities have programs that even come close to this specialization. A national effort to establish Semiconductor Process Engineering Centers of Excellence — ideally co-located with or near actual semiconductor facilities — is required. These centers should have hands-on cleanroom facilities, advanced characterization equipment, and industry partnerships that provide students with practical training.
The IITs are obvious anchors for this effort, but the scale required goes far beyond what the IITs alone can deliver. Regional technical universities — NITs, state engineering colleges — need to be upgraded to offer relevant programs. A national scholarship program specifically for semiconductor engineering would attract top talent to this field, which currently has a glamour deficit compared to software and data science.
AI and ML Research Capacity: India needs to dramatically expand the number of PhD-level researchers working on AI fundamentals — architecture development, training methodology, reasoning, safety, and interpretability. Currently, the number of such researchers in India is a small fraction of what exists in the United States or China. Building PhD programs, providing competitive fellowships, and establishing AI research institutes that can retain top researchers in India (rather than losing them to Silicon Valley) is essential.
The brain drain problem is real: India produces many of the world's best AI researchers, but most of them build their careers at American or European companies and universities. Creating conditions that make it attractive for them to work in India — competitive salaries (through industry partnerships), world-class research environments, and meaningful research problems — is essential.
Apprenticeship and Technical Training: The semiconductor industry requires not just PhDs and engineers but a large workforce of skilled technicians — people who can operate lithography machines, maintain etch chambers, perform precision metrology, and manage clean room protocols. This is a different skill level from research or design, but it is critical for manufacturing at scale. South Korea and Taiwan have extensive vocational and technical training programs specifically designed for semiconductor manufacturing. India needs to build equivalent programs — ideally in partnership with companies like Micron and Tata Semiconductors who are investing in Indian facilities and will need this workforce.
Semiconductor Design Training: Chip design using modern EDA tools is a skill that can be taught at scale and built on India's existing software engineering talent base. Expanding VLSI design education — the curriculum that teaches how to design chips using hardware description languages (VHDL, Verilog, SystemVerilog) and EDA tools — across the IITs, NITs, and private engineering colleges would dramatically expand the pool of chip design talent. Companies like Synopsys and Cadence have academic programs that make their tools available to universities; India should maximize utilization of these programs.
The Institutional Architecture — Building Organizations That Can Execute
One of the deepest problems with India's semiconductor and technology strategy has been institutional: the right policies have sometimes been formulated but the institutions capable of implementing them effectively have been absent. Building the right institutional architecture is as important as getting the policies right.
India Semiconductor Mission — Empowerment and Resources: The India Semiconductor Mission (ISM) has been established as the nodal agency, but it needs to be significantly empowered — with the ability to make rapid decisions, negotiate directly with foreign companies, coordinate across multiple ministries (Commerce, Finance, Education, Science and Technology, Labour), and hold state governments accountable for their commitments. The ISM should have a CEO with real decision-making authority, a staff of senior technical and commercial experts (including people drawn from the private sector on short-term appointments), and a direct reporting line to the Prime Minister's Office for strategic decisions.
Technology Acquisition Strategy: India needs a sophisticated strategy for acquiring semiconductor technology — whether through licensing, joint ventures, talent acquisition, or research partnerships. This requires an institutional capability for technology intelligence — understanding where the global technology frontier is, which specific technologies India needs, and which potential partners are willing to license or transfer technology on what terms. The ISM should have a dedicated Technology Intelligence unit staffed by people with deep technical and commercial expertise.
Regulatory Fast-Tracking: Semiconductor fabs require numerous environmental, land use, and infrastructure approvals that currently move through India's bureaucratic system at a pace incompatible with commercial timelines. Investors consider India's regulatory environment a significant risk factor. A "Semiconductor Corridor" regulatory regime — analogous to the special regulatory treatment given to defense manufacturing under the "Defence Industrial Corridors" initiative — with single-window clearance, time-bound approvals, and guaranteed infrastructure provision would be a major competitive advantage.
Research-Industry Linkage: India needs stronger mechanisms for translating academic research into industrial application. The current gap between university research and industry is wide — universities produce research papers while industries import foreign technology rather than licensing domestic research. Institutional mechanisms for research-industry collaboration — similar to IMEC in Belgium (a world-leading semiconductor research institute that bridges academic research and industrial application) — would be valuable.
Geopolitical Negotiation Capacity: Building domestic semiconductor capability requires extensive international negotiations — technology transfer agreements with TSMC or Samsung, participation in US CHIPS Act allied partnerships, cooperation with Japan's semiconductor revitalization program, and engagement with the Netherlands regarding semiconductor equipment. These negotiations require sophisticated technical and commercial expertise combined with diplomatic skill. India's diplomatic service has excellent generalists but may lack the technical specialists needed for semiconductor diplomacy. Building this capacity — whether in the Foreign Ministry, the Ministry of Commerce, or the ISM — is essential.
One of the deepest problems with India's semiconductor and technology strategy has been institutional: the right policies have sometimes been formulated but the institutions capable of implementing them effectively have been absent. Building the right institutional architecture is as important as getting the policies right.
India Semiconductor Mission — Empowerment and Resources: The India Semiconductor Mission (ISM) has been established as the nodal agency, but it needs to be significantly empowered — with the ability to make rapid decisions, negotiate directly with foreign companies, coordinate across multiple ministries (Commerce, Finance, Education, Science and Technology, Labour), and hold state governments accountable for their commitments. The ISM should have a CEO with real decision-making authority, a staff of senior technical and commercial experts (including people drawn from the private sector on short-term appointments), and a direct reporting line to the Prime Minister's Office for strategic decisions.
Technology Acquisition Strategy: India needs a sophisticated strategy for acquiring semiconductor technology — whether through licensing, joint ventures, talent acquisition, or research partnerships. This requires an institutional capability for technology intelligence — understanding where the global technology frontier is, which specific technologies India needs, and which potential partners are willing to license or transfer technology on what terms. The ISM should have a dedicated Technology Intelligence unit staffed by people with deep technical and commercial expertise.
Regulatory Fast-Tracking: Semiconductor fabs require numerous environmental, land use, and infrastructure approvals that currently move through India's bureaucratic system at a pace incompatible with commercial timelines. Investors consider India's regulatory environment a significant risk factor. A "Semiconductor Corridor" regulatory regime — analogous to the special regulatory treatment given to defense manufacturing under the "Defence Industrial Corridors" initiative — with single-window clearance, time-bound approvals, and guaranteed infrastructure provision would be a major competitive advantage.
Research-Industry Linkage: India needs stronger mechanisms for translating academic research into industrial application. The current gap between university research and industry is wide — universities produce research papers while industries import foreign technology rather than licensing domestic research. Institutional mechanisms for research-industry collaboration — similar to IMEC in Belgium (a world-leading semiconductor research institute that bridges academic research and industrial application) — would be valuable.
Geopolitical Negotiation Capacity: Building domestic semiconductor capability requires extensive international negotiations — technology transfer agreements with TSMC or Samsung, participation in US CHIPS Act allied partnerships, cooperation with Japan's semiconductor revitalization program, and engagement with the Netherlands regarding semiconductor equipment. These negotiations require sophisticated technical and commercial expertise combined with diplomatic skill. India's diplomatic service has excellent generalists but may lack the technical specialists needed for semiconductor diplomacy. Building this capacity — whether in the Foreign Ministry, the Ministry of Commerce, or the ISM — is essential.
The International Partnerships Strategy
No country builds advanced semiconductor capability entirely alone in the modern era — the technology is too complex and globally distributed. India's path to semiconductor capability runs through a carefully structured set of international partnerships.
The United States Partnership: The US-India relationship is India's most important strategic technology partnership. The Initiative on Critical and Emerging Technology (iCET), launched by President Biden and Prime Minister Modi in early 2023, specifically identified semiconductors and AI as priority areas for collaboration. Under iCET, there is potential for:
• US technical assistance in developing India's semiconductor ecosystem
• Access to US government-funded semiconductor research (DARPA, NIST)
• Potential CHIPS Act-adjacent partnerships — if India can be positioned as a "trusted partner" in the US semiconductor security framework, this could unlock technology access and potentially even equipment access at preferential terms
• Space for Indian companies in US government AI procurement
The US also has leverage over TSMC and Samsung's technology transfer decisions — if the US government signals support for TSMC or Samsung establishing more advanced partnerships with Indian companies, this would significantly accelerate India's technology access.
The Japan Partnership: Japan is a natural partner for India in semiconductors, both because of the strategic alignment between the two countries (both democracies, both concerned about China's rise) and because of Japan's strengths in exactly the areas where India needs help: semiconductor materials, equipment, and mature-node chip manufacturing. India-Japan semiconductor cooperation could include:
• Japanese semiconductor equipment companies establishing manufacturing or maintenance centers in India
• Joint development of semiconductor materials with Japanese partners
• Renesas (which has already partnered for an OSAT in India) potentially deepening to chip manufacturing
• Technology transfer from Japanese chip designers (Socionext, Toshiba Electronic Devices) for mature-node products
The South Korean Partnership: Samsung and SK Hynix are the key South Korean companies whose engagement India needs most desperately for DRAM. India-South Korea semiconductor cooperation has been limited — both companies have focused their manufacturing investments on South Korea, China (in SK Hynix's case), and (for Samsung) Vietnam and other Southeast Asian locations. Attracting significant Samsung or SK Hynix investment to India — particularly for memory manufacturing — would require extraordinary incentives and political-level commitment from both sides.
The Taiwan Consideration: TSMC is the most valuable potential partner for India's foundry ambitions, and the Tata-PSMC partnership (with PSMC being a Taiwanese foundry) represents a first step. Deepening the India-Taiwan semiconductor relationship — potentially to include TSMC engagement (which would require careful navigation of geopolitical sensitivities, as India maintains formal relations with the PRC rather than the Republic of China) — would be a major strategic achievement.
The European Dimension: The European Union's Chips Act and India's interests align in several ways. European semiconductor equipment companies (ASML) and materials companies are critical to India's fab aspirations. European chip designers (Infineon, STMicroelectronics, NXP) design the automotive and industrial chips that India will increasingly need as it develops its EV and industrial sectors. An India-EU semiconductor partnership agreement — building on the broader Trade and Technology Council that the EU has established with the US — would be valuable.
No country builds advanced semiconductor capability entirely alone in the modern era — the technology is too complex and globally distributed. India's path to semiconductor capability runs through a carefully structured set of international partnerships.
The United States Partnership: The US-India relationship is India's most important strategic technology partnership. The Initiative on Critical and Emerging Technology (iCET), launched by President Biden and Prime Minister Modi in early 2023, specifically identified semiconductors and AI as priority areas for collaboration. Under iCET, there is potential for:
• US technical assistance in developing India's semiconductor ecosystem
• Access to US government-funded semiconductor research (DARPA, NIST)
• Potential CHIPS Act-adjacent partnerships — if India can be positioned as a "trusted partner" in the US semiconductor security framework, this could unlock technology access and potentially even equipment access at preferential terms
• Space for Indian companies in US government AI procurement
The US also has leverage over TSMC and Samsung's technology transfer decisions — if the US government signals support for TSMC or Samsung establishing more advanced partnerships with Indian companies, this would significantly accelerate India's technology access.
The Japan Partnership: Japan is a natural partner for India in semiconductors, both because of the strategic alignment between the two countries (both democracies, both concerned about China's rise) and because of Japan's strengths in exactly the areas where India needs help: semiconductor materials, equipment, and mature-node chip manufacturing. India-Japan semiconductor cooperation could include:
• Japanese semiconductor equipment companies establishing manufacturing or maintenance centers in India
• Joint development of semiconductor materials with Japanese partners
• Renesas (which has already partnered for an OSAT in India) potentially deepening to chip manufacturing
• Technology transfer from Japanese chip designers (Socionext, Toshiba Electronic Devices) for mature-node products
The South Korean Partnership: Samsung and SK Hynix are the key South Korean companies whose engagement India needs most desperately for DRAM. India-South Korea semiconductor cooperation has been limited — both companies have focused their manufacturing investments on South Korea, China (in SK Hynix's case), and (for Samsung) Vietnam and other Southeast Asian locations. Attracting significant Samsung or SK Hynix investment to India — particularly for memory manufacturing — would require extraordinary incentives and political-level commitment from both sides.
The Taiwan Consideration: TSMC is the most valuable potential partner for India's foundry ambitions, and the Tata-PSMC partnership (with PSMC being a Taiwanese foundry) represents a first step. Deepening the India-Taiwan semiconductor relationship — potentially to include TSMC engagement (which would require careful navigation of geopolitical sensitivities, as India maintains formal relations with the PRC rather than the Republic of China) — would be a major strategic achievement.
The European Dimension: The European Union's Chips Act and India's interests align in several ways. European semiconductor equipment companies (ASML) and materials companies are critical to India's fab aspirations. European chip designers (Infineon, STMicroelectronics, NXP) design the automotive and industrial chips that India will increasingly need as it develops its EV and industrial sectors. An India-EU semiconductor partnership agreement — building on the broader Trade and Technology Council that the EU has established with the US — would be valuable.
The DRAM-Specific Strategy — Tackling the Memory Deficit
DRAM deserves a specific strategy, not just inclusion in a general semiconductor policy, because the barriers to entry, the strategic partners available, and the domestic demand dynamics are different from other semiconductor segments.
India's DRAM strategy should proceed in three phases:
Phase 1 (2024-2028) — Foundation and Partnership: India should negotiate a strategic DRAM technology partnership with Micron Technology, building on the relationship established through the Micron OSAT investment in Sanand. A technology licensing agreement with Micron — allowing India to begin manufacturing DRAM at 1Y or 1Z generation nodes (using processes that Micron has already advanced beyond but that represent current state-of-the-art for mature DRAM) — would be the foundation. The government should offer 60-70% capital subsidy for a dedicated DRAM wafer fab in India, with guaranteed domestic procurement from government agencies and defense establishments to provide initial demand security.
Phase 2 (2028-2035) — Scale and Process Development: With a functioning DRAM fab, the focus shifts to scaling production, improving yields, and beginning to develop domestic process capabilities. The fab should target domestic supply of LPDDR4 and LPDDR5 (mobile DRAM) and DDR4 and DDR5 (server DRAM) — the most commonly used specifications in India's growing digital economy. Domestic demand from smartphone assembly (if component localization requirements are strengthened), data centers, and government computing would provide a captive market.
Phase 3 (2035-2045) — Frontier Participation: By the mid-2030s, if Phase 1 and 2 have been successfully executed, India should be positioning to participate in more advanced DRAM development — potentially including High Bandwidth Memory (HBM) development for AI applications. This phase would require domestic R&D investment, international research partnerships, and potentially a second, more advanced DRAM fab.
The HBM dimension is particularly strategic for India's AI ambitions. HBM is the memory technology that makes large-scale AI training possible — it is the reason Nvidia's H100 GPUs can achieve the memory bandwidth required for trillion-parameter model training. If India could develop domestic HBM manufacturing capability — even at one generation behind the frontier — it would dramatically reduce the bottleneck in domestic AI compute infrastructure.
