Cognitive BriefSubscribe
Deep Tech · Data centres and energy · Analysis · 4 min

India's AI Race Just Became a Real Estate Story

TCS just committed ₹70,000 crore to one Hyderabad campus. The real bet: AI's bottleneck in India is land and power, not chips.

Cognitive Brief Team
India's AI Race Just Became a Real Estate Story

On Saturday, TCS subsidiary HyperVault announced plans to invest up to ₹70,000 crore, alongside its partners, on a single AI data centre campus in Hyderabad's Bharat Future City. The project will span 264 acres and eventually reach 1 gigawatt of capacity, built out in phases as customer demand and technology requirements evolve.

Read that again: one campus, one company, ₹70,000 crore. That is roughly $8 billion at current exchange rates, more than what most Indian startups raise in a decade, going into concrete, cooling systems, and power infrastructure rather than a model or an app.

This is the tell. India's AI story keeps getting told as a software story: sovereign LLMs, GPU allocations, foundation-model grants. The HyperVault announcement is a reminder that underneath all of that sits a much duller, much more capital-intensive layer, and right now that layer is where the real constraint lives.

The gap this is actually trying to close

India generates close to a fifth of the world's data but has historically held only a low single-digit percentage of global data centre capacity, according to TechCrunch's reporting when TCS first brought in TPG as a HyperVault investor last November. Real estate consultancy Colliers has projected the country's data centre capacity will roughly triple by 2030 from current levels. Those two numbers together describe a country generating far more digital exhaust than it has anywhere to process it.

That gap is what HyperVault exists to close, and Hyderabad is not its first move. TCS brought in TPG last November, with both partners committing up to $2 billion in equity and TPG's own contribution running to roughly ₹8,820 crore, plus a plan to raise several billion more in debt. That earlier commitment covered HyperVault's business broadly. This week's announcement is the first time a specific site, size, and rupee figure have been attached to an actual location.

Who's actually going to fill it

HyperVault already has an anchor tenant. OpenAI announced in February that it would become the platform's first customer, starting with 100 megawatts of capacity with room to scale toward the full 1 gigawatt. The deal sits inside OpenAI's broader "OpenAI for India" programme and its global Stargate initiative, and OpenAI has said the arrangement lets it run models inside India for lower latency and to meet data residency and compliance requirements, including government workloads.

Neither TCS nor OpenAI has said how much of the new Hyderabad campus is earmarked for that relationship specifically, or when the first phase goes live. But the sequencing tells its own story: sign the anchor customer, then build the specific campus sized to serve them.

"We are building infrastructure for where AI is going."

That's HyperVault CEO Deepesh Kiran Nanda, describing a facility designed for rack densities above 170 kilowatts, which he says will be competitive with anything built anywhere in the world. That density figure matters more than it sounds. Older data centres were built for a world of modest, evenly distributed compute loads. AI training and inference clusters draw power in dense, concentrated bursts, which is why liquid cooling and larger power blocks, not just more floor space, are the actual engineering problem HyperVault is solving for.

The state government's stake in this

Telangana Chief Minister A. Revanth Reddy met the TCS delegation, led by CEO K. Krithivasan, at the state secretariat on the day of the announcement and called the project "historic" for both Telangana and the country. The state's calculation is straightforward: HyperVault expects the campus to generate around 7,000 direct and indirect jobs across construction, engineering, power, cooling, and operations, on top of whatever downstream ecosystem forms around a facility this size.

This is also not Hyderabad's first AI data centre commitment. Telangana signed a separate agreement with Japan's NTT Data and Neysa Networks in 2025 for a 400 megawatt cluster in the same broad category. Multiple, large, competing bets by the same state government are themselves a signal: state administrations are treating AI compute capacity the way they used to treat manufacturing plants, as the infrastructure that decides which regions get the next decade's jobs.

What actually determines whether this works

The honest, falsifiable version of this story is narrower than the headline number. TCS has committed capital and secured land. It has not yet said when the first phase comes online or how much of it OpenAI will actually fill. Watch two things over the next twelve months: whether HyperVault or a competitor announces a second India campus of comparable scale, which would confirm this is the start of a real build-out rather than a one-off; and whether the Hyderabad site's first phase hits its own construction timeline, since data centre projects at this scale have a well-documented habit of slipping by years, not months.

If both hold, India's AI story genuinely becomes a hardware and infrastructure story as much as a model story. If they don't, ₹70,000 crore of announced investment will end up being a more useful headline than a useful campus, at least for a while.

In this storyTCSOpenAIHyperVault
Cognitive Brief Team

Keep reading

All Deep Tech →

The Cognitive Brief Digest

The stories that matter in AI, deep tech and capital, plus what moved in The Ledger. Free, in your inbox.