By BarathVector Editorial — 2026-03-07

The Technopolar Tilt: In the GPU Arms Race, India Arrives With a Spreadsheet

By BarathVector Editorial

There is a session at the Raisina Dialogue 2026 -- currently underway in New Delhi from March 5 to 7 -- that examines deterrence in the Taiwan Strait through the lens of semiconductor dependence. The organisers call it "Pax Silica." The phrase is borrowed from a thesis that has gained quiet traction in foreign policy circles: that the thing preventing a Chinese invasion of Taiwan is not American aircraft carriers but the fact that TSMC fabricates over 90 per cent of the world's most advanced chips. Blow up Taiwan, and you blow up the global economy. Silicon, not missiles, keeps the peace.

It is a clarifying idea. And if you sit with it long enough, it becomes a deeply uncomfortable one for India.

Because India -- the nation that built Aadhaar for 1.4 billion people, that processes 21.7 billion UPI transactions in a single month, that rolled out CoWIN to vaccinate a population larger than Europe and North America combined -- imports roughly 90 per cent of its semiconductors. Every chip in every phone that taps a QR code at a Chennai chai stall was designed in California, fabricated in Hsinchu, and packaged in Penang. India wrote the software. Someone else made the machine it runs on.

In a technopolar world, that is not a quirky asymmetry. It is a strategic emergency.

The Samskara Paradox

The theme of this year's Raisina Dialogue is "Samskara" -- a Sanskrit term loosely meaning the impressions that shape future action. The Observer Research Foundation and the Ministry of External Affairs, who co-organise the event, have assembled delegates from 110 countries to discuss how technological disruption, strategic competition, and economic security are reshaping global politics. Finland's President Alexander Stubb is the chief guest. The guest list reads like a who's-who of the post-liberal international order: defence ministers, tech CEOs, former heads of state, military commanders, and the inevitable battalion of think-tank analysts.

The core thesis animating the proceedings is blunt: we have entered a "technopolar" world, where influence is determined not by the size of your army but by your position in the AI and semiconductor supply chain. Traditional alliances matter less than who controls the fabs, the rare earths, the GPU clusters, and the submarine cables.

India's position in this new order is genuinely paradoxical. On one axis, it is a frontrunner. On the other, it is a dependent.

The Software Superpower That Can't Make a Chip

Start with what India has built. The numbers are not merely impressive; they are civilisational in scale.

UPI processed 21.7 billion transactions in January 2026 alone -- a new monthly record -- worth Rs 28.33 lakh crore. Annualised, the system is on track for 379 billion transactions in FY2026-27, accounting for roughly 90 per cent of all digital payments in India. No country on earth has achieved this kind of financial inclusion at this velocity. Not China. Not the United States. Not anyone.

Aadhaar covers 1.4 billion identities and handles 2.5 billion authentication transactions per month. It has saved the government an estimated Rs 3.48 trillion in subsidy leakage -- the kind of figure that would make a European finance minister weep with envy. The UIDAI recently launched an AI-powered "Invisible Shield" platform for biometric deduplication, and Aadhaar is being integrated into Google Wallet and Samsung Wallet.

India Stack -- the layered architecture of Aadhaar, UPI, DigiLocker, and eKYC -- is now studied and replicated by governments from Singapore to Brazil. At the India AI Impact Summit in February 2026, which drew delegates from over 100 countries, the phrase "Manhattan Project for AI for public good" was used without irony. Ninety-two countries endorsed the India AI Impact Summit Declaration. Google's Sundar Pichai announced a USD 15 billion AI hub in Visakhapatnam. India set a Guinness World Record for AI responsibility pledges.

And yet.

India has zero operational advanced semiconductor fabrication plants. It designs some chips. It assembles some chips. It does not manufacture the advanced logic chips that power AI inference, military systems, or the very UPI servers that process those 21.7 billion transactions. The silicon beneath India's digital miracle is entirely borrowed.