DRAM deserves a specific strategy, not just inclusion in a general semiconductor policy, because the barriers to entry, the strategic partners available, and the domestic demand dynamics are different from other semiconductor segments.
India's DRAM strategy should proceed in three phases:
Phase 1 (2024-2028) — Foundation and Partnership: India should negotiate a strategic DRAM technology partnership with Micron Technology, building on the relationship established through the Micron OSAT investment in Sanand. A technology licensing agreement with Micron — allowing India to begin manufacturing DRAM at 1Y or 1Z generation nodes (using processes that Micron has already advanced beyond but that represent current state-of-the-art for mature DRAM) — would be the foundation. The government should offer 60-70% capital subsidy for a dedicated DRAM wafer fab in India, with guaranteed domestic procurement from government agencies and defense establishments to provide initial demand security.
Phase 2 (2028-2035) — Scale and Process Development: With a functioning DRAM fab, the focus shifts to scaling production, improving yields, and beginning to develop domestic process capabilities. The fab should target domestic supply of LPDDR4 and LPDDR5 (mobile DRAM) and DDR4 and DDR5 (server DRAM) — the most commonly used specifications in India's growing digital economy. Domestic demand from smartphone assembly (if component localization requirements are strengthened), data centers, and government computing would provide a captive market.
Phase 3 (2035-2045) — Frontier Participation: By the mid-2030s, if Phase 1 and 2 have been successfully executed, India should be positioning to participate in more advanced DRAM development — potentially including High Bandwidth Memory (HBM) development for AI applications. This phase would require domestic R&D investment, international research partnerships, and potentially a second, more advanced DRAM fab.
The HBM dimension is particularly strategic for India's AI ambitions. HBM is the memory technology that makes large-scale AI training possible — it is the reason Nvidia's H100 GPUs can achieve the memory bandwidth required for trillion-parameter model training. If India could develop domestic HBM manufacturing capability — even at one generation behind the frontier — it would dramatically reduce the bottleneck in domestic AI compute infrastructure.
The AI Strategy — From Fast Follower to Frontier Participant
India's AI strategy needs to honestly assess what is realistically achievable in what timeframe and build a coherent approach rather than spreading resources thinly.
Tier 1 — Where India Can and Must Lead: Indian language AI is the domain where India has a unique advantage and where global companies are underserving India's needs. India has 22 scheduled languages, hundreds of regional languages, and a vast digital communications ecosystem in languages that are underrepresented in global AI training data. Building the best AI models for Hindi, Tamil, Telugu, Bengali, Marathi, Kannada, and other major Indian languages is both strategically important and commercially attractive. The market for AI applications in Indian languages — healthcare, education, agriculture, financial services, government services — is enormous and underserved.
India should invest heavily in Indian language AI through:
• Massive investment in high-quality Indian language datasets (government documents, books, newspapers, academic texts, healthcare records in multiple languages)
• Dedicated Indian language model training using the national AI compute infrastructure
• Regulatory requirements for government AI applications to use India-approved AI systems with strong multilingual capability
• International partnerships to include Indian language data in global training sets (negotiating with OpenAI, Google, Anthropic to prioritize Indian language capability development)
Tier 2 — AI Applications for India's Priority Sectors: AI for agriculture, healthcare, education, and financial services — sectors where India has massive domestic demand and where AI can have transformational impact — should be priority investment areas. AI for crop disease detection, soil quality assessment, and weather-based farming advice for India's 150+ million farming families; AI for triage and diagnostic assistance in India's overstretched healthcare system; AI for personalized learning in India's schools; AI for credit assessment and financial fraud detection in India's growing digital finance ecosystem — these are high-value, high-impact applications where India can both meet domestic needs and develop globally exportable solutions.
Tier 3 — Strategic AI for Defense and Security: AI for defense and national security applications deserves a dedicated, classified program with sufficient resources and the right governance framework. AI for signals intelligence, satellite image analysis, cyber defense, battlefield logistics, and strategic decision support — these are areas where India must develop domestic capability for obvious strategic reasons, and where access to foreign AI systems cannot be assumed. A dedicated defence AI program, run by DRDO with support from the ISM and with access to the national AI compute infrastructure, is essential.
Tier 4 — Foundational AI Research: India needs to invest in fundamental AI research — new architectures, new training methods, AI safety and interpretability, and the theory underlying AI systems. This is the hardest tier to succeed in because it requires not just money but world-class researchers who could earn far more in Silicon Valley. The strategy requires a combination of: (a) creating world-class research environments at selected Indian institutions, (b) building mechanisms for diaspora researchers to contribute to Indian AI research from wherever they are, (c) international research partnerships with leading global AI labs, and (d) patience — world-class research programs take a decade to build.
PART SEVEN: THE IMPACT ON DAILY LIFE OF INDIANS
India's AI strategy needs to honestly assess what is realistically achievable in what timeframe and build a coherent approach rather than spreading resources thinly.
Tier 1 — Where India Can and Must Lead: Indian language AI is the domain where India has a unique advantage and where global companies are underserving India's needs. India has 22 scheduled languages, hundreds of regional languages, and a vast digital communications ecosystem in languages that are underrepresented in global AI training data. Building the best AI models for Hindi, Tamil, Telugu, Bengali, Marathi, Kannada, and other major Indian languages is both strategically important and commercially attractive. The market for AI applications in Indian languages — healthcare, education, agriculture, financial services, government services — is enormous and underserved.
India should invest heavily in Indian language AI through:
• Massive investment in high-quality Indian language datasets (government documents, books, newspapers, academic texts, healthcare records in multiple languages)
• Dedicated Indian language model training using the national AI compute infrastructure
• Regulatory requirements for government AI applications to use India-approved AI systems with strong multilingual capability
• International partnerships to include Indian language data in global training sets (negotiating with OpenAI, Google, Anthropic to prioritize Indian language capability development)
Tier 2 — AI Applications for India's Priority Sectors: AI for agriculture, healthcare, education, and financial services — sectors where India has massive domestic demand and where AI can have transformational impact — should be priority investment areas. AI for crop disease detection, soil quality assessment, and weather-based farming advice for India's 150+ million farming families; AI for triage and diagnostic assistance in India's overstretched healthcare system; AI for personalized learning in India's schools; AI for credit assessment and financial fraud detection in India's growing digital finance ecosystem — these are high-value, high-impact applications where India can both meet domestic needs and develop globally exportable solutions.
Tier 3 — Strategic AI for Defense and Security: AI for defense and national security applications deserves a dedicated, classified program with sufficient resources and the right governance framework. AI for signals intelligence, satellite image analysis, cyber defense, battlefield logistics, and strategic decision support — these are areas where India must develop domestic capability for obvious strategic reasons, and where access to foreign AI systems cannot be assumed. A dedicated defence AI program, run by DRDO with support from the ISM and with access to the national AI compute infrastructure, is essential.
Tier 4 — Foundational AI Research: India needs to invest in fundamental AI research — new architectures, new training methods, AI safety and interpretability, and the theory underlying AI systems. This is the hardest tier to succeed in because it requires not just money but world-class researchers who could earn far more in Silicon Valley. The strategy requires a combination of: (a) creating world-class research environments at selected Indian institutions, (b) building mechanisms for diaspora researchers to contribute to Indian AI research from wherever they are, (c) international research partnerships with leading global AI labs, and (d) patience — world-class research programs take a decade to build.
PART SEVEN: THE IMPACT ON DAILY LIFE OF INDIANS
How Technology Dependency Shapes Every Indian's Life Today
The discussion of semiconductors and AI can seem abstract — a matter for technologists, economists, and diplomats. But the consequences of India's technology deficit are deeply, personally felt by every Indian every day. Understanding these concrete impacts is essential to building the political will for the difficult, expensive, long-term work of technological development.
The Smartphone and Its Hidden Technology Chain: The smartphone is the most transformative technology in India's recent history. With over 600 million smartphone users, India has used the smartphone to leapfrog many stages of technological development — mobile internet has brought digital services to populations that never had personal computers or fixed-line internet. UPI has made India the global leader in real-time digital payments. WhatsApp has transformed social communication. YouTube has become the primary entertainment medium for hundreds of millions of Indians.
But every smartphone — whether assembled in India or imported — contains a semiconductor supply chain that is entirely foreign. The processor that runs UPI transactions was designed by Qualcomm in San Diego and manufactured by TSMC in Taiwan. The DRAM that stores WhatsApp messages was manufactured by Samsung in South Korea. The NAND flash that stores photos was made by SK Hynix. The display driver chip was made by Novatek in Taiwan. The Wi-Fi chip was made by Qualcomm or MediaTek. The cellular modem was made by Qualcomm.
When India imports 200 million smartphones per year (approximately 50-60 million are now assembled domestically, but with imported components), the billions of dollars that India pays for those phones flow almost entirely to foreign companies and foreign workers. If India manufactured the semiconductor components in those phones domestically, a much larger share of that economic value would be captured domestically — creating Indian jobs, generating Indian tax revenues, and building Indian technological capacity.
The price of smartphones in India also partly reflects the global supply chain's cost structure. If India had domestic semiconductor manufacturing, it could potentially offer lower-cost smartphones to its population, accelerating digital inclusion among the still-unconnected population (approximately 400-500 million Indians have never accessed the internet). Digital inclusion is directly correlated with economic opportunity — access to digital financial services, agricultural information, job matching platforms, and education — so reducing smartphone costs would have cascading positive effects on India's poorest citizens.
Healthcare — AI and Technology in Medicine: India's healthcare system is under enormous strain — with a doctor-to-population ratio of approximately 1:1,456 against the WHO recommendation of 1:1,000, and vast geographic disparities between urban and rural healthcare access. AI offers the potential to dramatically extend the reach of healthcare — AI diagnostic tools that can identify tuberculosis from X-rays, detect diabetic retinopathy from fundus photographs, and triage patients based on symptoms can effectively act as a first layer of healthcare access for populations without convenient access to doctors.
But these AI tools, if developed on foreign AI platforms (which most currently are, given the dominance of US AI models), create dependency and privacy concerns. Developing domestic AI diagnostic tools — using India's enormous healthcare data (with appropriate privacy protections) — would allow India to build systems specifically calibrated to Indian disease patterns, Indian genetic variations, and India's specific healthcare context. Foreign AI models, trained primarily on Western data, may not perform optimally for Indian patients.
The semiconductor dependency also affects India's medical device industry. India imports approximately 85% of its medical devices by value — including ventilators, CT scanners, MRI machines, ICU monitors, and diagnostic laboratory equipment. These devices all rely on semiconductors. During the COVID-19 pandemic, when ventilator demand spiked globally, India's inability to manufacture ventilators domestically — because it lacked domestic semiconductor and component supply chains — was a direct cause of preventable deaths. Building domestic medical device manufacturing capacity requires, as a necessary condition, building domestic semiconductor and component supply.
Agriculture — The Digital Green Revolution: India's agricultural sector — which still employs approximately 45% of the workforce, though contributing only 15-18% of GDP — stands to be transformed by AI and digital technology. Precision agriculture — using AI-powered analysis of satellite data, soil sensors, weather data, and crop images to optimize planting, irrigation, fertilization, and pest control — can dramatically improve agricultural productivity. The technology already exists; the challenge is deploying it at scale for India's 150+ million small and marginal farmers.
Currently, the AI tools most applicable to Indian agriculture are primarily developed by foreign companies (John Deere's AI precision agriculture systems, Microsoft's FarmBeats, Google's agriculture AI projects) or by Indian startups using foreign AI infrastructure. If India had domestic AI capability — particularly for regional language interfaces and models trained on Indian agricultural data — these tools could be more effectively tailored to India's diverse agricultural conditions.
The semiconductor dependency also affects the sensors and IoT devices used in smart agriculture — soil moisture sensors, weather stations, automated irrigation controllers — all of which rely on microcontrollers and connectivity chips that India currently imports entirely.
Financial Services — The UPI Revolution and Its Technology Underpinning: The Unified Payments Interface (UPI) is arguably India's most globally admired technological achievement of the last decade. Processing over 10 billion transactions per month, with average transaction values measured in hundreds of rupees, UPI has transformed India into the world's most active real-time payments market. It has enabled digital financial services to reach populations previously excluded from the formal financial system.
But UPI runs on servers. Those servers use processors made by Intel or AMD (American companies), DRAM made by Samsung or SK Hynix (Korean companies), SSDs made with NAND flash from Japanese and Korean companies, and network equipment made by Cisco or Juniper (American companies) or Huawei (Chinese company, whose use in critical infrastructure raises separate concerns). The National Payments Corporation of India (NPCI) — which operates UPI — runs its infrastructure on imported hardware.
This is not merely a financial vulnerability. It is a digital sovereignty question. India's most critical financial infrastructure — the system through which billions of Indian transactions flow every day — is fundamentally dependent on foreign hardware. If sanctions, export controls, or supply chain disruptions were to prevent India from accessing these hardware components, UPI infrastructure could not be expanded, maintained, or replaced. Understanding this dependency should galvanize action toward domestic semiconductor production — particularly the server-grade processors and memory chips that critical infrastructure depends on.
Education — The AI Tutor and Its Possibilities: India's education system faces enormous challenges: inadequate school infrastructure, teacher shortages particularly in science and mathematics, wide urban-rural quality gaps, and high dropout rates. AI-powered personalized education — systems that can identify each student's learning gaps and provide customized instruction — offers a potentially transformative solution.
But the effectiveness of AI education tools for Indian students depends critically on language. India's students learn in Hindi, Tamil, Telugu, Bengali, Kannada, Malayalam, Marathi, and numerous other languages. AI tutoring systems that function primarily in English are of limited value for the hundreds of millions of Indian students whose primary language is not English. Building high-quality AI education tools for Indian languages — with the depth of knowledge and pedagogical sophistication needed to genuinely help struggling students — requires exactly the kind of Indian language AI infrastructure that India is only beginning to build.
The digital divide in India is starkly visible in education. Urban, English-medium private school students have access to AI tutoring tools, YouTube educational content, and online education platforms that simply do not function as well — if at all — in regional Indian languages. Bridging this gap requires not just internet access and devices but AI systems that genuinely work in the languages those students speak. This is an equity issue as much as a technology issue.
Manufacturing and Jobs — The Industry 4.0 Transition: India's manufacturing sector faces a crucial challenge: automation and AI are transforming manufacturing globally, but India's comparative advantage has traditionally been low labor costs. As automation reduces the relative cost advantage of low labor costs, India must compete on other dimensions — quality, flexibility, innovation — which require more advanced manufacturing technology.
The automotive sector illustrates this clearly. India is the world's third-largest automotive market and a significant automotive manufacturer. The shift to electric vehicles is driving massive demand for automotive semiconductors — power management chips, motor controller chips, battery management chips, ADAS (Advanced Driver Assistance Systems) chips, and display chips. All of these are currently imported. As the Indian automotive industry electrifies, its semiconductor dependency — and vulnerability to supply chain disruptions — will increase.
