The Story
HyperVault, the data centre subsidiary of Tata Consultancy Services, has secured 264 acres in Hyderabad for an AI data centre campus of up to 1 gigawatt, disclosed in a regulatory filing. HyperVault and its partners are expected to invest up to ₹70,000 crore to build and operate it.
The campus is planned for the Bharat Future City area and will be built in phases, in line with customer demand and technology requirements. It is designed for frontier AI companies and hyperscalers, supporting high-density, liquid-cooled GPU deployments for training, inference and other advanced workloads. TCS says the build will use green energy and water-neutral design principles, and expects it to generate several thousand direct and indirect jobs along with demand across power, cooling, networking, construction and operations.
TCS chief executive K Krithivasan framed the campus as part of an infrastructure-to-intelligence strategy, pairing AI-ready infrastructure with the company's cloud, engineering and AI capabilities, and pointed to Hyderabad's scale, talent and ecosystem.
The agreement traces back to a conversation between Telangana chief minister A. Revanth Reddy and Tata Sons chairman N Chandrasekaran at Davos in January, with state IT minister D Sridhar Babu leading negotiations since. Reddy called the project historic for the state and for India, arguing that access to compute is becoming public infrastructure.
HyperVault's financing was arranged last year. In November 2025, TCS brought in the investment firm TPG, with the two committing up to ₹18,000 crore over the following years and debt forming part of the structure. TPG's contribution was put at up to ₹8,820 crore, roughly $1 billion. That arrangement covered HyperVault's wider data centre business rather than this campus specifically.
Why It Matters
The number that gives this its scale is not the rupee figure. It is the gigawatt.
India's entire installed data centre capacity sits in the region of one and a half gigawatts, accumulated across two decades in Mumbai, Chennai, Bengaluru, Hyderabad and Noida. HyperVault is proposing a single campus of up to one gigawatt. At full build-out, one site would be roughly comparable to everything the country has managed so far.
That is what makes the Hyderabad decision more than a choice of address. AI training is not latency-sensitive the way transaction processing is, so a training campus does not need to sit near financial centres or subsea cable landings. What it needs is land, power and water, in quantities Mumbai cannot supply at any sensible price. Telangana can, and it has moved quickly: a January conversation at Davos, land secured within months, which is not the usual pace of Indian industrial land acquisition.
For TCS the strategic logic is defensible. It already sells cloud migration, engineering and AI services to large enterprises. Owning the compute those services run on captures a layer it currently rents from Amazon, Microsoft and Google, and it gives the company an answer when a client asks where its models will actually run. Infrastructure-to-intelligence is a real position rather than a slogan, provided the infrastructure gets used.
The Strategic Read
Two things are worth separating before anyone treats ₹70,000 crore as spending.
The first is what has actually been committed. In November 2025, TCS and TPG together committed up to ₹18,000 crore to HyperVault's data centre business, with debt part of the financing and TPG's share around ₹8,820 crore. This announcement puts the Hyderabad figure at up to ₹70,000 crore from HyperVault and unnamed partners. That is roughly four times the committed capital, for a single campus, and it is phrased as an expectation. The build is explicitly phased against customer demand. The number therefore describes a full build-out that happens if the demand arrives, not capital anyone has allocated.
The second is who the customers are. The campus is aimed at frontier AI companies and hyperscalers. India does not have frontier AI labs operating at gigawatt scale, and the hyperscalers that do are building their own capacity globally. The realistic demand is narrower: data-residency rules pushing regulated Indian workloads onshore, the government's sovereign AI ambitions, TCS's own enterprise clients, and international customers drawn by Indian power and land costs against American ones. Each of those is real. None of them is currently a gigawatt.
Then there is power, which is the constraint that actually binds. A gigawatt of continuous draw is on the order of five percent of Telangana's peak demand. Securing that reliably, and doing it on green energy as promised, means building or contracting renewable generation at a scale comparable to the campus itself, plus firming for the hours when solar is not producing. That is a separate multi-year project with its own capital requirement, and nothing published so far explains how it sits inside the ₹70,000 crore headline.
The last question is what this does to TCS. IT services runs on high margins, negative working capital and almost no fixed assets. Data centres are the reverse: heavy capex, long payback, debt on the balance sheet. Bringing TPG in is precisely how a company keeps that off its own books, and the structure suggests TCS understands the trade it is making. Shareholders will still ask whether a services business should be building power-hungry real estate at all, and the answer rests entirely on whether the phasing is real.
For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.