38,000 GPUs and a Prayer

The hardware deficit extends beyond chip fabrication into the raw computational infrastructure that AI demands. At the India AI Impact Summit, IT Minister Ashwini Vaishnaw announced that India currently has 38,000 GPUs under the IndiaAI Mission, with 20,000 more to be added "in the coming weeks." The government's ambition is to scale to 200,000 GPUs, with a target of attracting over USD 200 billion in AI infrastructure investment by 2028.

These are serious numbers. They are also, by global standards, modest.

Microsoft alone has committed to spending USD 80 billion on AI data centres in a single fiscal year. Meta's GPU cluster at its Louisiana facility reportedly exceeds 100,000 H100 equivalents. China, despite US export controls, has stockpiled an estimated 500,000 to one million high-end GPUs through grey-market channels and domestic alternatives. India's 38,000 GPUs -- offered to startups at the subsidised rate of Rs 65 per hour -- represent a commendable democratisation effort. But they do not represent sovereign AI capability at scale.

The gap is structural. India's semiconductor market is projected to reach USD 55-63 billion by 2026, but the overwhelming majority of that demand is met by imports. The India Semiconductor Mission 2.0, announced in the Union Budget on February 1, 2026, shifts focus from chip assembly to full-scale manufacturing, equipment production, and ecosystem development. The government allocated Rs 1,000 crore for the mission. Ten semiconductor projects across six states have been approved, representing Rs 1.60 lakh crore in investment. Tata Electronics plans to bring its Assam facility to commercial readiness by mid-2026. Micron, CG Power, and Kaynes Technology are also entering commercial operations.

All of this is necessary. None of it is sufficient. The plants coming online are overwhelmingly in assembly, testing, and packaging -- the lower end of the semiconductor value chain. India is not fabricating 3nm or 5nm logic chips. It is not going to fabricate them this decade. The question is whether it can close the gap before the technopolar order calcifies into permanent hierarchy.

The Rs 7.85 Lakh Crore Question

India's defence budget for FY2026-27 offers a revealing lens on the technopolar tension. At Rs 7.85 lakh crore -- approximately USD 86.7 billion -- it is the highest allocation ever, a 15.19 per cent increase over the previous year, and the single largest ministerial share of the Union Budget at 14.67 per cent.

The capital allocation for modernisation stands at Rs 2,19,306 crore, a 21.8 per cent jump. The "Other Equipment" head -- covering network-centric warfare tools and ground-based systems -- surged 30.3 per cent to Rs 82,217 crore. Aircraft and aero engines received Rs 63,733 crore. DRDO's allocation rose to Rs 29,100 crore, with Rs 17,250 crore earmarked for capital R&D expenditure.

These are numbers that signal intent. India is investing heavily in indigenous defence under Atmanirbhar Bharat. But the hard question that the Raisina panellists are dancing around is this: how much of that "indigenous" defence technology runs on imported chips? How many of those network-centric warfare tools depend on semiconductor supply chains that route through East Asian fabs over which India has zero leverage?

The next war -- if it comes -- will not be fought with tanks. It will be fought with autonomous drones, AI-powered surveillance, cyber weapons, and electronic warfare systems. Every one of those capabilities is a semiconductor-dependent capability. A nation that imports 90 per cent of its chips is a nation that has outsourced a non-trivial portion of its defence sovereignty to foreign fabrication lines.

IMEC: The Corridor That Could

This is where the India-Middle East-Europe Economic Corridor enters the frame -- not merely as a trade route, but as the connective tissue of a technopolar counter-strategy.

IMEC, first announced at the G20 in September 2023 under India's presidency, is designed as a multimodal corridor linking India to Europe via the UAE, Saudi Arabia, Jordan, and Israel. It is explicitly positioned as a counter to China's Belt and Road Initiative, and its ambitions extend beyond shipping containers. The corridor envisions integrated rail, shipping, data cables, hydrogen pipelines, and electricity interconnects.