Industry 4.0 — the integration of IoT sensors, AI-based process optimization, digital twins, and robotics into manufacturing — is being adopted rapidly by global manufacturers but slowly in India. One barrier is the cost and availability of the semiconductor components needed for Industry 4.0 technology — PLCs (Programmable Logic Controllers), IoT gateways, industrial robots — all of which are currently imported. Domestic semiconductor manufacturing, starting with the industrial microcontrollers and communication chips used in these systems, would accelerate India's adoption of Industry 4.0 technologies.
Energy — Semiconductor-Dependent Clean Energy Transition: India has made ambitious commitments to clean energy — 500GW of renewable energy capacity by 2030, a target of 30% electric vehicle penetration by 2030, and a net-zero target by 2070. All of these transitions depend critically on semiconductors.
Solar panels are not pure silicon — they are semiconductor devices. More relevantly, the power electronics that convert solar panel DC output to grid AC, and the battery management systems that manage energy storage, are based on specialized power semiconductors — silicon carbide (SiC) and gallium nitride (GaN) chips — that India currently imports almost entirely. The EV transition similarly requires power electronics (motor drives, onboard chargers, DC-DC converters) and battery management chips. India's clean energy transition will dramatically increase India's semiconductor import dependency unless steps are taken to build domestic power electronics semiconductor manufacturing.
Power semiconductors — SiC and GaN devices specifically — represent an opportunity for India because these are more mature technologies than leading-edge logic, and because India has genuine materials capabilities (silicon carbide crystal growth, for example) that could be leveraged. An India-specific power semiconductor manufacturing initiative — targeting the SiC and GaN chips needed for solar inverters, EV power trains, and industrial motor drives — would directly support India's energy transition while building domestic semiconductor capability.
Defense and National Security — The Most Critical Dependency: India's defense establishment faces perhaps the most acute semiconductor dependency of any sector. Modern weapons systems — missiles, fighter aircraft, warships, armored vehicles, electronic warfare systems, surveillance drones — are, at their core, sophisticated computers. The Tejas fighter jet contains millions of lines of software running on advanced chips. The Brahmos missile is guided by real-time computing systems. The DRDO's radar systems use high-performance signal processing chips. The Indian Navy's sonar systems, the Indian Army's communication infrastructure — all of it runs on imported semiconductors.
During periods of geopolitical tension — precisely when these systems are most needed — supply chain access to critical defense electronics could be restricted. The United States, which is India's primary defense technology partner today, is also the country that wields export control authority over much of the critical semiconductor technology in India's defense systems. While the US-India relationship is currently strong, strategic autonomy requires India to develop domestic supply of critical defense electronics.
India's defence semiconductor requirements span a broad spectrum: high-performance processors for signal processing and navigation; specialized memory for data recording; radiation-hardened chips for space applications; power amplifier chips for radar and communications; and AI accelerator chips for automated threat detection and decision support. Building domestic supply of these components — even if not at the commercial frontier in terms of manufacturing node — is a national security imperative.
PART EIGHT: THE ECONOMIC IMPLICATIONS
The discussion of semiconductors and AI can seem abstract — a matter for technologists, economists, and diplomats. But the consequences of India's technology deficit are deeply, personally felt by every Indian every day. Understanding these concrete impacts is essential to building the political will for the difficult, expensive, long-term work of technological development.
The Smartphone and Its Hidden Technology Chain: The smartphone is the most transformative technology in India's recent history. With over 600 million smartphone users, India has used the smartphone to leapfrog many stages of technological development — mobile internet has brought digital services to populations that never had personal computers or fixed-line internet. UPI has made India the global leader in real-time digital payments. WhatsApp has transformed social communication. YouTube has become the primary entertainment medium for hundreds of millions of Indians.
But every smartphone — whether assembled in India or imported — contains a semiconductor supply chain that is entirely foreign. The processor that runs UPI transactions was designed by Qualcomm in San Diego and manufactured by TSMC in Taiwan. The DRAM that stores WhatsApp messages was manufactured by Samsung in South Korea. The NAND flash that stores photos was made by SK Hynix. The display driver chip was made by Novatek in Taiwan. The Wi-Fi chip was made by Qualcomm or MediaTek. The cellular modem was made by Qualcomm.
When India imports 200 million smartphones per year (approximately 50-60 million are now assembled domestically, but with imported components), the billions of dollars that India pays for those phones flow almost entirely to foreign companies and foreign workers. If India manufactured the semiconductor components in those phones domestically, a much larger share of that economic value would be captured domestically — creating Indian jobs, generating Indian tax revenues, and building Indian technological capacity.
The price of smartphones in India also partly reflects the global supply chain's cost structure. If India had domestic semiconductor manufacturing, it could potentially offer lower-cost smartphones to its population, accelerating digital inclusion among the still-unconnected population (approximately 400-500 million Indians have never accessed the internet). Digital inclusion is directly correlated with economic opportunity — access to digital financial services, agricultural information, job matching platforms, and education — so reducing smartphone costs would have cascading positive effects on India's poorest citizens.
Healthcare — AI and Technology in Medicine: India's healthcare system is under enormous strain — with a doctor-to-population ratio of approximately 1:1,456 against the WHO recommendation of 1:1,000, and vast geographic disparities between urban and rural healthcare access. AI offers the potential to dramatically extend the reach of healthcare — AI diagnostic tools that can identify tuberculosis from X-rays, detect diabetic retinopathy from fundus photographs, and triage patients based on symptoms can effectively act as a first layer of healthcare access for populations without convenient access to doctors.
But these AI tools, if developed on foreign AI platforms (which most currently are, given the dominance of US AI models), create dependency and privacy concerns. Developing domestic AI diagnostic tools — using India's enormous healthcare data (with appropriate privacy protections) — would allow India to build systems specifically calibrated to Indian disease patterns, Indian genetic variations, and India's specific healthcare context. Foreign AI models, trained primarily on Western data, may not perform optimally for Indian patients.
The semiconductor dependency also affects India's medical device industry. India imports approximately 85% of its medical devices by value — including ventilators, CT scanners, MRI machines, ICU monitors, and diagnostic laboratory equipment. These devices all rely on semiconductors. During the COVID-19 pandemic, when ventilator demand spiked globally, India's inability to manufacture ventilators domestically — because it lacked domestic semiconductor and component supply chains — was a direct cause of preventable deaths. Building domestic medical device manufacturing capacity requires, as a necessary condition, building domestic semiconductor and component supply.
Agriculture — The Digital Green Revolution: India's agricultural sector — which still employs approximately 45% of the workforce, though contributing only 15-18% of GDP — stands to be transformed by AI and digital technology. Precision agriculture — using AI-powered analysis of satellite data, soil sensors, weather data, and crop images to optimize planting, irrigation, fertilization, and pest control — can dramatically improve agricultural productivity. The technology already exists; the challenge is deploying it at scale for India's 150+ million small and marginal farmers.
Currently, the AI tools most applicable to Indian agriculture are primarily developed by foreign companies (John Deere's AI precision agriculture systems, Microsoft's FarmBeats, Google's agriculture AI projects) or by Indian startups using foreign AI infrastructure. If India had domestic AI capability — particularly for regional language interfaces and models trained on Indian agricultural data — these tools could be more effectively tailored to India's diverse agricultural conditions.
The semiconductor dependency also affects the sensors and IoT devices used in smart agriculture — soil moisture sensors, weather stations, automated irrigation controllers — all of which rely on microcontrollers and connectivity chips that India currently imports entirely.
Financial Services — The UPI Revolution and Its Technology Underpinning: The Unified Payments Interface (UPI) is arguably India's most globally admired technological achievement of the last decade. Processing over 10 billion transactions per month, with average transaction values measured in hundreds of rupees, UPI has transformed India into the world's most active real-time payments market. It has enabled digital financial services to reach populations previously excluded from the formal financial system.
But UPI runs on servers. Those servers use processors made by Intel or AMD (American companies), DRAM made by Samsung or SK Hynix (Korean companies), SSDs made with NAND flash from Japanese and Korean companies, and network equipment made by Cisco or Juniper (American companies) or Huawei (Chinese company, whose use in critical infrastructure raises separate concerns). The National Payments Corporation of India (NPCI) — which operates UPI — runs its infrastructure on imported hardware.
This is not merely a financial vulnerability. It is a digital sovereignty question. India's most critical financial infrastructure — the system through which billions of Indian transactions flow every day — is fundamentally dependent on foreign hardware. If sanctions, export controls, or supply chain disruptions were to prevent India from accessing these hardware components, UPI infrastructure could not be expanded, maintained, or replaced. Understanding this dependency should galvanize action toward domestic semiconductor production — particularly the server-grade processors and memory chips that critical infrastructure depends on.
Education — The AI Tutor and Its Possibilities: India's education system faces enormous challenges: inadequate school infrastructure, teacher shortages particularly in science and mathematics, wide urban-rural quality gaps, and high dropout rates. AI-powered personalized education — systems that can identify each student's learning gaps and provide customized instruction — offers a potentially transformative solution.
But the effectiveness of AI education tools for Indian students depends critically on language. India's students learn in Hindi, Tamil, Telugu, Bengali, Kannada, Malayalam, Marathi, and numerous other languages. AI tutoring systems that function primarily in English are of limited value for the hundreds of millions of Indian students whose primary language is not English. Building high-quality AI education tools for Indian languages — with the depth of knowledge and pedagogical sophistication needed to genuinely help struggling students — requires exactly the kind of Indian language AI infrastructure that India is only beginning to build.
The digital divide in India is starkly visible in education. Urban, English-medium private school students have access to AI tutoring tools, YouTube educational content, and online education platforms that simply do not function as well — if at all — in regional Indian languages. Bridging this gap requires not just internet access and devices but AI systems that genuinely work in the languages those students speak. This is an equity issue as much as a technology issue.
Manufacturing and Jobs — The Industry 4.0 Transition: India's manufacturing sector faces a crucial challenge: automation and AI are transforming manufacturing globally, but India's comparative advantage has traditionally been low labor costs. As automation reduces the relative cost advantage of low labor costs, India must compete on other dimensions — quality, flexibility, innovation — which require more advanced manufacturing technology.
The automotive sector illustrates this clearly. India is the world's third-largest automotive market and a significant automotive manufacturer. The shift to electric vehicles is driving massive demand for automotive semiconductors — power management chips, motor controller chips, battery management chips, ADAS (Advanced Driver Assistance Systems) chips, and display chips. All of these are currently imported. As the Indian automotive industry electrifies, its semiconductor dependency — and vulnerability to supply chain disruptions — will increase.
Industry 4.0 — the integration of IoT sensors, AI-based process optimization, digital twins, and robotics into manufacturing — is being adopted rapidly by global manufacturers but slowly in India. One barrier is the cost and availability of the semiconductor components needed for Industry 4.0 technology — PLCs (Programmable Logic Controllers), IoT gateways, industrial robots — all of which are currently imported. Domestic semiconductor manufacturing, starting with the industrial microcontrollers and communication chips used in these systems, would accelerate India's adoption of Industry 4.0 technologies.
Energy — Semiconductor-Dependent Clean Energy Transition: India has made ambitious commitments to clean energy — 500GW of renewable energy capacity by 2030, a target of 30% electric vehicle penetration by 2030, and a net-zero target by 2070. All of these transitions depend critically on semiconductors.
Solar panels are not pure silicon — they are semiconductor devices. More relevantly, the power electronics that convert solar panel DC output to grid AC, and the battery management systems that manage energy storage, are based on specialized power semiconductors — silicon carbide (SiC) and gallium nitride (GaN) chips — that India currently imports almost entirely. The EV transition similarly requires power electronics (motor drives, onboard chargers, DC-DC converters) and battery management chips. India's clean energy transition will dramatically increase India's semiconductor import dependency unless steps are taken to build domestic power electronics semiconductor manufacturing.
Power semiconductors — SiC and GaN devices specifically — represent an opportunity for India because these are more mature technologies than leading-edge logic, and because India has genuine materials capabilities (silicon carbide crystal growth, for example) that could be leveraged. An India-specific power semiconductor manufacturing initiative — targeting the SiC and GaN chips needed for solar inverters, EV power trains, and industrial motor drives — would directly support India's energy transition while building domestic semiconductor capability.
Defense and National Security — The Most Critical Dependency: India's defense establishment faces perhaps the most acute semiconductor dependency of any sector. Modern weapons systems — missiles, fighter aircraft, warships, armored vehicles, electronic warfare systems, surveillance drones — are, at their core, sophisticated computers. The Tejas fighter jet contains millions of lines of software running on advanced chips. The Brahmos missile is guided by real-time computing systems. The DRDO's radar systems use high-performance signal processing chips. The Indian Navy's sonar systems, the Indian Army's communication infrastructure — all of it runs on imported semiconductors.
During periods of geopolitical tension — precisely when these systems are most needed — supply chain access to critical defense electronics could be restricted. The United States, which is India's primary defense technology partner today, is also the country that wields export control authority over much of the critical semiconductor technology in India's defense systems. While the US-India relationship is currently strong, strategic autonomy requires India to develop domestic supply of critical defense electronics.
India's defence semiconductor requirements span a broad spectrum: high-performance processors for signal processing and navigation; specialized memory for data recording; radiation-hardened chips for space applications; power amplifier chips for radar and communications; and AI accelerator chips for automated threat detection and decision support. Building domestic supply of these components — even if not at the commercial frontier in terms of manufacturing node — is a national security imperative.
PART EIGHT: THE ECONOMIC IMPLICATIONS
The Trade Deficit — Technology as India's Hidden Current Account Drain
India's current account deficit — the excess of imports over exports — is one of the persistent macroeconomic vulnerabilities of the Indian economy. Energy (oil and gas) and gold have traditionally been the largest contributors to India's import bill. But electronics has emerged as the third-largest import category, and the semiconductor component of the electronics import bill is growing rapidly.
India's electronics imports were approximately $70 billion in FY 2022-23. This includes imported mobile phones (approximately $10 billion, although growing domestic assembly is reducing this), imported components for domestic electronics assembly (approximately $30-40 billion, including PCBs, displays, and semiconductors), and imported consumer and industrial electronics. The semiconductor component of these electronics imports — including discrete chips, integrated circuits, and memory modules — was approximately $25-30 billion.
For perspective: India's IT services exports — which are celebrated as a major economic success — were approximately $227 billion in FY 2022-23. But a portion of this is offset by electronics imports. India's net technology trade balance (IT services exports minus electronics imports) is positive, but the electronics import bill is growing faster than IT services exports — meaning the net position is deteriorating.
If India had domestic semiconductor manufacturing capability — even at a modest scale — the trade implications would be significant. Every dollar of semiconductor production substituting for imports directly improves India's current account. At the scale of production that the Tata-PSMC fab represents (28nm mature node chips), the import substitution effect might be on the order of $2-3 billion per year by 2030. Larger-scale semiconductor manufacturing — approaching South Korea's level, which exports approximately $130 billion per year in semiconductors — would be transformatively positive for India's trade position.
The employment creation implications are also significant. Each semiconductor fab creates approximately 1,500-2,500 direct jobs in high-skill manufacturing, plus 3-5x indirect jobs in the supply chain and services ecosystem. But more importantly, semiconductor manufacturing creates the ecosystem of suppliers, engineers, and institutions that enables further technology development — the multiplier effects extend far beyond direct employment.
