India built its modern economy on a single competitive advantage. AI is about to remove it.
In the last week of February 2026, India's Nifty IT index hit a two-year low. TCS, Infosys, Wipro, HCL Tech, and Tech Mahindra had collectively shed over 20% of their value since January. Foreign portfolio investors pulled $18 billion out of Indian equities in 2025 alone, the worst year of foreign outflows on record, with IT stocks bearing the sharpest outflows of any sector.
Here is a number that captures the moment precisely: the combined market capitalization of India's five largest IT services companies is roughly $241 billion. Anthropic, a single AI lab founded in 2021, closed its Series G round in February 2026 at a $380 billion valuation. One company building the replacement is now worth more than the entire industry being replaced.
The Indian government, for its part, hosted an AI summit in mid-February. Altman, Amodei, Hassabis and Pichai flew in and praised India's "tech talent." There were photo opportunities. There was no strategy.
This piece is about why that absence of strategy matters more than most currently appreciate.
How India Got Here
In the summer of 1991, the Reserve Bank of India loaded 47 tonnes of gold onto a chartered plane and flew it to the Bank of England. Another 20 tonnes had already been shipped to the Union Bank of Switzerland weeks earlier. The collateral secured $600 million in emergency loans, enough to cover a few weeks of imports and stave off sovereign default. India's foreign exchange reserves had fallen below $1 billion, sufficient to finance roughly three weeks of the country's import bill. The government that authorized the airlift collapsed shortly after.
This was not a policy choice. It was triage. The reforms that followed were conditions attached to a $2.2 billion IMF bailout: currency devaluation, the dismantling of import licensing, tariff reduction, the opening of sectors to foreign investment. The reforms India would later celebrate as a bold economic vision were, at their origin, the terms of a financial rescue. For forty years after independence, India had operated under the License Raj: a system where virtually every economic activity required government permission. Indian manufacturing, insulated from global competition for decades, was globally uncompetitive. The Soviet Union, India's largest trading partner, had just disintegrated. The Gulf War had spiked oil prices and sharply reduced remittances from Indian workers in the Middle East.
The liberalization that India would later celebrate as a bold economic vision was, at its origin, a country with its back against the wall.
What happened next was one of the more improbable economic developments of the late twentieth century. India's engineering colleges had been producing technically competent, English-speaking graduates for decades, originally to staff the public-sector enterprises of the Nehruvian socialist model. Then three things converged in rapid succession: the internet went mainstream, fiber optic cables dropped the cost of transcontinental data transmission, and the Y2K panic sent every major corporation scrambling for programmers. India had, almost by accident, exactly what the market demanded.
The firms that became India's IT giants, TCS, Infosys, and Wipro, had existed before 1991 but were small. Y2K was the catalyst. And critically, the relationships persisted after the crisis passed. Western companies discovered that outsourcing worked. The dot-com bust of 2000–2001 then accelerated the trend rather than killing it: American companies needed to cut costs, and Indian offshore delivery was the mechanism. The bust that destroyed Silicon Valley wealth creation built the Indian IT services industry.
Industry revenue grew from roughly $6 billion in 2001 to $24 billion by 2006, to $67 billion by 2009, to $118 billion by 2015, to $283 billion by 2025, with exports accounting for $224 billion of that total.1 Employment scaled from hundreds of thousands to 5.8 million direct workers, with an estimated 15 to 20 million employed indirectly. Entire cities were remade: Bangalore became a global technology hub. Hyderabad built HITEC City from scratch. Pune, Chennai, Gurgaon were all reshaped by IT services money.
The wealth creation was real and its social effects were profound. A new, broadly meritocratic upper-middle class emerged. An engineering degree became the golden ticket. The pull started early in the pipeline: India's aspiration structure channeled its most capable students toward engineering and computer science from their mid-teens, treating an IT career as the default path to middle-class prosperity. Families across the country invested everything in getting their children into engineering programs because the math was transparent: a job at Infosys meant a salary five to ten times the national average, an apartment in a good neighborhood, a car, international travel. A social contract formed around this, one that nobody wrote down but everyone understood: India's path to prosperity runs through technical education, into IT employment, and out into middle-class consumption.