The cynics -- and there are many -- point out that IMEC has been long on rhetoric and short on implementation. India has not established a dedicated implementing body. Funding commitments remain vague. UAE-Saudi policy competition complicates coordination. And the Israel-Palestine situation introduces a geopolitical volatility that no infrastructure project can simply engineer around.

But IMEC received a significant boost in early 2026. The India-EU Free Trade Agreement, signed in January, gives the corridor commercial rationale. The "IMEC Plus" framework now seeks to integrate African gateways via Egypt. A Trump White House official, Ricky Gill, is reportedly planning a summit-level visit to India focused specifically on IMEC. During Modi's visit to Israel in February 2026, the corridor was prominently discussed. And the 2025 Trieste Summit positioned the Italian port as IMEC's European terminus.

Here is the argument that India's strategic planners need to internalise: IMEC is not just a trade corridor. It is a data corridor. The submarine cables, the 5G infrastructure, the energy interconnects -- these are the physical layer of a digital supply chain. If India cannot fabricate its own chips, it can at least control the routes along which chips, data, and energy flow. IMEC, done right, gives India infrastructural leverage in the technopolar order even while the fabrication gap persists.

Done wrong -- or done slowly -- it becomes another PowerPoint presentation at another multilateral summit.

The Pax Silica Dilemma

Return, for a moment, to the Raisina session on Taiwan. The "Pax Silica" thesis holds that semiconductor dependence is itself a form of deterrence: China does not invade Taiwan because it cannot afford to lose access to TSMC. In January 2026, Taiwan and the US signed a statement endorsing Pax Silica principles, formalising what was previously an implicit strategic understanding.

India's relationship with this framework is that of a bystander with high stakes and low agency. India depends on the same Taiwanese fabs that China covets. If the Taiwan Strait destabilises -- through blockade, invasion, or coercion -- India's digital infrastructure faces supply-chain disruption with no domestic fallback. The UPI system, the Aadhaar authentication network, the AI compute clusters, the defence electronics -- all of it runs on silicon that India cannot produce.

This is not hypothetical risk-modelling. This is the current architecture of Indian sovereignty, and it has a single point of failure located on an island 3,000 kilometres away, across a strait that a nuclear-armed revisionist power considers its own territory.

The Seven Sovereign AI Models and the Missing Fab

At the India AI Impact Summit, several sovereign AI models were unveiled -- Indian-built foundation models tested against global benchmarks and, by some metrics, outperforming larger international systems. This is genuinely significant. India is demonstrating that it can build competitive AI at the model layer.

But a sovereign AI model running on imported GPUs in a data centre powered by imported chips is sovereign only in the narrow sense of intellectual property. The compute layer, the hardware layer, the fabrication layer -- these remain foreign dependencies. Sovereignty, in the technopolar era, is a full-stack proposition. You need the model, the data, the compute, the chips, and the rare earths. India has the first two. It is building the third. It does not have the fourth. And it has barely begun thinking about the fifth.

The Rs 1,000 crore allocated to Semiconductor Mission 2.0 is a statement of direction, not a statement of scale. For context, TSMC's capital expenditure in 2025 alone exceeded USD 30 billion. Intel's fab investments in the US and Europe total over USD 100 billion across the decade. South Korea's semiconductor support package exceeds USD 470 billion. India's Rs 1,000 crore -- roughly USD 115 million -- is a rounding error in these ledgers.

The ten approved semiconductor projects, at Rs 1.60 lakh crore in total investment, are more meaningful. But the timeline matters. Tata Electronics in Assam reaching commercial readiness by mid-2026 is a milestone for assembly and packaging, not for advanced logic fabrication. India is entering the semiconductor supply chain at the low-value segments precisely when the technopolar order is being determined by whoever controls the high-value segments.

A Digital Colony That Writes Excellent Code

There is an uncomfortable historical parallel here. In the 19th century, India was the world's largest producer of raw cotton. It exported bales to Lancashire, which turned them into textiles, which were sold back to India at a markup. India had the raw material. Britain had the machines. The value accrued to the machines.