India's current account deficit — the excess of imports over exports — is one of the persistent macroeconomic vulnerabilities of the Indian economy. Energy (oil and gas) and gold have traditionally been the largest contributors to India's import bill. But electronics has emerged as the third-largest import category, and the semiconductor component of the electronics import bill is growing rapidly.
India's electronics imports were approximately $70 billion in FY 2022-23. This includes imported mobile phones (approximately $10 billion, although growing domestic assembly is reducing this), imported components for domestic electronics assembly (approximately $30-40 billion, including PCBs, displays, and semiconductors), and imported consumer and industrial electronics. The semiconductor component of these electronics imports — including discrete chips, integrated circuits, and memory modules — was approximately $25-30 billion.
For perspective: India's IT services exports — which are celebrated as a major economic success — were approximately $227 billion in FY 2022-23. But a portion of this is offset by electronics imports. India's net technology trade balance (IT services exports minus electronics imports) is positive, but the electronics import bill is growing faster than IT services exports — meaning the net position is deteriorating.
If India had domestic semiconductor manufacturing capability — even at a modest scale — the trade implications would be significant. Every dollar of semiconductor production substituting for imports directly improves India's current account. At the scale of production that the Tata-PSMC fab represents (28nm mature node chips), the import substitution effect might be on the order of $2-3 billion per year by 2030. Larger-scale semiconductor manufacturing — approaching South Korea's level, which exports approximately $130 billion per year in semiconductors — would be transformatively positive for India's trade position.
The employment creation implications are also significant. Each semiconductor fab creates approximately 1,500-2,500 direct jobs in high-skill manufacturing, plus 3-5x indirect jobs in the supply chain and services ecosystem. But more importantly, semiconductor manufacturing creates the ecosystem of suppliers, engineers, and institutions that enables further technology development — the multiplier effects extend far beyond direct employment.
The Innovation Economy — How Semiconductor Capability Enables Startup Success
The relationship between semiconductor manufacturing capability and startup innovation is not immediately obvious but is deeply important. Countries and regions with significant semiconductor manufacturing capability — Silicon Valley, Taiwan, South Korea, Israel — have consistently produced a disproportionate share of technology innovation. This is not coincidence.
Semiconductor manufacturing capability creates several conditions that enable innovation:
Engineering Talent Density: Areas with semiconductor manufacturing attract and develop a specific type of highly skilled engineering talent — people who understand hardware at a deep level. This talent is the seed of hardware-focused startup ecosystems. ARM was founded by engineers with deep hardware backgrounds. NVIDIA was founded by engineers who had worked in the semiconductor industry. The dense concentration of semiconductor engineers in Silicon Valley was a prerequisite for the hardware innovation ecosystem that emerged there.
India's relative absence from semiconductor manufacturing means India lacks this talent density in hardware engineering. India has abundant software engineers but relatively few hardware engineers — engineers who design chips, develop embedded systems, or understand the physics of electronic devices at a deep level. Consequently, Indian startups tend to be software-focused — apps, platforms, SaaS products — rather than hardware-focused. This is not a cultural or capability limitation; it is a structural consequence of India's industry composition. Building semiconductor manufacturing would build hardware engineering talent and enable a hardware startup ecosystem.
Design-Manufacturing Proximity: Chip designers who are physically close to manufacturing facilities have significant advantages — they can iterate designs more quickly, develop deeper understanding of manufacturing constraints, and create designs that take better advantage of specific process capabilities. TSMC's extraordinary success as a foundry is partly attributable to the tight feedback loop between chip designers in Taiwan and TSMC's fabs. India's semiconductor design community, largely working for global companies' captive design centers, does not benefit from this proximity because the fabs are all offshore.
Supply Chain Ecosystem: Manufacturing semiconductor chips requires a supply chain of materials, components, and services. The ecosystem of companies that develops around semiconductor manufacturing — equipment service companies, materials suppliers, specialized logistics providers, calibration and metrology services — creates a dense web of economic activity that supports broad industrial development. Building semiconductor manufacturing in India would trigger the development of this ecosystem, creating opportunities for Indian companies in adjacent industries.
IP Generation: Manufacturing new semiconductors requires the development of new intellectual property — process patents, design patents, trade secrets. Companies and countries that manufacture semiconductors accumulate IP portfolios that have significant economic value and provide competitive moats. India's current focus on IT services generates relatively little fundamental IP — service businesses build skill and reputation rather than IP-based competitive advantages. Semiconductor manufacturing would enable IP generation that compounds over time, creating durable competitive advantages.
The relationship between semiconductor manufacturing capability and startup innovation is not immediately obvious but is deeply important. Countries and regions with significant semiconductor manufacturing capability — Silicon Valley, Taiwan, South Korea, Israel — have consistently produced a disproportionate share of technology innovation. This is not coincidence.
Semiconductor manufacturing capability creates several conditions that enable innovation:
Engineering Talent Density: Areas with semiconductor manufacturing attract and develop a specific type of highly skilled engineering talent — people who understand hardware at a deep level. This talent is the seed of hardware-focused startup ecosystems. ARM was founded by engineers with deep hardware backgrounds. NVIDIA was founded by engineers who had worked in the semiconductor industry. The dense concentration of semiconductor engineers in Silicon Valley was a prerequisite for the hardware innovation ecosystem that emerged there.
India's relative absence from semiconductor manufacturing means India lacks this talent density in hardware engineering. India has abundant software engineers but relatively few hardware engineers — engineers who design chips, develop embedded systems, or understand the physics of electronic devices at a deep level. Consequently, Indian startups tend to be software-focused — apps, platforms, SaaS products — rather than hardware-focused. This is not a cultural or capability limitation; it is a structural consequence of India's industry composition. Building semiconductor manufacturing would build hardware engineering talent and enable a hardware startup ecosystem.
Design-Manufacturing Proximity: Chip designers who are physically close to manufacturing facilities have significant advantages — they can iterate designs more quickly, develop deeper understanding of manufacturing constraints, and create designs that take better advantage of specific process capabilities. TSMC's extraordinary success as a foundry is partly attributable to the tight feedback loop between chip designers in Taiwan and TSMC's fabs. India's semiconductor design community, largely working for global companies' captive design centers, does not benefit from this proximity because the fabs are all offshore.
Supply Chain Ecosystem: Manufacturing semiconductor chips requires a supply chain of materials, components, and services. The ecosystem of companies that develops around semiconductor manufacturing — equipment service companies, materials suppliers, specialized logistics providers, calibration and metrology services — creates a dense web of economic activity that supports broad industrial development. Building semiconductor manufacturing in India would trigger the development of this ecosystem, creating opportunities for Indian companies in adjacent industries.
IP Generation: Manufacturing new semiconductors requires the development of new intellectual property — process patents, design patents, trade secrets. Companies and countries that manufacture semiconductors accumulate IP portfolios that have significant economic value and provide competitive moats. India's current focus on IT services generates relatively little fundamental IP — service businesses build skill and reputation rather than IP-based competitive advantages. Semiconductor manufacturing would enable IP generation that compounds over time, creating durable competitive advantages.
The GDP Growth Model — Technology as the New Productivity Driver
India's economic ambitions are well-known: Prime Minister Modi's vision of a "Developed India" (Viksit Bharat) by 2047 implies that India should achieve high-income status — typically defined as GNI per capita above $13,000 — within 25 years. India's per capita GNI in 2023 was approximately $2,400. Achieving high-income status by 2047 requires compounding per capita income at approximately 7-8% per year for 25 years.
Sustaining that growth rate requires continuously increasing productivity — producing more output per unit of labor and capital input. Technology is the primary driver of productivity improvement at the national level over long periods. The countries that have maintained high growth rates for decades — South Korea, Taiwan, Singapore — have done so by continuously upgrading their technological capabilities, moving from low-skill manufacturing to high-skill manufacturing to technology development. India must follow a similar trajectory.
The semiconductor and AI sectors are central to this productivity upgrading story:
Direct Productivity Contribution: Semiconductor manufacturing and AI development are high-productivity industries — they create enormous output value relative to the number of workers employed. A semiconductor fab employing 2,000 workers might generate $1-2 billion in annual revenue, implying revenue per employee of $500,000-$1,000,000. This is dramatically higher than the productivity of most sectors that currently employ most Indians.
Indirect Productivity Spillovers: AI adoption across India's economy — in agriculture, manufacturing, services, and government — would generate productivity improvements that are difficult to estimate precisely but could be very large. McKinsey estimates that AI could add $1.5-3 trillion to India's GDP by 2035. These estimates are speculative but indicative of the scale of potential productivity impact.
Export Diversification: India's export base is currently too concentrated in IT services (which face potential disruption from AI automation), pharmaceuticals, and commodities. Semiconductor and electronics manufacturing would diversify India's export base, making the economy more resilient and adding high-value export sectors that support sustained current account improvement.
PART NINE: THE GLOBAL ORDER RESHAPING AND INDIA'S POSITION
India's economic ambitions are well-known: Prime Minister Modi's vision of a "Developed India" (Viksit Bharat) by 2047 implies that India should achieve high-income status — typically defined as GNI per capita above $13,000 — within 25 years. India's per capita GNI in 2023 was approximately $2,400. Achieving high-income status by 2047 requires compounding per capita income at approximately 7-8% per year for 25 years.
Sustaining that growth rate requires continuously increasing productivity — producing more output per unit of labor and capital input. Technology is the primary driver of productivity improvement at the national level over long periods. The countries that have maintained high growth rates for decades — South Korea, Taiwan, Singapore — have done so by continuously upgrading their technological capabilities, moving from low-skill manufacturing to high-skill manufacturing to technology development. India must follow a similar trajectory.
The semiconductor and AI sectors are central to this productivity upgrading story:
Direct Productivity Contribution: Semiconductor manufacturing and AI development are high-productivity industries — they create enormous output value relative to the number of workers employed. A semiconductor fab employing 2,000 workers might generate $1-2 billion in annual revenue, implying revenue per employee of $500,000-$1,000,000. This is dramatically higher than the productivity of most sectors that currently employ most Indians.
Indirect Productivity Spillovers: AI adoption across India's economy — in agriculture, manufacturing, services, and government — would generate productivity improvements that are difficult to estimate precisely but could be very large. McKinsey estimates that AI could add $1.5-3 trillion to India's GDP by 2035. These estimates are speculative but indicative of the scale of potential productivity impact.
Export Diversification: India's export base is currently too concentrated in IT services (which face potential disruption from AI automation), pharmaceuticals, and commodities. Semiconductor and electronics manufacturing would diversify India's export base, making the economy more resilient and adding high-value export sectors that support sustained current account improvement.
PART NINE: THE GLOBAL ORDER RESHAPING AND INDIA'S POSITION
Technology and the New Geopolitics — Understanding the Power Shift
The 21st century's central geopolitical competition is a competition for technological supremacy. This represents a fundamental shift from the Cold War's nuclear-based bipolarity or the post-Cold War's focus on liberal democratic order. The new ordering principle of global power is technological capability — the ability to develop, produce, and deploy the technologies that determine economic productivity, military effectiveness, and cultural influence.
The United States has dominated the global technology landscape since the Second World War — not because of any particular cultural superiority but because of a specific institutional ecology: a well-funded research university system (MIT, Stanford, Caltech, Harvard, Cornell), a military-industrial research complex (DARPA, ARPA-E, national laboratories), venture capital markets willing to fund high-risk technological bets, open immigration attracting global talent, and a market large enough to provide the initial scale for technology companies to achieve global reach.
China's rise as a technology power — while still incomplete and contested — represents the first serious challenge to US technological supremacy since the Soviet space race. China's approach has been different from the US model: centralized state investment and coordination, massive scale of engineering education output (producing more STEM graduates per year than the US), aggressive technology acquisition through joint ventures and, in some cases, industrial espionage, and a protected domestic market that gives Chinese technology companies a large home base before competing globally.
The technology competition between the United States and China has now become the organizing framework for global geopolitics. The US-China rivalry is not primarily about territory (unlike most 20th century great power conflicts) but about who controls the technological infrastructure on which the 21st century economy and military order will be built. Semiconductors — as the physical substrate of that infrastructure — are the central battleground.
The weaponization of semiconductor supply chains — US export controls on chips and equipment to China — represents a fundamentally new form of geopolitical coercion. For the first time in history, a nation is using control over the production of a basic economic input (rather than oil, or gold, or a territory) as a primary instrument of strategic competition. This is possible because semiconductor manufacturing is uniquely concentrated: the choke points — ASML's EUV machines, TSMC's advanced fabs, US EDA tools — are controlled by a handful of companies in a handful of countries.
The 21st century's central geopolitical competition is a competition for technological supremacy. This represents a fundamental shift from the Cold War's nuclear-based bipolarity or the post-Cold War's focus on liberal democratic order. The new ordering principle of global power is technological capability — the ability to develop, produce, and deploy the technologies that determine economic productivity, military effectiveness, and cultural influence.
The United States has dominated the global technology landscape since the Second World War — not because of any particular cultural superiority but because of a specific institutional ecology: a well-funded research university system (MIT, Stanford, Caltech, Harvard, Cornell), a military-industrial research complex (DARPA, ARPA-E, national laboratories), venture capital markets willing to fund high-risk technological bets, open immigration attracting global talent, and a market large enough to provide the initial scale for technology companies to achieve global reach.
China's rise as a technology power — while still incomplete and contested — represents the first serious challenge to US technological supremacy since the Soviet space race. China's approach has been different from the US model: centralized state investment and coordination, massive scale of engineering education output (producing more STEM graduates per year than the US), aggressive technology acquisition through joint ventures and, in some cases, industrial espionage, and a protected domestic market that gives Chinese technology companies a large home base before competing globally.
The technology competition between the United States and China has now become the organizing framework for global geopolitics. The US-China rivalry is not primarily about territory (unlike most 20th century great power conflicts) but about who controls the technological infrastructure on which the 21st century economy and military order will be built. Semiconductors — as the physical substrate of that infrastructure — are the central battleground.
The weaponization of semiconductor supply chains — US export controls on chips and equipment to China — represents a fundamentally new form of geopolitical coercion. For the first time in history, a nation is using control over the production of a basic economic input (rather than oil, or gold, or a territory) as a primary instrument of strategic competition. This is possible because semiconductor manufacturing is uniquely concentrated: the choke points — ASML's EUV machines, TSMC's advanced fabs, US EDA tools — are controlled by a handful of companies in a handful of countries.
India's Strategic Position — The Non-Aligned Technology Power?
India's traditional foreign policy stance has been strategic autonomy — the refusal to join either of the major Cold War blocs, articulated through the Non-Aligned Movement. This instinct for strategic autonomy is deeply embedded in India's foreign policy culture and remains relevant today. But technology geopolitics is testing this instinct in new and challenging ways.
