Tata Consultancy Services has secured 264 acres in Hyderabad for a HyperVault AI data-centre campus planned to reach up to 1GW. The site removes a location question and makes the project more credible as a build. The commercial question is now the first phase: its cost, the load contracted for Hyderabad and the schedule for utility power. TCS announced the site on September 5 alongside an investment ceiling of up to ₹70,000 crore for HyperVault and its partners.
The land is new; the 1GW ambition is not
TCS and TPG had already announced a HyperVault platform capable of more than 1GW. The November 2025 agreement envisaged up to ₹18,000 crore of partner equity, including as much as ₹8,820 crore from TPG, with debt financing alongside it. The Hyderabad announcement therefore adds a physical site to an existing programme; the 1GW demand and funding plan remain the prior story.
The capital figures describe different layers. The ₹70,000 crore figure is a maximum investment expectation for the Hyderabad campus. The ₹18,000 crore is a multi-year, platform-level partner-equity envelope. If fully drawn, the arithmetic TCS share would be ₹9,180 crore; that figure belongs to the platform, not the Hyderabad allocation. The March 9 exchange filing records the cash actually subscribed at closing: TPG paid ₹199.36 crore for a 49% fully diluted stake, and HyperVault ceased to be wholly owned by TCS. TCS received no direct consideration. Reuters reported that the venture also contemplated roughly $4.5 billion to $5 billion of debt.
The March 9 subscription records the cash paid at closing; later phases can add financing exposure. The public record leaves the Hyderabad-specific debt mix, shareholder loans, guarantees, recourse, overrun support and refinancing obligations unquantified. The campus ceiling is therefore a project-scale marker, while the funded TPG amount is an entry point in the capital base.
Phase one is the investable unit
TCS management previously gave a rough platform-level guide of about $1 billion for each 150MW, a five-to-seven-year path to 1GW and first revenue in 18 to 24 months. For capex planning, applying that ratio as a consistency check gives approximately $667 million for 100MW, $1 billion for 150MW, $2 billion for 300MW, $3.3 billion for 500MW and $6.7 billion for 1GW. The figures are a platform-level check against the funding discontinuity between a demand-matched first phase and the ultimate master plan. TCS’s October 2025 earnings call said occupants would bring their own compute and storage while HyperVault supplied the passive infrastructure.
The demand ladder is narrower than the 1GW headline. OpenAI said it would be HyperVault’s first customer, beginning with 100MW and retaining an option to scale to 1GW. TCS later described that initial 100MW as committed. On a same-scope comparison, 100MW is 10% of Hyderabad’s maximum capacity; neither company has assigned that load to this site. TCS’s April earnings call referred to customer commitments and typical anchor workloads of 100–200MW, but also described design alignment, security work, site due diligence and commercial structuring. The measure that will change the valuation is contracted Hyderabad load in phase one, with the platform option treated separately.
Power is the physical gate. The site release promises high-density liquid cooling, green energy and water-neutral design, while a campus earns data-centre revenue only after utility supply, grid interconnection, substations, transformers and switchgear, cooling and water systems, network connectivity, permits, critical equipment, construction and energization are in place. A Hyderabad commissioning date and site-specific power allocation remain unreported. Until those links are secured, the project has a site milestone but no demonstrated operating path.
Hyderabad’s 1GW maximum equals the original HyperVault platform ambition, leaving the cost-concentration question live. The disclosures leave open whether the campus is the whole programme, one location within a larger network or a change in the programme’s geography. The first interpretation concentrates capital, customers and execution in one site; the second leaves the platform’s total funding need larger than the Hyderabad number suggests.
TCS needs infrastructure returns, not just strategic adjacency
The bull case is credible: scarce AI capacity, TPG’s outside capital, large customer relationships and Tata ecosystem partners could support high utilization, while data-centre deployments may pull through integration and managed-services work. TCS said HyperVault partnerships helped enable two very large net-new IT-services deals, but it disclosed no value, margin or duration. Its more than $2.3 billion of annualized AI-services revenue is a services figure, not HyperVault data-centre revenue.
The bear case is equally concrete. A slow power connection or delayed equipment can push first revenue out while financing costs accumulate; an option-heavy demand book can leave expensive capacity underused; and concentrated customers, rapid hardware cycles or returns below TCS’s asset-light core business can weaken the rationale for absorbing infrastructure risk. Reuters noted investor concern about that overlap when the venture was announced.
Hyderabad makes HyperVault more credible as a project, but not yet more valuable as an investment. The decisive evidence will be phase-one contracted load at the site, secured power with a dated energization path and the TCS equity required to reach first revenue. Until those numbers appear, 1GW and ₹70,000 crore describe scale; they do not establish returns.