AI’s hidden cost for India

Every January, Davos produces a familiar genre of writing: forecasts, anxieties, optimism dressed as realism. In 2026, artificial intelligence will dominate that genre almost completely. At the World Economic Forum, AI is discussed as a productivity engine, a labour-market disruptor, a geopolitical accelerant. All of that is true. But from India’s perspective, something more prosaic - and more dangerous — is taking shape beneath the rhetoric. India may be drifting towards an AI-driven current account problem.

To understand why AI changes the picture, it helps to begin with what has kept India stable for years. India has lived with a chronic merchandise trade deficit for decades. Yet the current account has remained broadly manageable because India exports something else: services and skills. Software services, business process outsourcing, consulting, engineering work, and remittances from skilled labour abroad have functioned as India’s macroeconomic shock absorbers. When goods trade bleeds, services income cauterises the wound. It rests on a simple asymmetry: India imports physical inputs and exports human cognition. Artificial intelligence unsettles that asymmetry. AI is often described as a technology. In macroeconomic terms, it behaves more like infrastructure. At scale, AI is not a one-time purchase. It is a continuous service: cloud systems, model access, inference capacity, security updates, compliance layers, constant upgrades. These services are paid for every month, often in foreign currency, and increasingly to a very small set of global providers.

What matters is not the stock price story, but what those firms are buying with that capital: data centres, long-term power contracts, advanced chips, proprietary models. This is physical, durable infrastructure. Once infrastructure is built, pricing power follows. For India, this means AI adoption does not merely improve productivity; it creates a recurring import stream — one that does not show up clearly in customs data, but accumulates across sectors. In balance-of-payments terms, AI begins to resemble energy more than software.

Firms integrate AI into logistics, finance, marketing, compliance. Government departments deploy AI tools for fraud detection, analytics, and service delivery. Each deployment seems rational, even cost-saving. But collectively, they embed foreign intelligence infrastructure more deeply into the economy. This is how structural dependence forms: not through a single decision, but through thousands of sensible ones. Unlike oil, AI imports do not spike visibly with prices. They rise steadily with usage. Capital inflows can mask the effect for years. By the time pressure appears in the exchange rate or reserves, the dependency is already baked in. The problem is that AI does not merely add an import bill. It collides directly with the export engine that has sustained India’s external balance. Tasks that once justified cross-border billing — code review, customer support, analytics, documentation, compliance — are increasingly bundled into AI platforms and priced per query or per token. The more capable the model, the less pricing power remains with the human intermediary.

This creates a slow pincer movement in the current account:  On one side, AI services imports rise, On the other hand, the unit value of services exports comes under pressure. India could train millions of AI-literate workers and still see a weaker services surplus, not because demand collapses, but because rents have moved upstream.

There is another dimension that deserves more attention: the Indian state itself. India’s governance model is now digitally intensive at population scale - payments, identity, welfare, taxation, compliance. As AI is layered onto these systems, the state becomes a large and long-term consumer of AI services. This has fiscal implications, but also external ones. Licensing costs, cloud contracts, security audits, and vendor lock-in gradually create foreign-currency obligations embedded in governance itself. The WEF’s Global Cybersecurity Outlook for 2026 notes that public-sector cyber resilience lags private-sector resilience even as AI-driven threats grow. For India, where digital systems operate at unprecedented scale, cyber failure is not just a technical risk. It is a macroeconomic and political risk. A state that cannot function without rented intelligence is not resilient. It is fragile. Nothing here implies that India should slow AI adoption. That would be self-defeating. The warning is narrower and more serious.  AI changes the structure of India’s external balance, and pretending otherwise will make policy reactive rather than strategic. Countries that own AI infrastructure will export cognition. Countries that do not will import it - permanently. Growth will continue in both.

Power will not. India has encountered this pattern before with commodities and capital goods. The difference now is that AI embeds itself into decision-making itself. Once embedded, exit is not merely expensive; it becomes institutionally disruptive. WEF 2026 does not say India is falling behind. It says something subtler: the global AI economy is settling into an infrastructure hierarchy. India is participating energetically in that economy. What remains unresolved is whether it will own enough of it to protect its external balance, its fiscal stability, and its strategic autonomy.

The writer is a theoretical physicist at the University of North Carolina at Chapel Hill, US, and the author of the forthcoming book The Last Equation Before Silence; views are personal