Meanwhile, the rest of the economy developed a peculiar shape. Manufacturing's share of GDP peaked at 17.9% in 1995, four years after liberalization, and then declined. By 2010 it was 17%. By 2023, 13%. By 2024, it reached an all-time low of 12.5%.2 Agriculture, which still employs roughly 45% of the workforce, contributes only about 15% of GDP. The IT sector grew so fast and paid so well that it pulled India's best-educated workers away from everything else.
By the mid-2020s, India's economy had a distinctive structural profile: a globally competitive IT services sector generating $224 billion in export revenue, sitting on top of a domestic economy that had not fundamentally transformed since liberalization. The IT sector accounts for roughly 7.3% of GDP, 43–45% of total services exports, and its workers, the top 5–10% of earners, drive a disproportionate share of urban discretionary consumption. They are the marginal consumer in India's domestic growth story.
India's IT services exports also do specific mechanical work in the country's balance of payments. India runs a persistent goods trade deficit of $250 to $280 billion annually, importing far more oil, gold, and electronics than it exports in textiles and pharmaceuticals. What keeps this manageable is the services trade surplus, roughly $134 billion, overwhelmingly generated by IT, supplemented by $125 billion in annual remittances from the Indian diaspora.3 These inflows finance the goods deficit and allow the Reserve Bank of India to accumulate the $725 billion in foreign exchange reserves that serve as the country's financial buffer.4
The Proposition
Strip away the historical detail and India's modern economic story reduces to a single proposition: educated Indian workers can perform cognitive work for global clients at a lower cost than the alternative.
Everything rests on this. The $224 billion in services exports. The services surplus that plugs the goods trade deficit. The foreign exchange reserves built over three decades. The urban middle class and its consumption. The real estate markets in Bangalore and Hyderabad. The tax base (only 7–8% of Indians pay income tax, and IT professionals are disproportionately represented). The aspiration structure that channels 1.5 million engineering graduates per year into a specific career path.
The proposition worked because of a specific set of conditions: cheap labor, English fluency, the internet, and the absence of machine intelligence. The first three were structural. The fourth was circumstantial. And the circumstance just changed.
A note on scope. This piece anchors on IT services professionals because they are the most measurable, most concentrated, and most directly threatened cohort. But the underlying argument applies more broadly: to business process outsourcing, knowledge process outsourcing, financial back-office operations, legal process outsourcing, and the growing network of Global Capability Centers. The common thread is the same proposition: educated Indians performing cognitive work at a computer for foreign clients at a fraction of the Western cost. Any work that is primarily digital, primarily cognitive, and primarily performed at a screen falls within the blast radius. The piece uses "IT services" as shorthand, but the structural exposure extends well beyond any single sector label.
The Technology
There is a tendency, still widespread as of early 2026, to evaluate AI as a software product. People interact with chatbots, see code suggestions, watch demos. They assess what the tools can do today and project linearly from there. They think in terms of productivity software that makes existing workers faster. This fundamentally misreads what is happening. What has arrived is not a tool that augments cognitive work. It is a replacement for cognitive work itself.
Every frontier AI lab is working toward the same goal: automate coding as completely as possible, then use automated coding to accelerate AI research itself. This is the self-improvement loop. Better AI builds better AI builds better AI.5 Thus far, the rate of capability improvement has been driven by scaling laws and the sequential speed of human researchers. What happens when the models start improving themselves is that the rate shifts from steep to exponential.
This matters for the India story because of what it implies about the timeline. You can argue, as former RBI Governor Raghuram Rajan does, that technology adoption lags innovation. He invokes the telephone exchange: automated switching was invented in the 1920s, the last human operator was removed in the 1980s. Fair point. Enterprise IT is spectacularly sticky. Large companies do not rip out their systems overnight.
But here is the counterpoint: even if adoption lags, and it does, the innovation curve is exponential. This means that even the lagging adoption represents a massive structural disruption. Every quarter that passes without full deployment does not buy the incumbent economy time. It accumulates unrealized capability: a growing reservoir of displacement potential that builds pressure regardless of whether any individual company has flipped the switch yet. The question is not whether AI will be adopted. It is how much structural change has already been locked in by the time adoption catches up. For an economy as dependent on the pre-AI model as India's, this distinction is existential.