In the 21st century, India has the world's most sophisticated digital governance infrastructure. It has the developers, the architects, the data scientists. It exports software services worth over USD 250 billion annually. But it imports the chips on which all of that software runs. The value, once again, accrues to the machines.

The phrase "digital colony" is provocative, and it should be. India is not a colony -- it is a sovereign democracy with genuine agency. But the structural dependency on imported silicon creates a form of technological subordination that is incompatible with the superpower ambitions that Raisina Dialogue panels so enthusiastically project.

The technopolar order does not care about your democratic credentials or your civilisational heritage. It cares about whether you can fabricate a 3nm chip. India cannot. Not yet.

Forty Years of Talking About It

There is a question that no panel at Raisina will ask, because it is too uncomfortable. It is this: why?

Taiwan founded TSMC in 1987. South Korea poured national will into Samsung's semiconductor division in the early 1980s. Both were smaller economies than India. Both had fewer engineers. Neither had a civilisational tradition of mathematical excellence stretching back to Aryabhata. And yet, by the time a teenager in 1980s India was reading about Taiwan's chip miracle, the race was already being lost.

India was not asleep. Rajiv Gandhi understood, perhaps more clearly than any Indian leader before or since, that technology was destiny. He liberalised computer imports, launched the Centre for Development of Telematics (C-DOT), pushed computerisation into government, and seeded the IT revolution that would eventually produce Infosys, Wipro, and TCS. The Semiconductor Complex Limited (SCL) in Mohali was established in 1983 -- India's first and, for decades, only chip fabrication facility. Gandhi saw the future. He did not get to build it. An assassin's bomb in 1991 ended the most technologically literate premiership India has ever had.

What followed was a study in structural failure.

The 1991 liberalisation, necessary and transformative as it was, made a fateful choice: software over hardware. Services over manufacturing. The logic was seductive -- software required minimal capital, India had English-speaking engineers, and the margins were extraordinary. Why build fabs that cost billions and take a decade to mature when you could build IT companies that cost thousands and generated revenue in months? Bangalore became a global back office. The chip question was quietly shelved.

SCL Mohali suffered a devastating fire in 1989 that destroyed its clean room. It was rebuilt, slowly, but never upgraded beyond 180nm technology -- a node that the rest of the world had abandoned by 2005. Today, SCL fabricates chips for ISRO and defence applications at geometries that would be considered antique in Hsinchu or Suwon. It is a monument to what happens when strategic assets are maintained with maintenance budgets instead of moonshot investments.

The pattern repeated with numbing regularity. In 2007, SemIndia proposed a USD 3 billion fab in Hyderabad. It collapsed. In 2014, Jaiprakash Associates and IBM consortium promised a fab in Uttar Pradesh. It evaporated. HSMC (Hindustan Semiconductor Manufacturing Corporation) was announced with fanfare. It went nowhere. Each failure shared common ingredients: inadequate government support, byzantine land acquisition, unreliable power infrastructure, and a political class that could not distinguish a nanometre from a kilometre.

This is the crux. Building a semiconductor industry requires three things that Indian politics has historically been unable to deliver simultaneously: patient capital, institutional continuity, and technical literacy at the decision-making level.

Patient capital means committing tens of billions of dollars to projects that will not generate returns for a decade. Indian electoral cycles run on five-year horizons. A Chief Minister who breaks ground on a fab in year one will not cut the ribbon on production in year five. The next government will. In a political culture where infrastructure projects are campaign props, semiconductor fabs are terrible politics -- invisible, incomprehensible to voters, and profitable only for your successor.

Institutional continuity means keeping the same team, the same policy framework, and the same incentive structure in place across administrations. Taiwan's Industrial Technology Research Institute (ITRI) incubated TSMC over fifteen years of consistent government support. South Korea's semiconductor drive survived military dictatorships, democratic transitions, and financial crises without losing focus. India's semiconductor policy has been rewritten, renamed, and relaunched with every new government. The India Semiconductor Mission is the latest iteration of an ambition that has been announced, in various forms, at least six times since the 1990s.