The US-China technology conflict is creating pressure on all countries to choose sides. The US export control regime on advanced semiconductors requires all suppliers — not just American companies but also Dutch, Japanese, and Korean companies operating under US technology — to comply with US restrictions on China. Countries that do not align with the US technology regime face restrictions on their own access to US technology. This creates real pressure on India: to fully participate in US-led technology cooperation (including the iCET framework and potential integration into the US CHIPS Act ecosystem), India may need to restrict its use of Chinese technology in critical infrastructure — something India has already been doing in the telecom sector (banning Huawei and ZTE from 5G networks) but has not done comprehensively.
India's strategic position is complex. On one hand, India has genuine strategic alignment with the United States — shared democratic values, shared concerns about Chinese power, growing defense cooperation, and the Quad framework (US, Japan, Australia, India). This alignment argues for deep technology cooperation with the US, Japan, and other democracies.
On the other hand, China is India's largest trading partner (or among the top two or three, depending on how trade is measured) and India's second-largest source of FDI in electronics (through contract manufacturing). India also has significant economic interests in maintaining working relationships with China in sectors where Chinese companies supply components and equipment that India needs. Completely aligning with the US technology regime against China would impose real economic costs on India.
The optimal Indian strategy is probably not pure alignment or pure neutrality but a sophisticated, issue-by-issue approach: deep cooperation with the US in defense technology, AI safety, and critical infrastructure security; selective engagement with Chinese technology companies in non-critical commercial domains; and aggressive pursuit of its own indigenous technology development to reduce dependencies on both sides.
Chapter 34: The Quad as Technology Alliance — Leveraging India's Democratic Partnerships
The Quad — the strategic grouping of the United States, Japan, Australia, and India — has evolved from a maritime security dialogue to a comprehensive strategic partnership with a significant technology dimension. The Quad's Critical and Emerging Technology Working Group has identified semiconductors, AI, quantum computing, and 5G as priority areas for cooperation. Maximizing the benefits of this partnership is central to India's technology strategy.
The specific Quad synergies for India's semiconductor and AI ambitions are significant:
US Technology Access: The US can facilitate India's access to leading US semiconductor companies' technology, to US government-funded research outputs, and potentially to the US CHIPS Act allied partner framework. US export controls have also been calibrated to treat India as a "trusted country" for most dual-use technology — a status that should be preserved and deepened.
Japan Technology and Investment: Japan's semiconductor equipment and materials companies — ASML's Dutch EUV machines aside, Japanese companies like Tokyo Electron, Shin-Etsu, Sumco, and JSR are critical to semiconductor manufacturing. Deepening India-Japan technology cooperation could give Indian semiconductor fabs preferential access to Japanese equipment and materials, and potentially Japanese technical assistance in process development.
Australia's Rare Earth Partnership: Australia is one of the world's largest producers of rare earth elements — materials like neodymium, praseodymium, and dysprosium that are critical for electronics, electric vehicles, and defense systems. China currently dominates rare earth processing (even if not mining), creating a supply chain vulnerability. An Australia-India rare earth partnership — in which Australian rare earth mining is complemented by Indian processing and manufacturing capability — would strengthen both countries' supply chains and create a non-Chinese alternative in a critical materials market.
Quad Semiconductor Coordination: The Quad countries could coordinate on semiconductor investment — for example, by agreeing that specific segments of the semiconductor value chain will be developed in specific countries to avoid redundancy and create complementary capabilities. Japan might focus on materials and equipment, Australia on raw materials, the US on leading-edge design and advanced fabrication, and India on mature-node fabrication, OSAT, and AI chip design.
Chapter 35: The China Factor — Managing the Neighbor While Building Resilience
China's relationship with India is perhaps the most complex of any bilateral relationship in the world — a neighbor of 3,400 km shared border, a fellow ancient civilization, a recent adversary in three border conflicts (1962, 1967, 1987), and simultaneously a major economic partner. The Galwan Valley clash of June 2020 — in which 20 Indian soldiers and an unknown number of Chinese troops died in hand-to-hand combat — has deeply damaged the political relationship between the two countries, and the border standoff in Ladakh that triggered it remains partially unresolved.
In the technology domain, India has made several significant decisions to reduce Chinese technology presence in critical infrastructure:
• Banning 59 Chinese mobile apps (including TikTok, WeChat, and numerous others) in June 2020, primarily citing security concerns
• Restricting Huawei and ZTE from Indian 5G networks in 2023
• Tightening FDI regulations to require government approval for investments from China and Hong Kong bordering countries, effectively slowing or halting most Chinese FDI since 2020
• Restricting the import of Chinese drones (Dajiang Innovation, the dominant global drone manufacturer, has been effectively restricted from Indian government and military procurement)
These measures reflect a genuine security assessment — the risk that Chinese-made technology in critical infrastructure could be used for surveillance or sabotage — but they also have economic costs. Chinese electronics companies had been making significant investments in Indian manufacturing — Xiaomi, OPPO, Vivo, Realme — and some of this investment has been disrupted.
India's approach to China in technology must balance several competing considerations:
• Chinese companies supply important components for India's electronics assembly industry (displays, batteries, PCBs) that are difficult to source elsewhere in the short term
• Chinese AI companies (Baidu, Alibaba Cloud, Huawei's AI division) have capabilities that Indian companies might benefit from in non-critical applications
• Restricting all Chinese technology comprehensively would impose significant economic costs
• Failing to restrict Chinese technology in critical infrastructure creates genuine security risks
The optimal approach is a risk-stratified framework: comprehensive exclusion of Chinese technology from defense, critical infrastructure, and sensitive communications; selective evaluation of Chinese technology in commercial non-critical applications; and active development of domestic and allied-nation alternatives to reduce dependency over time.
Chapter 36: The Global Order — Technology Multipolarity vs. US Hegemony
The current global technology order is built on US dominance: American companies design most of the world's software, control most of the EDA tools, lead in advanced chip design, and maintain significant influence over semiconductor manufacturing standards. US export controls have demonstrated that this dominance can be weaponized — the US can effectively cut off any country from the most advanced technology if it chooses to do so.
This dominance is not permanent. China's technology rise — while still incomplete — is genuine. Europe's Chips Act reflects European determination to reclaim technological sovereignty. Japan and South Korea are deepening their technology independence. India's aspirations add another voice to the chorus of nations seeking to reduce dependence on any single technology power.
The long-term trajectory of the global technology order is toward multipolarity — not the US-China binary that today's rhetoric suggests, but a more genuinely multipolar order with multiple technology powers each with significant capabilities in different domains. This multipolar technology order would feature:
• US dominance in advanced chip design, AI algorithms, and quantum computing research
• Taiwan (or TSMC, wherever it operates) as the leading advanced semiconductor manufacturer
• South Korea as the dominant memory chip manufacturer
• Japan as the dominant materials and equipment supplier
• China as a significant producer of mature-node chips and a leading AI power in Chinese-language applications
• The European Union as a significant producer of automotive and industrial chips (through STMicroelectronics, Infineon, NXP)
• And India — if the right choices are made — as a significant producer of mature-node logic chips, packaging services, and AI infrastructure, and as the leading provider of AI solutions for the Global South
This multipolar technology order would be more resilient than the current highly concentrated one — more supply chains, more geographic diversity, more competition — but it would also be more complex to navigate. India's strategic challenge is to position itself as a significant node in this multipolar order, not merely as a consumer at the periphery.
India's traditional foreign policy stance has been strategic autonomy — the refusal to join either of the major Cold War blocs, articulated through the Non-Aligned Movement. This instinct for strategic autonomy is deeply embedded in India's foreign policy culture and remains relevant today. But technology geopolitics is testing this instinct in new and challenging ways.
The US-China technology conflict is creating pressure on all countries to choose sides. The US export control regime on advanced semiconductors requires all suppliers — not just American companies but also Dutch, Japanese, and Korean companies operating under US technology — to comply with US restrictions on China. Countries that do not align with the US technology regime face restrictions on their own access to US technology. This creates real pressure on India: to fully participate in US-led technology cooperation (including the iCET framework and potential integration into the US CHIPS Act ecosystem), India may need to restrict its use of Chinese technology in critical infrastructure — something India has already been doing in the telecom sector (banning Huawei and ZTE from 5G networks) but has not done comprehensively.
India's strategic position is complex. On one hand, India has genuine strategic alignment with the United States — shared democratic values, shared concerns about Chinese power, growing defense cooperation, and the Quad framework (US, Japan, Australia, India). This alignment argues for deep technology cooperation with the US, Japan, and other democracies.
On the other hand, China is India's largest trading partner (or among the top two or three, depending on how trade is measured) and India's second-largest source of FDI in electronics (through contract manufacturing). India also has significant economic interests in maintaining working relationships with China in sectors where Chinese companies supply components and equipment that India needs. Completely aligning with the US technology regime against China would impose real economic costs on India.
The optimal Indian strategy is probably not pure alignment or pure neutrality but a sophisticated, issue-by-issue approach: deep cooperation with the US in defense technology, AI safety, and critical infrastructure security; selective engagement with Chinese technology companies in non-critical commercial domains; and aggressive pursuit of its own indigenous technology development to reduce dependencies on both sides.
Chapter 34: The Quad as Technology Alliance — Leveraging India's Democratic Partnerships
The Quad — the strategic grouping of the United States, Japan, Australia, and India — has evolved from a maritime security dialogue to a comprehensive strategic partnership with a significant technology dimension. The Quad's Critical and Emerging Technology Working Group has identified semiconductors, AI, quantum computing, and 5G as priority areas for cooperation. Maximizing the benefits of this partnership is central to India's technology strategy.
The specific Quad synergies for India's semiconductor and AI ambitions are significant:
US Technology Access: The US can facilitate India's access to leading US semiconductor companies' technology, to US government-funded research outputs, and potentially to the US CHIPS Act allied partner framework. US export controls have also been calibrated to treat India as a "trusted country" for most dual-use technology — a status that should be preserved and deepened.
Japan Technology and Investment: Japan's semiconductor equipment and materials companies — ASML's Dutch EUV machines aside, Japanese companies like Tokyo Electron, Shin-Etsu, Sumco, and JSR are critical to semiconductor manufacturing. Deepening India-Japan technology cooperation could give Indian semiconductor fabs preferential access to Japanese equipment and materials, and potentially Japanese technical assistance in process development.
Australia's Rare Earth Partnership: Australia is one of the world's largest producers of rare earth elements — materials like neodymium, praseodymium, and dysprosium that are critical for electronics, electric vehicles, and defense systems. China currently dominates rare earth processing (even if not mining), creating a supply chain vulnerability. An Australia-India rare earth partnership — in which Australian rare earth mining is complemented by Indian processing and manufacturing capability — would strengthen both countries' supply chains and create a non-Chinese alternative in a critical materials market.
Quad Semiconductor Coordination: The Quad countries could coordinate on semiconductor investment — for example, by agreeing that specific segments of the semiconductor value chain will be developed in specific countries to avoid redundancy and create complementary capabilities. Japan might focus on materials and equipment, Australia on raw materials, the US on leading-edge design and advanced fabrication, and India on mature-node fabrication, OSAT, and AI chip design.
Chapter 35: The China Factor — Managing the Neighbor While Building Resilience
China's relationship with India is perhaps the most complex of any bilateral relationship in the world — a neighbor of 3,400 km shared border, a fellow ancient civilization, a recent adversary in three border conflicts (1962, 1967, 1987), and simultaneously a major economic partner. The Galwan Valley clash of June 2020 — in which 20 Indian soldiers and an unknown number of Chinese troops died in hand-to-hand combat — has deeply damaged the political relationship between the two countries, and the border standoff in Ladakh that triggered it remains partially unresolved.
In the technology domain, India has made several significant decisions to reduce Chinese technology presence in critical infrastructure:
• Banning 59 Chinese mobile apps (including TikTok, WeChat, and numerous others) in June 2020, primarily citing security concerns
• Restricting Huawei and ZTE from Indian 5G networks in 2023
• Tightening FDI regulations to require government approval for investments from China and Hong Kong bordering countries, effectively slowing or halting most Chinese FDI since 2020
• Restricting the import of Chinese drones (Dajiang Innovation, the dominant global drone manufacturer, has been effectively restricted from Indian government and military procurement)
These measures reflect a genuine security assessment — the risk that Chinese-made technology in critical infrastructure could be used for surveillance or sabotage — but they also have economic costs. Chinese electronics companies had been making significant investments in Indian manufacturing — Xiaomi, OPPO, Vivo, Realme — and some of this investment has been disrupted.
India's approach to China in technology must balance several competing considerations:
• Chinese companies supply important components for India's electronics assembly industry (displays, batteries, PCBs) that are difficult to source elsewhere in the short term
• Chinese AI companies (Baidu, Alibaba Cloud, Huawei's AI division) have capabilities that Indian companies might benefit from in non-critical applications
• Restricting all Chinese technology comprehensively would impose significant economic costs
• Failing to restrict Chinese technology in critical infrastructure creates genuine security risks
The optimal approach is a risk-stratified framework: comprehensive exclusion of Chinese technology from defense, critical infrastructure, and sensitive communications; selective evaluation of Chinese technology in commercial non-critical applications; and active development of domestic and allied-nation alternatives to reduce dependency over time.
Chapter 36: The Global Order — Technology Multipolarity vs. US Hegemony
The current global technology order is built on US dominance: American companies design most of the world's software, control most of the EDA tools, lead in advanced chip design, and maintain significant influence over semiconductor manufacturing standards. US export controls have demonstrated that this dominance can be weaponized — the US can effectively cut off any country from the most advanced technology if it chooses to do so.
This dominance is not permanent. China's technology rise — while still incomplete — is genuine. Europe's Chips Act reflects European determination to reclaim technological sovereignty. Japan and South Korea are deepening their technology independence. India's aspirations add another voice to the chorus of nations seeking to reduce dependence on any single technology power.
The long-term trajectory of the global technology order is toward multipolarity — not the US-China binary that today's rhetoric suggests, but a more genuinely multipolar order with multiple technology powers each with significant capabilities in different domains. This multipolar technology order would feature:
• US dominance in advanced chip design, AI algorithms, and quantum computing research
• Taiwan (or TSMC, wherever it operates) as the leading advanced semiconductor manufacturer
• South Korea as the dominant memory chip manufacturer
• Japan as the dominant materials and equipment supplier
• China as a significant producer of mature-node chips and a leading AI power in Chinese-language applications
• The European Union as a significant producer of automotive and industrial chips (through STMicroelectronics, Infineon, NXP)
• And India — if the right choices are made — as a significant producer of mature-node logic chips, packaging services, and AI infrastructure, and as the leading provider of AI solutions for the Global South
This multipolar technology order would be more resilient than the current highly concentrated one — more supply chains, more geographic diversity, more competition — but it would also be more complex to navigate. India's strategic challenge is to position itself as a significant node in this multipolar order, not merely as a consumer at the periphery.
India and the Global South — Technology Leadership Opportunity
One dimension of India's technology opportunity that deserves special emphasis is the potential for India to become the technology leader for the Global South — the developing nations of Africa, Southeast Asia, Latin America, the Middle East, and smaller Asian nations that are seeking technology development pathways suited to their specific circumstances.
