TCS and TPG Commit $7.4 Billion to 1 GW AI Campus in Hyderabad
HyperVault, a joint venture between TCS and TPG, has committed $7.4 billion to a 1 GW AI campus in Hyderabad. The project repositions India's largest IT company as a compute infrastructure provider for AI labs worldwide.
700 billion rupees, equivalent to $7.4 billion committed by a single joint venture for a single campus in Hyderabad, say more about the speed at which demand for AI infrastructure has escalated than any analyst projection. For Tata Consultancy Services, India’s largest technology services company, Saturday's announcement represents a bet that diverges from what has defined its history: no longer selling engineering hours, but leasing megawatts to artificial intelligence labs around the world.
What HyperVault Will Build
Established as a joint venture between TCS and American private equity firm TPG with an initial investment of $1 billion, HyperVault has acquired 107 hectares (264 acres) in the technology corridor of Hyderabad, in the state of Telangana. The campus will design capacity of up to 1 gigawatt, configured for high-density GPU deployments aimed at training, inference, and advanced computing, according to an official statement from TCS. For reference, the largest individual campuses of American hyperscalers rarely exceed 200 megawatts at a single address. Telangana will have five times that.
Execution will occur in phases aligned with customer demand. The design adopts water neutrality. TCS estimates the generation of several thousand direct and indirect jobs in the state.
Who Will Pay for This Compute
HyperVault names its audience without euphemisms: frontier AI companies and hyperscalers. OpenAI, Anthropic, Meta, and Google, all with expanding development centers in India, make up the most likely pool of anchor tenants, although TCS has not disclosed specific contracts in Saturday's announcement.
Hyderabad already hosts regional centers for Amazon, Google, Microsoft, and Meta. For an American AI lab looking to serve clients in the Asia-Pacific without duplicating infrastructure in the U.S., the city offers acceptable latency for inference, energy costs lower than those in Northern Virginia, and high-performance engineering with an attractive cost-quality ratio. For European and Japanese companies, the campus represents a geopolitical alternative: AI capacity outside American centers, outside the jurisdiction of China, and with more defensible data sovereignty in front of European regulators and the Japanese government.
Why TCS and TPG Separated the Risk
The joint venture structure distributes the risks of construction and tenant acquisition between two distinct capital profiles: TCS brings decades of relationships with global technology clients; TPG provides capital with a typical private equity fund divestment horizon of five to seven years. The commitment of 700 billion rupees comes from the entire HyperVault consortium, not just the two founding partners, indicating that other investors or tenants have already been incorporated into the structure.
The most direct comparison is not to AWS or Azure. It is with the large-scale pre-leasing models used by specialized data center developers like Digital Realty and Equinix: the anchor tenant signs a 10 to 15-year reserved capacity contract, which finances the construction and justifies the fund's capital commitment. If HyperVault replicates this model with one or two AI labs as anchors, the project becomes financially defensible without relying on organic leasing growth in the spot market.
In the United States, lawmakers are increasingly monitoring the concentration of AI infrastructure among three hyperscalers. A large-scale campus operated by a non-American company diversifies this geometry, at least for labs willing to work with latency of 200 to 240 milliseconds between Hyderabad and San Francisco.
The number that will validate or call into question the project is not in hectares or gigawatts. It is in long-term leasing contracts with well-known names. A 1 GW campus without anchor tenants is a liability of $7.4 billion, not an asset. The next statement that matters will not come from TCS or TPG: it will come from an AI company announcing where it will train its next models.