The more fundamental problem with Rajan's defense is his invocation of the lump of labor fallacy, which holds that previous waves of automation created more jobs than they destroyed because cheaper services generate more demand. This is historically true for technologies that automate narrow tasks. It does not hold for artificial general intelligence.6 The "general" in AGI means general-purpose cognitive replacement. When the machine can perform most of the cognitive tasks that displaced workers would naturally migrate to, and can improve at those tasks faster than humans can retrain, the fallacy is on the person citing the fallacy.
Rajan, to his credit, has articulated the counterarguments to the India bear case more rigorously than perhaps anyone else, compressing them into a single Bloomberg interview in February 2026 with Haslinda Amin and Menaka Doshi.7 The fact that I engage with his arguments at length is a reflection of their quality, not a dismissal. But his defense does not survive scrutiny on the specific question of India's structural exposure.
His argument that an Indian consultant equipped with AI tools still costs one-fifth of a Western consultant with the same tools misunderstands the nature of the disruption. The relevant comparison is not an Indian consultant versus an American consultant, each augmented by AI. It is one American consultant orchestrating dozens of AI agents versus a team of Indian consultants doing the work manually. And here the currency dynamics become directly relevant: the American consultant, earning and spending in dollars, can deploy AI agents priced in dollars at a fraction of their revenue. The Indian firm, earning in rupees but paying for AI in dollars, faces a structural cost disadvantage in the very tools needed to stay competitive.
More fundamentally, as AI handles a larger share of the actual cognitive output, human labor shifts from being the primary cost to a minority cost in the equation. The human becomes valuable not as a doer of work but as an orchestrator: a manager of AI agents, a translator of business context into technical direction. And if the human's role is judgment and orchestration rather than volume of output, it becomes worth paying a premium for the best person in that role regardless of geography. The cost advantage that defined Indian IT, more bodies for fewer dollars, loses its meaning when the bodies are no longer the unit of production.
The Currency Trap
There is a specific mechanism by which this disruption becomes self-reinforcing, and it has to do with how AI services are priced.
The entire Indian IT model worked because of a currency asymmetry. Revenue in dollars, costs in rupees. TCS bills JPMorgan in USD, pays its engineers in INR. The spread between what a dollar buys in Bangalore versus what it buys in New York is the margin. Every time the rupee weakened, Indian IT margins actually improved: cheaper inputs, same-priced outputs. For thirty years, currency depreciation was a tailwind.
AI flips this. Tokens, API calls, cloud compute are all priced in USD. There is no purchasing power parity discount on a Claude API call. A token costs the same whether you are calling it from San Francisco or Bangalore. The "labor" moved from humans in a low-cost economy to silicon in USD-denominated data centers.
The asymmetry does not disappear. It inverts. Indian companies go from "our inputs are cheap in dollar terms" to "our inputs are expensive in rupee terms." The structural advantage that built the industry becomes a structural disadvantage in the successor model.
The obvious response is: build the data centers in India. And there are moves in this direction, with data localization rules potentially accelerating the trend. But this does not solve the currency problem. The inputs for those data centers are still priced in dollars. Nvidia GPUs are dollar-denominated. The networking equipment, the cooling infrastructure, the advanced power systems are largely imported. India does not manufacture these components domestically at scale. Building a data center in Bangalore does not make the chips inside it cheaper. It may actually make the problem worse, because now you are importing more dollar-denominated capital goods while your dollar-earning capacity is shrinking.
And here is where the vicious cycle kicks in. As AI disrupts IT services exports, fewer dollars flow into India. Fewer dollars mean the rupee weakens. A weaker rupee makes dollar-priced AI tools more expensive for Indian companies trying to adopt them to stay competitive. Which makes it harder to compete. Which means more revenue loss. Which means fewer dollars. Which means more rupee weakness. Which makes AI tools even more expensive.
At the national level, the flow does not just shrink. It threatens to reverse direction. India goes from being a net dollar earner in cognitive services to a net dollar spender on cognitive technology. Instead of Indian engineers generating dollar inflows by selling their labor, Indian enterprises hemorrhage dollars buying compute and AI services from American providers.