The technology solutions developed by Silicon Valley are designed for Western contexts — English language, high-bandwidth connectivity, high per-capita income, strong regulatory institutions. They are often poorly suited to the realities of developing country contexts: multiple local languages, patchy connectivity, limited digital literacy, different economic structures, and different regulatory frameworks.
India's experience developing technology solutions for its own massive, diverse, multilingual, and partially-connected population gives it a distinctive competence in building "appropriate technology" for developing country contexts. UPI — India's real-time payments platform — is being adopted and adapted in multiple countries (Singapore, France, UAE, the UK are accepting UPI payments; Bhutan, Nepal, Mauritius, Sri Lanka have adopted UPI-based systems). Aadhaar's digital identity model is being studied and adapted by multiple developing countries. India's CoWIN vaccination management platform was offered to other countries during COVID-19.
In the AI domain, India's focus on multilingual AI, low-bandwidth AI deployment, and AI for agriculture and healthcare creates capabilities that are directly applicable to other developing country contexts. An Indian AI company that has built a highly capable Hindi language model and an agricultural advisory AI system for Indian farmers has a product that could be adapted for Nigerian Hausa-language markets, Bangladeshi Bengali markets, or Indonesian Bahasa markets — with less adaptation required than would be needed starting from an English-language US model.
This "India Stack for the world" narrative — leveraging India's unique experience of building digital public infrastructure at massive scale and with limited resources — is beginning to gain traction in Indian foreign policy circles. The G20 presidency that India held in 2023 allowed India to champion digital public infrastructure as a global development priority, and the Global Digital Public Infrastructure Repository (GDPIR) established under India's G20 presidency is a concrete step toward sharing India's digital governance experience with the world.
In semiconductors, India's potential role as a manufacturer of mature-node chips for the Global South — affordable chips for affordable devices, manufactured closer to home in the Global South rather than imported from East Asia — is a distinctive niche that India could serve. Many Global South countries have similar supply chain vulnerabilities to India and would welcome a closer-to-home, democratically governed semiconductor manufacturing alternative.
One dimension of India's technology opportunity that deserves special emphasis is the potential for India to become the technology leader for the Global South — the developing nations of Africa, Southeast Asia, Latin America, the Middle East, and smaller Asian nations that are seeking technology development pathways suited to their specific circumstances.
The technology solutions developed by Silicon Valley are designed for Western contexts — English language, high-bandwidth connectivity, high per-capita income, strong regulatory institutions. They are often poorly suited to the realities of developing country contexts: multiple local languages, patchy connectivity, limited digital literacy, different economic structures, and different regulatory frameworks.
India's experience developing technology solutions for its own massive, diverse, multilingual, and partially-connected population gives it a distinctive competence in building "appropriate technology" for developing country contexts. UPI — India's real-time payments platform — is being adopted and adapted in multiple countries (Singapore, France, UAE, the UK are accepting UPI payments; Bhutan, Nepal, Mauritius, Sri Lanka have adopted UPI-based systems). Aadhaar's digital identity model is being studied and adapted by multiple developing countries. India's CoWIN vaccination management platform was offered to other countries during COVID-19.
In the AI domain, India's focus on multilingual AI, low-bandwidth AI deployment, and AI for agriculture and healthcare creates capabilities that are directly applicable to other developing country contexts. An Indian AI company that has built a highly capable Hindi language model and an agricultural advisory AI system for Indian farmers has a product that could be adapted for Nigerian Hausa-language markets, Bangladeshi Bengali markets, or Indonesian Bahasa markets — with less adaptation required than would be needed starting from an English-language US model.
This "India Stack for the world" narrative — leveraging India's unique experience of building digital public infrastructure at massive scale and with limited resources — is beginning to gain traction in Indian foreign policy circles. The G20 presidency that India held in 2023 allowed India to champion digital public infrastructure as a global development priority, and the Global Digital Public Infrastructure Repository (GDPIR) established under India's G20 presidency is a concrete step toward sharing India's digital governance experience with the world.
In semiconductors, India's potential role as a manufacturer of mature-node chips for the Global South — affordable chips for affordable devices, manufactured closer to home in the Global South rather than imported from East Asia — is a distinctive niche that India could serve. Many Global South countries have similar supply chain vulnerabilities to India and would welcome a closer-to-home, democratically governed semiconductor manufacturing alternative.
The Digital Currency and Financial Infrastructure Dimension
An underappreciated dimension of the technology sovereignty question is the financial infrastructure layer. The global financial system — SWIFT payments, US dollar denomination of commodities, US-controlled financial data — has been revealed as a geopolitical weapon through the use of financial sanctions. The US disconnection of Russian banks from SWIFT after the 2022 invasion of Ukraine demonstrated the extraordinary coercive power that control over financial infrastructure provides.
India's Reserve Bank has been developing the Digital Rupee (e-RUPI, CBDC — Central Bank Digital Currency) partly with an eye to reducing India's dependence on dollar-denominated financial infrastructure. The Indian Rupee's use in bilateral trade — India has negotiated with several countries to settle trade in rupees rather than dollars — is another dimension of this strategy.
But digital financial infrastructure requires semiconductor support. The servers that process CBDC transactions, the network infrastructure that carries digital payments, the HSM (Hardware Security Modules) that protect cryptographic keys in financial systems — all require semiconductors. Building domestic semiconductor capability is therefore directly related to India's ability to build sovereign digital financial infrastructure.
The AI dimension of financial services adds another layer of dependency: AI systems used for fraud detection, credit risk assessment, algorithmic trading, and regulatory compliance in India's financial system are primarily built on foreign AI platforms and run on foreign cloud infrastructure. Developing domestic AI capability for financial services — on domestically controlled infrastructure — is a financial stability and sovereignty issue as much as a technology issue.
PART TEN: THE PATH FORWARD — A COMPREHENSIVE VISION
An underappreciated dimension of the technology sovereignty question is the financial infrastructure layer. The global financial system — SWIFT payments, US dollar denomination of commodities, US-controlled financial data — has been revealed as a geopolitical weapon through the use of financial sanctions. The US disconnection of Russian banks from SWIFT after the 2022 invasion of Ukraine demonstrated the extraordinary coercive power that control over financial infrastructure provides.
India's Reserve Bank has been developing the Digital Rupee (e-RUPI, CBDC — Central Bank Digital Currency) partly with an eye to reducing India's dependence on dollar-denominated financial infrastructure. The Indian Rupee's use in bilateral trade — India has negotiated with several countries to settle trade in rupees rather than dollars — is another dimension of this strategy.
But digital financial infrastructure requires semiconductor support. The servers that process CBDC transactions, the network infrastructure that carries digital payments, the HSM (Hardware Security Modules) that protect cryptographic keys in financial systems — all require semiconductors. Building domestic semiconductor capability is therefore directly related to India's ability to build sovereign digital financial infrastructure.
The AI dimension of financial services adds another layer of dependency: AI systems used for fraud detection, credit risk assessment, algorithmic trading, and regulatory compliance in India's financial system are primarily built on foreign AI platforms and run on foreign cloud infrastructure. Developing domestic AI capability for financial services — on domestically controlled infrastructure — is a financial stability and sovereignty issue as much as a technology issue.
PART TEN: THE PATH FORWARD — A COMPREHENSIVE VISION
The Ten-Year Roadmap — Specific Steps for 2025-2035
India's path from technology dependent to technology capable requires a carefully sequenced, adequately resourced, and institutionally well-supported effort over the next decade. The following roadmap outlines the key milestones and actions required:
2024-2026 — Foundation Phase:
• Complete the regulatory and infrastructure setup for the Micron OSAT facility in Gujarat
• Finalize land, power, and water commitments for the Tata-PSMC 28nm wafer fab project; begin construction
• Launch the IndiaAI compute infrastructure — first tranche of 10,000 GPUs, with a commitment to scale to 50,000 by 2027
• Establish the Semiconductor Design Centers of Excellence at five IITs (Delhi, Bombay, Madras, Kharagpur, Hyderabad) with full EDA tool suites and chip tape-out facilities
• Complete the National Semiconductor Research Fund setup and make first grants to research projects
• Launch an Indian Language AI initiative — dedicated program to develop high-quality datasets and competitive models for the top 10 Indian languages
• Negotiate iCET-based technology cooperation agreements with the US in at least two specific areas: (a) access to CHIPS Act-compatible technology transfer from US companies and (b) joint research in AI safety
2026-2028 — Build Phase:
• Commission the Micron OSAT facility — first semiconductor manufacturing product shipments from India
• Achieve first silicon from the Tata-PSMC fab (first test wafers, before commercial production)
• Scale IndiaAI compute to 50,000 GPUs, enabling meaningful AI model training runs
• Publish first competitive Indian language AI models for Hindi, Tamil, and Telugu
• Attract a second major semiconductor fab investment — ideally at a more advanced node than 28nm, targeting perhaps 12-16nm, through enhanced incentives and technology partnerships
• Launch a dedicated DRAM technology initiative with international partner (Micron preferred)
• Establish India's first semiconductor equipment manufacturing program — targeting selected components where Indian manufacturing capability is viable (process chemical supply, basic metrology equipment, handling systems)
2028-2032 — Scale Phase:
• Bring Tata-PSMC fab to full commercial production at 28nm — producing chips for domestic automotive, industrial, and consumer electronics markets
• Begin construction of a second, more advanced logic fab (12nm or below)
• Commission the first phase of DRAM manufacturing in India (LPDDR4 or equivalent)
• Achieve self-sufficiency in India-designed AI chips for specific applications (AI inference for healthcare, agricultural AI, government services)
• Develop a thriving Indian fabless chip design ecosystem — at least 50 Indian-owned chip design companies of meaningful scale
• Achieve 15-20% domestic content in electronics assembled in India (vs. <5% today)
• Establish India as a significant exporter of semiconductors — initially at $5-10 billion per year in OSAT and mature-node chip exports
2032-2035 — Consolidation Phase:
• Multiple Indian semiconductor fabs operating at mature and mid-tier nodes
• DRAM manufacturing at scale for domestic demand (LPDDR5 and DDR5)
• India as a net exporter of some semiconductor categories (packaging, mature-node logic, power semiconductors)
• Frontier Indian AI companies — competitive with global mid-tier companies in Indian language AI and specialized domains
• Semiconductor design ecosystem with annual revenue exceeding $10 billion from Indian-owned companies
Chapter 40: The Funding Architecture — How to Finance India's Technology Ambition
The scale of investment required — approximately $50-100 billion over ten years — cannot come from government alone, nor from pure private markets given the long time horizons and high risk of semiconductor manufacturing. It requires a carefully structured blend of public and private financing.
Government Direct Subsidies and Grants: The existing PLI scheme for semiconductors should be significantly enhanced — the total pool should be increased from ₹76,000 crore to ₹2-3 lakh crore ($25-35 billion) to provide adequate incentives for the scale of investment needed. The subsidy rate should be differentiated — higher (60-70%) for the most advanced investments (sub-10nm fabs, DRAM manufacturing) and lower (20-30%) for more mature technology investments. Subsidies should be structured as grants paid on achievement of specific production milestones, not upfront payments, to ensure actual production rather than announcements.
Sovereign Technology Fund: India should establish a dedicated Sovereign Technology Fund — modeled on Singapore's Temasek or South Korea's Korea Investment Corporation — with an initial corpus of $10-15 billion, drawing from India's foreign exchange reserves and the National Investment and Infrastructure Fund (NIIF). This fund would make equity investments in semiconductor and AI companies — both Indian companies and joint ventures with foreign partners — with the explicit mandate to build strategic technology capability rather than maximize financial return. The fund should be managed by experienced investment professionals with deep technology sector expertise, not by civil servants.
Development Finance Institution Lending: The EXIM Bank of India, SIDBI, and the National Bank for Financing Infrastructure and Development (NaBFID) should be given specific mandates and increased capitalization to provide long-term, concessional lending to semiconductor manufacturing projects. Semiconductor fabs need 15-20 year debt financing at competitive interest rates — the typical commercial bank lending tenure of 7-10 years is insufficient for assets with 20-30 year productive lives.
International Development Finance: The US International Development Finance Corporation (DFC), the Japan International Cooperation Agency (JICA), and the Asian Development Bank (ADB) should be engaged as co-financing partners for India's semiconductor projects. Positioning semiconductor manufacturing in India as a global supply chain resilience investment — which benefits all countries dependent on current concentrated supply chains — strengthens the case for international development finance support.
Private Equity and Venture Capital: The semiconductor design ecosystem — fabless chip companies, AI startups, semiconductor IP companies — should be financed primarily by private equity and venture capital, with government support through tax incentives (100% R&D deduction for semiconductor and AI research) and regulatory frameworks that make India an attractive investment destination. The emergence of several Indian semiconductor-focused VC funds — IndiaQuotient, Stellaris, Blume, and global funds like Sequoia Capital India — represents a growing private capital ecosystem for tech startups.
Diaspora Engagement: India's diaspora — particularly the community of Indian-origin professionals working in semiconductor companies, AI companies, and technology investment firms in the United States — represents an extraordinary resource. Programs to engage diaspora expertise — through advisory roles, research partnerships, return migration incentives, and investment facilitation — should be a component of India's technology strategy. The Indian-origin leaders at Intel, Google, Microsoft, IBM, and countless other technology companies have knowledge, relationships, and potentially investment interest that could be mobilized for India's semiconductor ambitions.
India's path from technology dependent to technology capable requires a carefully sequenced, adequately resourced, and institutionally well-supported effort over the next decade. The following roadmap outlines the key milestones and actions required:
2024-2026 — Foundation Phase:
• Complete the regulatory and infrastructure setup for the Micron OSAT facility in Gujarat
• Finalize land, power, and water commitments for the Tata-PSMC 28nm wafer fab project; begin construction
• Launch the IndiaAI compute infrastructure — first tranche of 10,000 GPUs, with a commitment to scale to 50,000 by 2027
• Establish the Semiconductor Design Centers of Excellence at five IITs (Delhi, Bombay, Madras, Kharagpur, Hyderabad) with full EDA tool suites and chip tape-out facilities
• Complete the National Semiconductor Research Fund setup and make first grants to research projects
• Launch an Indian Language AI initiative — dedicated program to develop high-quality datasets and competitive models for the top 10 Indian languages
• Negotiate iCET-based technology cooperation agreements with the US in at least two specific areas: (a) access to CHIPS Act-compatible technology transfer from US companies and (b) joint research in AI safety
2026-2028 — Build Phase:
• Commission the Micron OSAT facility — first semiconductor manufacturing product shipments from India
• Achieve first silicon from the Tata-PSMC fab (first test wafers, before commercial production)
• Scale IndiaAI compute to 50,000 GPUs, enabling meaningful AI model training runs
• Publish first competitive Indian language AI models for Hindi, Tamil, and Telugu
• Attract a second major semiconductor fab investment — ideally at a more advanced node than 28nm, targeting perhaps 12-16nm, through enhanced incentives and technology partnerships
• Launch a dedicated DRAM technology initiative with international partner (Micron preferred)
• Establish India's first semiconductor equipment manufacturing program — targeting selected components where Indian manufacturing capability is viable (process chemical supply, basic metrology equipment, handling systems)
2028-2032 — Scale Phase:
• Bring Tata-PSMC fab to full commercial production at 28nm — producing chips for domestic automotive, industrial, and consumer electronics markets
• Begin construction of a second, more advanced logic fab (12nm or below)
• Commission the first phase of DRAM manufacturing in India (LPDDR4 or equivalent)
• Achieve self-sufficiency in India-designed AI chips for specific applications (AI inference for healthcare, agricultural AI, government services)
• Develop a thriving Indian fabless chip design ecosystem — at least 50 Indian-owned chip design companies of meaningful scale
• Achieve 15-20% domestic content in electronics assembled in India (vs. <5% today)
• Establish India as a significant exporter of semiconductors — initially at $5-10 billion per year in OSAT and mature-node chip exports
2032-2035 — Consolidation Phase:
• Multiple Indian semiconductor fabs operating at mature and mid-tier nodes
• DRAM manufacturing at scale for domestic demand (LPDDR5 and DDR5)
• India as a net exporter of some semiconductor categories (packaging, mature-node logic, power semiconductors)
• Frontier Indian AI companies — competitive with global mid-tier companies in Indian language AI and specialized domains
• Semiconductor design ecosystem with annual revenue exceeding $10 billion from Indian-owned companies
Chapter 40: The Funding Architecture — How to Finance India's Technology Ambition
The scale of investment required — approximately $50-100 billion over ten years — cannot come from government alone, nor from pure private markets given the long time horizons and high risk of semiconductor manufacturing. It requires a carefully structured blend of public and private financing.
