This one started as a banking question, not a tech one. I keep reading about India’s massive AI data center buildout — gigawatt campuses, tens of billions in investment. My old instinct from commercial banking kicked in: when everyone announces enormous projects, the real story is usually in the supply chain and the permits, not the press releases. So I went through the market data. What I found confirmed the instinct.
The scale gap is real. India’s entire installed data center capacity is around 1.1 to 1.5 gigawatts. North America’s main markets alone are at 8.1. Mumbai holds over a third of India’s total. The country is roughly 17% of the mature-market supply — which means both a huge runway and a long distance to cover.
The technical shift is brutal, and it’s physical. Old server racks run at about 8–10 kilowatts. AI racks demand 30, 50, even 100+. That much heat means air cooling stops working — you need liquid cooling, piped directly to the chips. India’s supply chain for that equipment is still ramping up. The market for one key cooling component is projected to hit $3.6 billion by 2031, which tells you where the money is actually flowing.
The execution risk is the whole story. Developers need up to 42 separate clearances to open a facility. Power availability dictates where campuses can even be built. Water is a hard constraint — data centers use staggering amounts of it for cooling, and India’s water-stressed regions will force different designs from day one. In the first half of last year, the whole country absorbed 97.9 megawatts — solid growth, but a fraction of the multi-gigawatt announcements.
Here’s the banker framing I’ll offer: the higher-certainty opportunities aren’t the flashy campuses. They’re the picks and shovels — cooling systems, power distribution, networking gear — which make money no matter which tech giant’s campus wins. And any timeline you hear for a new Indian AI campus should be mentally discounted for permitting delays and grid reality.
Two things guarantee demand won’t be the problem: India’s central bank mandates that payment data stays in the country, and the government’s AI mission is subsidizing a 10,000-GPU public compute push. Demand is locked in. Pipes, power, water, and paperwork are what’s scarce.