By Shayak Mazumder
There are two ways to read the Prime Minister’s appeal in Secunderabad on 10 May 2026. The first is the obvious one: gold, fuel, foreign travel, edible oil — discretionary dollar outflows the country can defer in a hard quarter. The second is more useful. The world has entered a generalised resource war, and a nation’s strategic position is now defined by what it cannot produce at home.
The war on Iran has pushed crude above $100 a barrel. India’s foreign exchange reserves have fallen from $728.5 billion before the war to $690.69 billion by 1 May 2026. The rupee is trading near 95. But energy is only the visible front.
A Stockholm International Peace Research Institute analysis this year describes critical minerals as having undergone “politicisation, securitisation, and militarisation” — linking them directly to conflict dynamics in Ukraine, the DRC, Greenland, and Venezuela. In February, the United States financed $1.3 billion of Pakistan’s Reko Diq copper project, in what The National Interest called a moment when “control over critical minerals is beginning to shape geopolitical influence as profoundly as energy once did.” The shortages of 2026 are no longer accidental. They are instrumented.
What India Can and Cannot Produce
India’s external accounts hold together for one reason. We earn dollars by exporting intelligence. In the first eight months of FY26, India ran a services trade surplus of $134.13 billion — up 15.28% year-on-year — cushioning a merchandise trade deficit of $223.14 billion. That services surplus is the load-bearing wall of India’s foreign exchange position.
Now place beside it a second number. Gartner forecasts India’s IT spending will reach $176.3 billion in 2026, with software spending alone growing 17.6% to $24.7 billion, driven by enterprises ramping up AI-enabled solutions. A large share of that software spend leaves the country in dollars — routed through SaaS contracts, foreign AI platforms, and per-token model usage.
India earns dollars by exporting intelligence. India is now spending an accelerating share of those dollars buying intelligence back, on subscription, with currency risk attached.
The Difference Between Buying and Renting
When India imports crude, the dollars leave once and the barrels arrive. When India imports gold, the dollars leave once and the metal sits in a vault. AI is the first major import category that bills you forever.
Every workflow on a foreign model is a workflow that pays in dollars for as long as it runs. The dependency is recursive: the more sophisticated the architecture, the harder the migration. You cannot stockpile API calls.
Layer in currency. A contract priced at $1 million last September costs noticeably more in rupees today, with no renegotiation and no exit. Multiply that across every Indian enterprise rebuilding workflows on foreign platforms, and you arrive at a structural commitment to an indefinite future of dollar-denominated operational expense.
What Other Countries Have Learned
The pattern is by now well documented. On 7 September 2010, a Chinese trawler collided with two Japanese coastguard vessels near the Senkaku Islands. The captain was arrested. Within days, China stopped rare earth exports to Japan. At the time, Japan was nearly 90% dependent — and prices rose roughly 10x in the year that followed, sending the automobile sector into panic. World Economic Forum
The dependency had been entirely rational on cost grounds for two decades.
It became strategically catastrophic in a week. Japan spent the next decade in expensive substitution. India has handled this category of problem before, and well. We did not import a payments network; we built UPI. We did not import identity; we built Aadhaar. The country has a digital public infrastructure other nations now study because, at a critical juncture, India chose to build rather than rent. The AI question is exactly that question, asked again.
The Decision in Front of Indian Enterprises
The Reserve Bank has begun to signal concern. In his February 2026 policy statement, Governor Sanjay Malhotra framed the year against “momentous actions on the geopolitical and trade-tariff fronts” and “heightened geopolitical tensions and elevated uncertainty.” In January, he was sharper on technology dependence: “Technology must embed compliance, not bypass it. Accountability must remain human.”
A financial institution running its most consequential reasoning on infrastructure it cannot inspect, modify, or guarantee continuity of, cannot fully discharge that accountability. The economics of AI are shifting fast. Frontier models are commoditising; token costs are collapsing year over year. Durable value will not sit in the model. It will sit in the layer above — orchestration, governance, agent design, enterprise memory — the connective tissue that makes AI work in a regulated business. That layer is best built close to the enterprises that will use it, in their language, on their data, under their regulators.
I have spent the last three years building one answer to that question. I am biased about how the answer should look. But the underlying observation is structural: in a year when the country is being asked to conserve foreign exchange across every category from gold to airline tickets, the fastest-growing dollar-billed import is the one no one has put on the list.
The PM’s appeal was made to consumers. The equivalent appeal to enterprises is being made quietly, in procurement reviews and architecture decisions, by CTOs defaulting to the path of least resistance — which is also, increasingly, the path of greatest currency exposure. The resource war of this decade is being fought across many fronts. The AI front is the one that arrives, monthly, on the invoice. It is also the one we still have time to choose.
– Shayak Mazumder is the founder and CEO of Adya AI.