Government Direct Subsidies and Grants: The existing PLI scheme for semiconductors should be significantly enhanced — the total pool should be increased from ₹76,000 crore to ₹2-3 lakh crore ($25-35 billion) to provide adequate incentives for the scale of investment needed. The subsidy rate should be differentiated — higher (60-70%) for the most advanced investments (sub-10nm fabs, DRAM manufacturing) and lower (20-30%) for more mature technology investments. Subsidies should be structured as grants paid on achievement of specific production milestones, not upfront payments, to ensure actual production rather than announcements.
Sovereign Technology Fund: India should establish a dedicated Sovereign Technology Fund — modeled on Singapore's Temasek or South Korea's Korea Investment Corporation — with an initial corpus of $10-15 billion, drawing from India's foreign exchange reserves and the National Investment and Infrastructure Fund (NIIF). This fund would make equity investments in semiconductor and AI companies — both Indian companies and joint ventures with foreign partners — with the explicit mandate to build strategic technology capability rather than maximize financial return. The fund should be managed by experienced investment professionals with deep technology sector expertise, not by civil servants.
Development Finance Institution Lending: The EXIM Bank of India, SIDBI, and the National Bank for Financing Infrastructure and Development (NaBFID) should be given specific mandates and increased capitalization to provide long-term, concessional lending to semiconductor manufacturing projects. Semiconductor fabs need 15-20 year debt financing at competitive interest rates — the typical commercial bank lending tenure of 7-10 years is insufficient for assets with 20-30 year productive lives.
International Development Finance: The US International Development Finance Corporation (DFC), the Japan International Cooperation Agency (JICA), and the Asian Development Bank (ADB) should be engaged as co-financing partners for India's semiconductor projects. Positioning semiconductor manufacturing in India as a global supply chain resilience investment — which benefits all countries dependent on current concentrated supply chains — strengthens the case for international development finance support.
Private Equity and Venture Capital: The semiconductor design ecosystem — fabless chip companies, AI startups, semiconductor IP companies — should be financed primarily by private equity and venture capital, with government support through tax incentives (100% R&D deduction for semiconductor and AI research) and regulatory frameworks that make India an attractive investment destination. The emergence of several Indian semiconductor-focused VC funds — IndiaQuotient, Stellaris, Blume, and global funds like Sequoia Capital India — represents a growing private capital ecosystem for tech startups.
Diaspora Engagement: India's diaspora — particularly the community of Indian-origin professionals working in semiconductor companies, AI companies, and technology investment firms in the United States — represents an extraordinary resource. Programs to engage diaspora expertise — through advisory roles, research partnerships, return migration incentives, and investment facilitation — should be a component of India's technology strategy. The Indian-origin leaders at Intel, Google, Microsoft, IBM, and countless other technology companies have knowledge, relationships, and potentially investment interest that could be mobilized for India's semiconductor ambitions.
The Cultural Shift Required — From Services to Creation
Perhaps the deepest challenge in India's technology transformation is cultural and attitudinal. India's technology sector has been shaped, for three decades, by the IT services model — a culture of client service, process execution, and cost efficiency. This culture has produced extraordinary results in its domain. But it is not the culture of technology creation.
Creating new semiconductors, training new AI models, developing new manufacturing processes — these activities require different cultural attributes: comfort with failure and uncertainty, long time horizons for payoff, willingness to challenge established approaches, deep technical curiosity rather than client orientation. These are not attributes that are absent from Indian culture — Indian mathematicians, physicists, and scientists have demonstrated them abundantly throughout history. But they are not well-represented in the dominant culture of India's IT services sector.
The change required is not just in policy or investment but in how Indian society values different types of technical achievement. The engineer who builds a better semiconductor manufacturing process should be celebrated as much as the engineer who builds a unicorn fintech app. The scientist who develops a new AI architecture should be as celebrated as the product manager who grows a startup to IPO. Shifting social values takes time and requires deliberate effort — through education, media, role model celebration, and institutional reward structures.
India's startup ecosystem has begun to create role models for technology entrepreneurship — but the celebrated Indian founders are predominantly software and services entrepreneurs. The emergence of hardware-focused Indian startups — in semiconductor design, deep tech, space technology, defense technology — is a positive trend that needs to be nurtured and amplified.
Perhaps the deepest challenge in India's technology transformation is cultural and attitudinal. India's technology sector has been shaped, for three decades, by the IT services model — a culture of client service, process execution, and cost efficiency. This culture has produced extraordinary results in its domain. But it is not the culture of technology creation.
Creating new semiconductors, training new AI models, developing new manufacturing processes — these activities require different cultural attributes: comfort with failure and uncertainty, long time horizons for payoff, willingness to challenge established approaches, deep technical curiosity rather than client orientation. These are not attributes that are absent from Indian culture — Indian mathematicians, physicists, and scientists have demonstrated them abundantly throughout history. But they are not well-represented in the dominant culture of India's IT services sector.
The change required is not just in policy or investment but in how Indian society values different types of technical achievement. The engineer who builds a better semiconductor manufacturing process should be celebrated as much as the engineer who builds a unicorn fintech app. The scientist who develops a new AI architecture should be as celebrated as the product manager who grows a startup to IPO. Shifting social values takes time and requires deliberate effort — through education, media, role model celebration, and institutional reward structures.
India's startup ecosystem has begun to create role models for technology entrepreneurship — but the celebrated Indian founders are predominantly software and services entrepreneurs. The emergence of hardware-focused Indian startups — in semiconductor design, deep tech, space technology, defense technology — is a positive trend that needs to be nurtured and amplified.
The Environmental and Resource Dimension
Semiconductor manufacturing is extraordinarily resource-intensive — in energy, water, and specialty materials. India's semiconductor ambitions must be planned with explicit attention to the environmental footprint of manufacturing at scale.
Energy: A typical 200mm wafer fab consumes approximately 50-100 MW of continuous power. A modern 300mm leading-edge fab may consume 200-500 MW. India's Tata-PSMC fab, when fully operational, will require a significant, dedicated power supply — and it must be absolutely continuous (even a momentary power interruption can ruin thousands of wafers in process). India's power supply reliability, particularly for industrial consumers outside the largest cities, is not yet at semiconductor-grade reliability. Ensuring that fab sites have dedicated power supply — potentially with on-site generation and backup — is a critical infrastructure requirement.
The energy mix for semiconductor manufacturing is also important from a sustainability and international credibility perspective. Leading chip customers (Apple, Qualcomm, Intel) have aggressive carbon neutrality commitments that they apply to their supply chains. A semiconductor fab in India powered primarily by coal would struggle to attract these customers. Siting semiconductor fabs in India's solar-rich regions (Rajasthan, Gujarat, Tamil Nadu) and ensuring renewable power purchase agreements is important for both environmental and commercial reasons.
Water: Semiconductor manufacturing requires ultra-pure water in large quantities — a 300mm fab may use 2-5 million gallons per day. India's water scarcity challenges are well-known, and siting semiconductor fabs must be done with explicit consideration of local water availability and the impact of fab water usage on local communities. Water recycling and reclamation systems — standard in modern semiconductor fabs — can dramatically reduce water consumption, but net water usage remains significant. Fab site selection must include water resource assessment, and fabs should implement best-practice water recycling from the outset.
Semiconductor manufacturing generates significant chemical waste. Proper waste treatment and disposal is both an environmental requirement and a safety necessity. India must ensure that semiconductor fabs operate to international environmental standards — not the lower standards that might be negotiable for some other industries. The reputation of Indian semiconductor manufacturing in global supply chains depends partly on demonstrated environmental compliance.
PART ELEVEN: SYNTHESIS AND CONCLUSION
Semiconductor manufacturing is extraordinarily resource-intensive — in energy, water, and specialty materials. India's semiconductor ambitions must be planned with explicit attention to the environmental footprint of manufacturing at scale.
Energy: A typical 200mm wafer fab consumes approximately 50-100 MW of continuous power. A modern 300mm leading-edge fab may consume 200-500 MW. India's Tata-PSMC fab, when fully operational, will require a significant, dedicated power supply — and it must be absolutely continuous (even a momentary power interruption can ruin thousands of wafers in process). India's power supply reliability, particularly for industrial consumers outside the largest cities, is not yet at semiconductor-grade reliability. Ensuring that fab sites have dedicated power supply — potentially with on-site generation and backup — is a critical infrastructure requirement.
The energy mix for semiconductor manufacturing is also important from a sustainability and international credibility perspective. Leading chip customers (Apple, Qualcomm, Intel) have aggressive carbon neutrality commitments that they apply to their supply chains. A semiconductor fab in India powered primarily by coal would struggle to attract these customers. Siting semiconductor fabs in India's solar-rich regions (Rajasthan, Gujarat, Tamil Nadu) and ensuring renewable power purchase agreements is important for both environmental and commercial reasons.
Water: Semiconductor manufacturing requires ultra-pure water in large quantities — a 300mm fab may use 2-5 million gallons per day. India's water scarcity challenges are well-known, and siting semiconductor fabs must be done with explicit consideration of local water availability and the impact of fab water usage on local communities. Water recycling and reclamation systems — standard in modern semiconductor fabs — can dramatically reduce water consumption, but net water usage remains significant. Fab site selection must include water resource assessment, and fabs should implement best-practice water recycling from the outset.
Semiconductor manufacturing generates significant chemical waste. Proper waste treatment and disposal is both an environmental requirement and a safety necessity. India must ensure that semiconductor fabs operate to international environmental standards — not the lower standards that might be negotiable for some other industries. The reputation of Indian semiconductor manufacturing in global supply chains depends partly on demonstrated environmental compliance.
PART ELEVEN: SYNTHESIS AND CONCLUSION
The Civilizational Stakes — Why This Matters for India's Future
This essay has traveled a long distance — from Aryabhata's 5th century mathematics to the 3nm process nodes of TSMC's most advanced fabs, from the British destruction of India's textile industry to US export controls on Nvidia GPUs, from the birth of DRAM in Robert Dennard's IBM lab to the High Bandwidth Memory inside Nvidia's H100s. The journey has been wide-ranging because the challenge is wide-ranging — it spans history, technology, economics, geopolitics, and culture.
But the central argument is simple: India stands at a civilizational inflection point. The technology decisions of the next ten to fifteen years will determine whether India becomes a genuine first-tier technology power — a nation that shapes the technological future rather than merely consuming it — or whether it remains a sophisticated consumer and services provider in a global technology order whose terms are set by others.
The stakes of getting this right extend far beyond economics. In the geopolitical context of the mid-21st century, nations without domestic technology capability are nations without sovereignty. They can be coerced through supply chain weaponization. They cannot ensure the security of their own critical infrastructure. They cannot develop the AI systems required for sovereign military decision-making. They cannot build the digital public infrastructure their citizens deserve without dependence on foreign platforms governed by foreign rules.
India's civilizational tradition — its deep intellectual heritage, its mathematical genius, its capacity for systemic thinking — is entirely adequate to the challenge of becoming a technology leader. The question is not capability but will: the political will to make the long-term, expensive, institutionally demanding investments required; the educational will to transform India's engineering culture from service to creation; the geopolitical will to use India's strategic position to leverage technology partnerships that accelerate domestic capability building.
This essay has traveled a long distance — from Aryabhata's 5th century mathematics to the 3nm process nodes of TSMC's most advanced fabs, from the British destruction of India's textile industry to US export controls on Nvidia GPUs, from the birth of DRAM in Robert Dennard's IBM lab to the High Bandwidth Memory inside Nvidia's H100s. The journey has been wide-ranging because the challenge is wide-ranging — it spans history, technology, economics, geopolitics, and culture.
But the central argument is simple: India stands at a civilizational inflection point. The technology decisions of the next ten to fifteen years will determine whether India becomes a genuine first-tier technology power — a nation that shapes the technological future rather than merely consuming it — or whether it remains a sophisticated consumer and services provider in a global technology order whose terms are set by others.
The stakes of getting this right extend far beyond economics. In the geopolitical context of the mid-21st century, nations without domestic technology capability are nations without sovereignty. They can be coerced through supply chain weaponization. They cannot ensure the security of their own critical infrastructure. They cannot develop the AI systems required for sovereign military decision-making. They cannot build the digital public infrastructure their citizens deserve without dependence on foreign platforms governed by foreign rules.
India's civilizational tradition — its deep intellectual heritage, its mathematical genius, its capacity for systemic thinking — is entirely adequate to the challenge of becoming a technology leader. The question is not capability but will: the political will to make the long-term, expensive, institutionally demanding investments required; the educational will to transform India's engineering culture from service to creation; the geopolitical will to use India's strategic position to leverage technology partnerships that accelerate domestic capability building.
The Urgency — Why Now Is the Critical Window
If the challenge of building semiconductor and AI capability has existed for decades, why is it particularly urgent now? What makes the 2024-2035 window particularly critical?
The Supply Chain Restructuring Window: The US-China technology conflict is driving a once-in-a-generation restructuring of global technology supply chains. Companies that have concentrated their manufacturing in Taiwan, South Korea, and China are being pressured — by export controls, by geopolitical risk, by customer demand for supply chain resilience — to diversify to additional locations. This restructuring is happening now, and the decisions being made now will shape the geography of semiconductor manufacturing for the next 30-50 years. If India does not capture a significant share of this supply chain restructuring now — in this window — it will face a new equilibrium of supply chain geography that may not include India, making a subsequent entry even more difficult.
