For two decades, India's place in the global economy has rested on a deceptively simple equation. The country buys far more physical goods from abroad than it sells — mostly oil, gas, and electronics — but plugs the gap by exporting software services. Engineers in Bengaluru and Hyderabad write code for American banks and European retailers; the dollars come back home; the books balance. In the 2024–25 fiscal year, this arrangement produced a current-account deficit of just 0.6% of GDP, one of the healthiest readings in years. A speculative research note now argues that both halves of that equation are about to break simultaneously, and that financial markets are pricing only the smaller half.
The export side is the part investors already see. Bank analysts at HSBC estimate that 8–10% of Indian IT revenue is exposed to automation, while Kotak forecasts roughly 3.5% pricing erosion as foreign clients demand discounts in exchange for AI-driven productivity. The damage is already visible: Tata Consultancy Services has cut 12,000 jobs, and Infosys has disclosed more than 300 internal AI agents delivering productivity gains of 5–15%. Spread across India's $224 billion IT export base, the note projects trend revenue losses of roughly $30 billion by 2030 and $60–90 billion by 2035.
But here's the catch: that is the smaller problem. The larger one is what replaces those jobs — a recurring import bill, paid in dollars, to a handful of American AI suppliers. The note identifies three engines. First, Indian IT firms themselves will be forced to reallocate 15–25% of their $130 billion industry wage bill to foreign model APIs simply to defend their margins; TCS alone, with 600,000 employees, could become a $2 billion-a-year customer of vendors like OpenAI and Anthropic. Second, consumer subscriptions: ChatGPT Go costs ₹399 a month in India, roughly twice the average wireless bill, and a country with one billion internet users could easily spend $30–40 billion a year on consumer AI by 2035. Third, enterprise software seats, cloud inference billed through Azure and Bedrock, and AI features embedded in existing SaaS products add tens of billions more.
A common objection is that hyperscaler investment in Indian data centers will keep the money onshore. The note rejects this. The chips inside those centers come from Nvidia. The electricity depends on imported coal, gas, and oil — India already buys 80% of its energy from abroad. The models themselves are owned offshore, with gross margins above 70% repatriated to American parents the same way Google's and Meta's Indian ad revenue is today. Domestic data centers, by this accounting, reduce the bill by perhaps 10–20%, not reverse it.
The arithmetic, if the note's mid-case projections hold, is sobering. By 2030, $80 billion in new AI imports plus $30 billion in lost IT exports produces a $110 billion drag on the balance of payments — about 2% of GDP. By 2035, the drag widens to 3–4% of GDP, the zone where rating agencies issue downgrades and currencies come under sustained pressure. India was last in that neighborhood during 2012–13, but escaped over a decade as oil prices and global interest rates normalized. The note's central warning is that AI offers no equivalent reversal: unlike oil-price spikes, software substitution compounds rather than cycles. Once a workflow migrates to a foreign API, it tends to stay there — and the rupee anchor that has held the currency near ₹85 to the dollar may slip toward ₹180–200 by the mid-2030s.
India's economy has balanced on a single trick for twenty years: sell foreign software, buy foreign oil. A speculative research note argues AI is about to shatter both halves of that equation at once.
A research note (cheekily bylined 'Claude, high on hallucinations') argues India faces a coming balance-of-payments crisis driven by AI. The thesis: investors are watching AI destroy Indian IT jobs, but they're ignoring a much bigger problem — a recurring dollar bill for AI services that India will import from American firms like OpenAI, Microsoft, and Google.
Today, India runs a tiny current-account deficit of just 0.6% of GDP, plugged by engineers writing code for foreign clients. The note projects that by 2035, AI could create $175–245 billion in new imports while erasing $60–90 billion of IT export revenue, dragging the deficit to 3–4% of GDP and pushing the rupee from around ₹85 to ₹180–200 per dollar.
The argument isn't really about AI — it's about who owns the toll booth on a road everyone has to drive down. India's IT industry was built on selling cheap, skilled labor; AI changes who collects the rent.
If you're a student anywhere in the world, this preview your future job market. The Indian IT boom was the template for 'safe, middle-class, future-proof career' for an entire generation. AI is testing whether that template survives. For Indian students specifically, it reshapes which engineering specialties pay, whether the rupee in your future paycheck holds its value, and whether sectors like fintech and SaaS can grow without being squeezed by dollar-priced AI bills. Globally, it's a stress test: what happens to any economy that depends on exporting digital labor when that labor gets automated by someone else's machine?
India was last in this neighborhood — a current-account deficit near 4% — during the 2012–13 'taper tantrum,' and it took a decade to climb back. But that crisis ended because oil prices fell and global money got cheaper again. An AI-driven deficit has no equivalent reversal valve. Watch for three signals: whether India tries to build sovereign foundation models (it's trying), whether American AI firms accept rupee-denominated pricing to gain market share, and whether rating agencies start flagging the 'invisible' AI import bill. The deeper question is one every digital-services exporter — the Philippines, Ireland, Poland — will eventually face: when the world's most valuable software runs on someone else's servers, what exactly are you selling?