The Domino Chain
If the disruption moves at the pace I believe it will, here is what the chain of consequences looks like.
IT services exports decline sharply. The services surplus, which currently plugs India's massive goods trade deficit, shrinks or disappears. The current account deficit widens into territory that markets treat as a crisis signal for an emerging economy.
The rupee falls. Not just because the current account is hemorrhaging, but because foreign portfolio investors read the same data and pull capital, hitting the capital account simultaneously. The Reserve Bank burns through reserves trying to slow the decline.8
The weaker rupee makes imports more expensive. Oil, electronics, industrial inputs all cost more in rupee terms. Inflation rises. The RBI is forced to raise interest rates to fight inflation at exactly the moment the economy needs loose monetary policy for growth. This is the classic emerging market trap: you cannot do both.
Real estate in tech cities like Bangalore, Hyderabad, Pune, and Gurgaon enters a demand collapse. Developers who built apartments and office parks for a permanently expanding IT workforce find that the expansion has reversed. Commercial vacancy rates spike. Residential prices correct sharply. Developers default on loans.
Banks feel the stress. The exposure is concentrated: personal and home loans to IT professionals, commercial real estate lending, corporate loans to IT firms. Private banks and NBFCs with heavy penetration in tech cities see delinquencies rise. Credit tightens. The stress spills into unrelated sectors.
The fiscal position deteriorates. IT professionals and companies are among India's largest taxpayers. Revenue falls at the precise moment the government needs to spend on safety nets, retraining, and infrastructure. The fiscal deficit widens.
This is not India's 2008. India's financial system is less leveraged than America's was; it does not have the PE-backed LBO daisy chains or the $13 trillion mortgage market. This is more like the UK after the financial crisis hollowed out the City of London: a structural shock to the highest-value sector that permanently lowers the growth trajectory, creates persistent unemployment in a formerly privileged class, and reshapes the political economy for a generation.
The Way Out (Each Path Is Harder Than It Sounds)
If the services export story dies, what is India's actual path forward? The honest answer is that there is no clean one visible.
The Manufacturing Pivot
Everyone reaches for this first. Make in India. PLI schemes. Apple is moving iPhone production. But India missed the window. The East Asian model, Japan in the 60s, Korea in the 80s, China in the 2000s, worked because those countries industrialized when manufacturing was labor-intensive and global trade was expanding. The world they industrialized into no longer exists. India skipped that step and went straight to services. It never built the hard infrastructure that manufacturing at scale requires. And the window is closing from the other direction: the factories India would need to build in 2026 are not the labor-intensive factories of 1990s China. They are increasingly automated. Robots do not care about currency depreciation. India would be trying to industrialize into a sector undergoing the same automation transformation that is killing services.10
The Domestic Digital Economy
UPI and Aadhaar are genuine achievements. They reduce friction in transactions. But reducing friction in a $4.2 trillion economy does not create new wealth. It moves existing wealth around more efficiently. If the people transacting on UPI have less money because their jobs disappeared, UPI just processes fewer and smaller transactions faster. It is plumbing, not an engine.
The Startup Ecosystem
This is where optimism is most warranted, and most fragile. Displaced, technically skilled workers could start companies. But startups need capital, and capital follows growth narratives. If the India growth story is impaired, venture funding dries up. Indian startup funding has already been declining from its 2021–2022 peaks. And the types of companies these displaced workers would start are disproportionately likely to be in the same sectors being disrupted by AI. A former TCS engineer probably starts a SaaS company. He is competing against the very force that displaced him.
India Builds Its Own AI
India does not have the compute infrastructure, the research talent pipeline, or the capital to compete with OpenAI, Anthropic, Google DeepMind, or Chinese labs. Frontier models require tens of billions in GPU investment. India is not in this race. It could build niche applications on existing models, but that is a much smaller opportunity and it does not employ millions.
What is left is the thing India has been neglecting: the sheer scale of its domestic physical economy. India is urbanizing at a massive rate. It needs tens of millions of housing units, roads, water systems, hospitals, schools. Construction employs 60+ million people and contributes 8% of GDP. Healthcare faces an acute shortage. These are real needs that require real work.
