The commitment
HyperVault, a subsidiary of Tata Consultancy Services, and its partners will invest up to 700bn rupees — about $7.41bn — in a one-gigawatt AI data centre campus in the Indian state of Telangana, the company said on 5 September. It has secured 264 acres in Hyderabad and will develop the site in phases.
The campus is aimed at AI companies and hyperscalers, supporting high-density GPU deployments for training and inference. At full build-out HyperVault expects it to rank among India’s largest AI infrastructure facilities.
Why Hyderabad
TCS chief executive K Krithivasan said Hyderabad offered the scale, talent and ecosystem needed to serve global customers. Telangana has been competitive on land assembly and clearances, and Hyderabad already hosts substantial capacity from Microsoft, Amazon and Indian operators, which means transmission and fibre routes exist rather than needing to be invented.
The number that matters is not the money
A gigawatt is the headline, and it is the part to hold lightly. Announced Indian data centre capacity has consistently run ahead of energised capacity, because power procurement, grid connection and water are slower than construction. TCS has not published a completion date for the full gigawatt, and describes phased development that follows customer demand and changing technology requirements — which is another way of saying the later phases are conditional.
That structure is now standard. Nscale, Volta and the hyperscalers all build in tranches against contracted demand. It protects the developer and it makes announced totals a poor guide to what will actually be running in 2028.
India’s position in the build-out
India has been a large consumer of AI services and a small host of the compute behind them, with most inference for Indian users served from Singapore or the United States. Data residency rules, latency and the cost of cross-border egress all argue for building locally, and the government has been offering incentives to encourage exactly this. A gigawatt in Telangana would be the largest single answer to that argument so far, and it would sit alongside Figure’s $3.5bn commitment to Nscale and Amazon’s tripled Nvidia order as evidence that capacity is being contracted years ahead of the demand it is meant to serve.
What to watch
The useful signals are the power purchase agreements and the grid connection date, neither of which has been disclosed. The second is the anchor tenant: a campus built for hyperscalers is worth what its first signed customer says it is. TCS is a services company moving into infrastructure, so the first named tenant will also indicate whether it is building for its own delivery business or as a landlord to the labs.