Technical literacy at the decision-making level means that the ministers, secretaries, and committee chairs who approve semiconductor investments must understand what they are approving. This is not elitism -- it is competence. When a politician evaluates a highway project, they can drive on a highway. When they evaluate a hospital, they have been to a hospital. When they evaluate a 5nm fab proposal, they are evaluating something they have never seen, cannot visualise, and whose economics they must take on faith from consultants who may or may not have conflicts of interest. The result is either paralysis or gullibility -- either nothing gets approved, or the wrong things get approved for the wrong reasons.

Taiwan did not build TSMC because it had superior engineers. It built TSMC because Morris Chang -- a semiconductor executive, not a politician -- was given institutional authority, patient capital, and political air cover by a government that understood the strategic stakes. South Korea did not build Samsung Foundry because of market forces. It built it because the state directed capital into semiconductors with the same intensity it directed capital into steel and shipbuilding a generation earlier.

India had the engineers. It had the mathematical talent. It had, in Rajiv Gandhi, a leader who grasped the thesis. What it did not have -- and arguably still does not have -- is a political system capable of making a thirty-year bet on a technology that most of its parliamentarians cannot explain.

The Rs 1,000 crore allocated to Semiconductor Mission 2.0 is not evidence that this has changed. It is evidence that it has not. A thousand crore for semiconductor sovereignty, in a budget that allocates Rs 7.85 lakh crore for defence, is not a bet. It is a gesture. And gestures do not fabricate chips.

What Must Be Done

The path forward is not mysterious. It is expensive, slow, and politically unglamorous -- which is precisely why it has not been pursued with sufficient urgency.

First, semiconductor investment must be scaled by at least an order of magnitude. The Rs 1,000 crore for Semiconductor Mission 2.0 needs to become Rs 1 lakh crore over the decade. This is not charity; it is infrastructure spending with national security implications. Defence budgets that allocate Rs 82,000 crore for "other equipment" should be asking hard questions about the provenance of the silicon inside that equipment.

Second, IMEC must be treated as critical national infrastructure, not a diplomatic talking point. India needs a dedicated implementing body, committed funding, and a timeline that is measured in construction milestones, not communique paragraphs. The submarine cable and data corridor components of IMEC deserve priority equal to the shipping and rail elements.

Third, India must build strategic semiconductor stockpiles and diversify supply chains beyond Taiwan. South Korea, Japan, and the emerging US fab capacity offer alternatives, but alternatives require trade agreements, joint ventures, and the kind of industrial diplomacy that India has historically been slow to execute.

Fourth, the GPU compute buildout must accelerate. The target of 200,000 GPUs is correct in direction but needs a deadline sharper than "by 2028." AI capability compounds. Every month of delay is a month in which other nations train larger models, deploy more capable systems, and entrench their advantages.

The Spreadsheet at a Gunfight

The title of this essay is not entirely fair. India has not arrived at the technopolar contest with merely a spreadsheet. It has arrived with Aadhaar, UPI, India Stack, CoWIN, a 200,000-GPU aspiration, ten semiconductor projects, an IMEC corridor, an Rs 7.85 lakh crore defence budget, and a New Delhi Declaration endorsed by 92 countries. That is not nothing. That is, by most measures, formidable.

But formidable in software. Formidable in governance. Formidable in ambition and articulation.

The hardware gap remains. And in a technopolar world, the hardware gap is the gap that matters. The next superpower will not be the nation with the best app. It will be the nation that makes the chip the app runs on.

India has approximately five years to close this gap before the technopolar hierarchy becomes permanent. The Raisina Dialogue is the right place to discuss it. The question is whether discussion will translate into fabrication -- not of arguments, but of silicon.

The spreadsheet is excellent. Now build the factory.

The Raisina Dialogue 2026 continues through March 7 in New Delhi. BarathVector will provide ongoing coverage of the technology and semiconductor discussions.