The AI Infrastructure Window: The AI industry is in its early formative phase — the period in which the infrastructure (compute, data, algorithms) is being built and the foundational companies are being established. Early participants in this infrastructure buildout will have advantages — in accumulated learning, in established platforms, in talent ecosystems — that late entrants will find difficult to overcome. India needs to establish itself as a significant participant in AI infrastructure now, while the field is still young enough that early investment can yield foundational advantages.
The Technology Divergence Window: The gap between the global technology frontier and India's current capability is widening — every year that Moore's Law advances (even at its slowing pace), the frontier moves further from where India currently stands. Building a 28nm fab in 2027-2028 is a bigger achievement relative to the frontier than building it would have been in 2015 — because the frontier has advanced further. The same is true for DRAM technology and AI compute. The cost of catching up increases as the gap widens. Acting now — when the gap, while large, is still bridgeable — is less costly than acting later when it may be unbridgeable.
The Geopolitical Alignment Window: The strategic alignment between India and the technology-controlling democracies (US, Japan, South Korea, Australia, the EU) is currently strong. The iCET framework, the Quad technology cooperation initiatives, the US desire for allied semiconductor manufacturing capacity — these create a favorable geopolitical environment for India to access technology, investment, and partnerships. This alignment is not guaranteed to persist indefinitely — geopolitical alignments shift. India should maximize the benefits of the current favorable alignment while it lasts.
The Domestic Demand Window: India's domestic market is growing rapidly — in smartphones, data centers, electric vehicles, defence electronics, and AI applications. Domestic demand is the primary commercial justification for domestic manufacturing. The demand is large enough now — and growing fast enough — to justify the scale of manufacturing investment required. Building manufacturing capacity to serve a rapidly growing domestic market creates a virtuous cycle of scale, learning, cost reduction, and competitiveness that makes Indian semiconductor manufacturers progressively more competitive globally. Starting this cycle now, when domestic demand is surging, is the optimal timing.
If the challenge of building semiconductor and AI capability has existed for decades, why is it particularly urgent now? What makes the 2024-2035 window particularly critical?
The Supply Chain Restructuring Window: The US-China technology conflict is driving a once-in-a-generation restructuring of global technology supply chains. Companies that have concentrated their manufacturing in Taiwan, South Korea, and China are being pressured — by export controls, by geopolitical risk, by customer demand for supply chain resilience — to diversify to additional locations. This restructuring is happening now, and the decisions being made now will shape the geography of semiconductor manufacturing for the next 30-50 years. If India does not capture a significant share of this supply chain restructuring now — in this window — it will face a new equilibrium of supply chain geography that may not include India, making a subsequent entry even more difficult.
The AI Infrastructure Window: The AI industry is in its early formative phase — the period in which the infrastructure (compute, data, algorithms) is being built and the foundational companies are being established. Early participants in this infrastructure buildout will have advantages — in accumulated learning, in established platforms, in talent ecosystems — that late entrants will find difficult to overcome. India needs to establish itself as a significant participant in AI infrastructure now, while the field is still young enough that early investment can yield foundational advantages.
The Technology Divergence Window: The gap between the global technology frontier and India's current capability is widening — every year that Moore's Law advances (even at its slowing pace), the frontier moves further from where India currently stands. Building a 28nm fab in 2027-2028 is a bigger achievement relative to the frontier than building it would have been in 2015 — because the frontier has advanced further. The same is true for DRAM technology and AI compute. The cost of catching up increases as the gap widens. Acting now — when the gap, while large, is still bridgeable — is less costly than acting later when it may be unbridgeable.
The Geopolitical Alignment Window: The strategic alignment between India and the technology-controlling democracies (US, Japan, South Korea, Australia, the EU) is currently strong. The iCET framework, the Quad technology cooperation initiatives, the US desire for allied semiconductor manufacturing capacity — these create a favorable geopolitical environment for India to access technology, investment, and partnerships. This alignment is not guaranteed to persist indefinitely — geopolitical alignments shift. India should maximize the benefits of the current favorable alignment while it lasts.
The Domestic Demand Window: India's domestic market is growing rapidly — in smartphones, data centers, electric vehicles, defence electronics, and AI applications. Domestic demand is the primary commercial justification for domestic manufacturing. The demand is large enough now — and growing fast enough — to justify the scale of manufacturing investment required. Building manufacturing capacity to serve a rapidly growing domestic market creates a virtuous cycle of scale, learning, cost reduction, and competitiveness that makes Indian semiconductor manufacturers progressively more competitive globally. Starting this cycle now, when domestic demand is surging, is the optimal timing.
The Final Reckoning — What India Owes Its Future
Every generation of Indians has faced its defining challenge. Nehru's generation faced the challenge of building a nation from the ruins of colonialism — of feeding a population, establishing democratic institutions, and asserting sovereignty. The generation of 1991 faced the challenge of economic reform — of breaking free from the License Raj and establishing India as a market economy capable of sustained growth. The generation of the 2000s faced the challenge of sustaining growth while addressing inequality and building basic infrastructure.
This generation — the generation of Indian leaders, engineers, entrepreneurs, investors, and citizens active in the 2020s and 2030s — faces the challenge of technological transformation. The challenge is to build the semiconductor, memory, and AI capabilities that will determine India's place in the 21st-century global order. It is not as dramatic as independence, not as politically fraught as economic liberalization, but its consequences are equally momentous.
The cost of getting it right — of making the investments, building the institutions, training the people, and negotiating the partnerships required — is large but finite and manageable. The cost of getting it wrong — of allowing another decade or two to pass while the technological gap with the leading nations widens, while India's digital economy remains built on foreign semiconductor foundations, while other nations' AI systems shape the information that reaches Indian citizens — is incalculable.
India's 1.4 billion citizens deserve sovereignty in the digital age. They deserve healthcare AI that was built with Indian data and understands Indian diseases. They deserve agricultural AI that speaks their language and understands their farming conditions. They deserve smartphones built with Indian chips. They deserve a digital economy whose foundational infrastructure is not beholden to the export control decisions of foreign governments. They deserve to live in a country that is a technology shaper, not merely a technology consumer.
This is achievable. India has the talent — demonstrated in every major technology company in the world. India has the market — the second-largest internet population in the world, growing at extraordinary speed. India has the mathematical tradition — five thousand years of intellectual achievement in exactly the domains that matter for computing and AI. India has growing geopolitical standing — as a major economy, the world's most populous democracy, and a pivotal player in the Indo-Pacific balance of power.
What India needs is the decision — the national commitment, backed by adequate resources, institutional machinery, and sustained political will — to build this capability. The decision to compete not just in the services layer of the technology economy but in its physical and algorithmic foundations. The decision to be a maker, not just a user, of the technologies that will define the century.
Aryabhata described the rotation of the Earth and calculated the length of the year to extraordinary precision — in the 5th century CE, a thousand years before Copernicus. Brahmagupta gave the world zero — the concept without which every computer program, every AI model, every semiconductor design process would be impossible. Srinivasa Ramanujan produced mathematical theorems of incomprehensible elegance with virtually no formal training — theorems that mathematicians are still exploring a century later.
The intellectual heritage is there. The global networks are there. The domestic demand is there. The geopolitical opportunity is there.
The moment is now.
EPILOGUE: THE LONG VIEW
There is a quotation, attributed to various sources but perhaps best associated with the long sweep of Chinese strategic thinking, that "the best time to plant a tree was twenty years ago. The second-best time is now."
India's semiconductor and AI tree should have been planted in the 1980s, when South Korea and Taiwan were planting theirs. It should have been planted in the 1990s, when the digital age was being born. It should have been planted in the 2000s, when the IT services boom might have been consciously paired with a hardware manufacturing strategy. It was not.
But the second-best time is now. And now — in the specific geopolitical context of 2024 and beyond, with supply chain restructuring creating openings, with democratic alliances creating technology transfer opportunities, with domestic demand providing commercial justification, and with political leadership that has articulated technological self-reliance as a national goal — the conditions for planting that tree are better than they have ever been.
The journey from where India stands today — with its first OSAT facility under construction, its first 28nm fab being planned, its AI mission being launched — to where India needs to be in 2040 and 2047 — a significant semiconductor producer, an AI frontier participant, a sovereign digital power — is long and demanding. It requires sustained effort across multiple electoral cycles, sustained investment across fiscal cycles, and sustained institutional commitment that outlasts individual political leaders.
But it is the necessary journey. The alternative — remaining technologically dependent in a world where technology is sovereignty — is not neutrality. It is vulnerability, dressed in the comfortable clothing of familiarity.
India has always been a civilization that thinks in centuries. The semiconductor challenge requires thinking not just in years but in decades. The investments made now will yield their full fruit in 2035, 2040, 2047. The engineers trained now will be the technology leaders of 2040. The fabs built now will be the foundation of the fabs built in 2035 and the knowledge base of the fabs built in 2045.
Plant the tree now. Tend it with patience, with adequate water (literally, as semiconductor fabs need a lot of water), with the right nutrients of investment and institutional support, and with the long-term commitment that distinguishes a civilization that builds for the future from one that merely manages the present.
The silicon of the Vidharbha or the Rajasthan desert is chemically identical to the silicon of Taiwan's Hsinchu Science Park. The difference is not in the material. The difference is in the institutions, the investments, the decades of accumulated knowledge, and the national will that transform silicon into sovereignty.
India has the silicon. India has the will. What remains is the execution.
This essay has attempted to cover, in one continuous argument, the historical roots, technical dimensions, policy failures, strategic opportunities, and human consequences of India's semiconductor, DRAM, and AI challenge. It draws on publicly available information about the global semiconductor industry, India's policy initiatives, the geopolitics of technology, and the economic development trajectories of technology-leading nations. The specific claims about future technological developments, economic outcomes, and geopolitical scenarios are analytical projections based on current trends and expert assessment, not certainties. The future — as always — will be shaped by the choices made in the present.
Every generation of Indians has faced its defining challenge. Nehru's generation faced the challenge of building a nation from the ruins of colonialism — of feeding a population, establishing democratic institutions, and asserting sovereignty. The generation of 1991 faced the challenge of economic reform — of breaking free from the License Raj and establishing India as a market economy capable of sustained growth. The generation of the 2000s faced the challenge of sustaining growth while addressing inequality and building basic infrastructure.
This generation — the generation of Indian leaders, engineers, entrepreneurs, investors, and citizens active in the 2020s and 2030s — faces the challenge of technological transformation. The challenge is to build the semiconductor, memory, and AI capabilities that will determine India's place in the 21st-century global order. It is not as dramatic as independence, not as politically fraught as economic liberalization, but its consequences are equally momentous.
The cost of getting it right — of making the investments, building the institutions, training the people, and negotiating the partnerships required — is large but finite and manageable. The cost of getting it wrong — of allowing another decade or two to pass while the technological gap with the leading nations widens, while India's digital economy remains built on foreign semiconductor foundations, while other nations' AI systems shape the information that reaches Indian citizens — is incalculable.
India's 1.4 billion citizens deserve sovereignty in the digital age. They deserve healthcare AI that was built with Indian data and understands Indian diseases. They deserve agricultural AI that speaks their language and understands their farming conditions. They deserve smartphones built with Indian chips. They deserve a digital economy whose foundational infrastructure is not beholden to the export control decisions of foreign governments. They deserve to live in a country that is a technology shaper, not merely a technology consumer.
This is achievable. India has the talent — demonstrated in every major technology company in the world. India has the market — the second-largest internet population in the world, growing at extraordinary speed. India has the mathematical tradition — five thousand years of intellectual achievement in exactly the domains that matter for computing and AI. India has growing geopolitical standing — as a major economy, the world's most populous democracy, and a pivotal player in the Indo-Pacific balance of power.
What India needs is the decision — the national commitment, backed by adequate resources, institutional machinery, and sustained political will — to build this capability. The decision to compete not just in the services layer of the technology economy but in its physical and algorithmic foundations. The decision to be a maker, not just a user, of the technologies that will define the century.
Aryabhata described the rotation of the Earth and calculated the length of the year to extraordinary precision — in the 5th century CE, a thousand years before Copernicus. Brahmagupta gave the world zero — the concept without which every computer program, every AI model, every semiconductor design process would be impossible. Srinivasa Ramanujan produced mathematical theorems of incomprehensible elegance with virtually no formal training — theorems that mathematicians are still exploring a century later.
The intellectual heritage is there. The global networks are there. The domestic demand is there. The geopolitical opportunity is there.
The moment is now.
EPILOGUE: THE LONG VIEW
There is a quotation, attributed to various sources but perhaps best associated with the long sweep of Chinese strategic thinking, that "the best time to plant a tree was twenty years ago. The second-best time is now."
India's semiconductor and AI tree should have been planted in the 1980s, when South Korea and Taiwan were planting theirs. It should have been planted in the 1990s, when the digital age was being born. It should have been planted in the 2000s, when the IT services boom might have been consciously paired with a hardware manufacturing strategy. It was not.
But the second-best time is now. And now — in the specific geopolitical context of 2024 and beyond, with supply chain restructuring creating openings, with democratic alliances creating technology transfer opportunities, with domestic demand providing commercial justification, and with political leadership that has articulated technological self-reliance as a national goal — the conditions for planting that tree are better than they have ever been.
The journey from where India stands today — with its first OSAT facility under construction, its first 28nm fab being planned, its AI mission being launched — to where India needs to be in 2040 and 2047 — a significant semiconductor producer, an AI frontier participant, a sovereign digital power — is long and demanding. It requires sustained effort across multiple electoral cycles, sustained investment across fiscal cycles, and sustained institutional commitment that outlasts individual political leaders.
But it is the necessary journey. The alternative — remaining technologically dependent in a world where technology is sovereignty — is not neutrality. It is vulnerability, dressed in the comfortable clothing of familiarity.
India has always been a civilization that thinks in centuries. The semiconductor challenge requires thinking not just in years but in decades. The investments made now will yield their full fruit in 2035, 2040, 2047. The engineers trained now will be the technology leaders of 2040. The fabs built now will be the foundation of the fabs built in 2035 and the knowledge base of the fabs built in 2045.
Plant the tree now. Tend it with patience, with adequate water (literally, as semiconductor fabs need a lot of water), with the right nutrients of investment and institutional support, and with the long-term commitment that distinguishes a civilization that builds for the future from one that merely manages the present.
The silicon of the Vidharbha or the Rajasthan desert is chemically identical to the silicon of Taiwan's Hsinchu Science Park. The difference is not in the material. The difference is in the institutions, the investments, the decades of accumulated knowledge, and the national will that transform silicon into sovereignty.
India has the silicon. India has the will. What remains is the execution.
This essay has attempted to cover, in one continuous argument, the historical roots, technical dimensions, policy failures, strategic opportunities, and human consequences of India's semiconductor, DRAM, and AI challenge. It draws on publicly available information about the global semiconductor industry, India's policy initiatives, the geopolitics of technology, and the economic development trajectories of technology-leading nations. The specific claims about future technological developments, economic outcomes, and geopolitical scenarios are analytical projections based on current trends and expert assessment, not certainties. The future — as always — will be shaped by the choices made in the present.
(This essay is written with the help of AI)
Comments
Post a Comment