But these sectors are low-productivity, low-wage, and largely informal. They do not generate export earnings. They do not produce high-income consumers. They do not attract foreign capital. The economy does not collapse if it falls back on them. It stagnates. Growth drops. The aspirational middle class shrinks instead of expanding. India becomes a much poorer version of what it thought it was becoming.
The Political Powder Keg
India produces 1.5 million engineering graduates per year. That pipeline was purpose-built for the IT services industry.9 Those graduates expected to get a TCS or Infosys offer, earn a starting salary that placed them in the top tier of Indian earners, and join the urban middle class. When campus hiring dries up and layoffs hit the breadwinners of families who built their entire financial lives around IT salaries, you do not just have an economic problem. You have 1.5 million young people every year entering a labor market that has no use for the skills they spent four years acquiring, on top of the millions already being displaced.
Economic stress of this nature, millions of educated young people with no economic prospects, has historical precedent. Weimar Germany's radicalization was partly driven by educated professionals whose economic status collapsed. The Arab Spring was fueled by young graduates who could not find work. The pattern is consistent: when an educated class is promised a trajectory and then that trajectory vanishes, the result is not quiet resignation. It is instability.
In India's specific case, the political combustion risk is amplified by an existing dynamic. The current government has, for over a decade, used communal and religious polarization as a political tool, a way to maintain support when economic delivery is insufficient. When the economy was growing and the IT sector was hiring, this was an overlay on a broadly functional system. When the economy contracts, when millions of educated young people cannot find work, and when the government's primary response is to press the communal division button rather than address structural economic failure, the mixture becomes significantly more volatile.
This is not a prediction. It is a structural observation. Populations under economic stress that are simultaneously subjected to identity-based political mobilization have a poor track record of stability. India has all of the ingredients.
The Honest Assessment
Let me be precise about what I am and am not arguing.
I am not arguing that India will collapse. India's financial system is less leveraged than most Western economies. Its $725 billion in reserves buy two to three years of managed depreciation.
But seventy percent of the population living in a rural or semi-urban economy does not insulate them from what is coming. This piece has laid out exactly how the second-order effects reach everyone: through a weaker currency that makes everything from cooking oil to fuel more expensive, through inflation that erodes purchasing power across all income levels, through a fiscal squeeze that starves public spending on the services that the poorest Indians depend on most, and through the evaporation of the urban consumption that IT workers drive, consumption that sustains jobs in retail, restaurants, transportation, construction, and domestic services far beyond the IT sector itself. The contagion does not stop at city limits.
What I am arguing is that India's 35-year growth model, the one that lifted tens of millions into the middle class, built Bangalore and Hyderabad into globally connected cities, generated the foreign exchange that kept the country solvent, and created an aspiration structure for an entire generation, is built on a single competitive advantage that is now being eroded by a technology that India did not create, cannot compete with, and must pay for in a currency it can no longer earn at the rate it needs to.
Every government since 1991 knew that India was overexposed to services exports and underinvested in manufacturing, agriculture, education, and physical infrastructure. They all talked about fixing it. None executed at the scale required. The window for gradual diversification is now closing much faster than anyone expected.
I believe the structural erosion will become undeniable within 2026 itself. Not as a sudden collapse, but as accelerating contract cancellations, hiring freezes hardening into layoffs, campus recruitment falling off a cliff, and the first visible protests from a generation that was promised a future that is not arriving. The timeline is not five years. It is months.
India did not choose to become a services economy. It backed into one, in a moment of crisis in 1991, and discovered an advantage that it rode for thirty-five years. That advantage was built on a gap: the gap between Indian wages and Western wages for equivalent cognitive work. AI does not narrow that gap. It makes the gap irrelevant.
What India needs now is something no Indian government has ever executed: rapid, large-scale economic restructuring in the span of a few years, of the kind that took Korea twenty years and China thirty. While managing a currency crisis, a fiscal squeeze, and a generation of young people who were promised a future that is not coming.
It is not impossible. But nobody currently in charge appears to understand the urgency, let alone possess the vision to act on it.