TCS HyperVault's 1 GW AI Campus: What Buyers Should Verify

What TCS and HyperVault announced

TCS said the Telangana campus would be developed on 264 acres in phases, aligned with customer demand and technology requirements. The company described capacity of up to 1 GW and investment of up to ₹70,000 crore by HyperVault and its partners. Reuters converted that figure to about US$7.41 billion using the exchange rate at the time of its 5 September report.

The campus is intended for high-density GPU training and inference. TCS says the design will use advanced liquid cooling and resilient power and fibre infrastructure, while applying green-energy and water-neutral design principles. HyperVault's product material also describes direct-to-chip liquid cooling and a broader plan for more than 1 GW of data-centre capacity across India.

One gigawatt is a target, not a delivery date

The words up to and phased are commercially decisive. They distinguish an ultimate campus envelope from capacity that has been constructed, energised, commissioned and made available to customers. A buyer cannot run a workload on an announced total, and an investor cannot treat the full investment ceiling as money already committed or spent.

That distinction is normal for large infrastructure projects. Power connections, grid upgrades, equipment availability, customer contracts and commissioning all shape the build sequence. The useful question is therefore not whether 1 GW sounds large. It is what each phase will deliver, by when, under which operating conditions and with what evidence.

Demand a metered evidence pack

Procurement, finance and technology teams should request a consistent evidence pack for every phase:

• Capacity: commissioned IT load, energisation date, customer-ready rack count and the difference between gross campus power and usable compute capacity.
• Power: contracted supply, grid and renewable sources, redundancy, backup duration, curtailment exposure and measured availability.
• Cooling: supported rack density, direct-to-chip compatibility, heat-rejection performance and results under sustained production load.
• Water and sustainability: withdrawals, consumption, local water stress, accounting boundary, offsets or replenishment method, and the period over which any water-neutral claim is assessed.
• Service readiness: carrier diversity, latency, security controls, acceptance tests, operating responsibility and contractual remedies when a milestone is missed.

Turn design principles into contract milestones

TCS's announcement uses the careful phrase green-energy and water-neutral design principles. Buyers should preserve that distinction. A design intention is not the same as measured operating performance, and a campus-level claim may not describe a customer's specific hall, workload or time period.

Contracts can convert ambition into evidence. Link deposits and workload migration to land and permitting completion, power and fibre availability, cooling validation, independent commissioning, customer acceptance and ongoing reporting. Define the measurement boundary for energy and water claims, who verifies the figures, how often they are reported and what happens if performance changes.

What business leaders should do next

Treat the Hyderabad announcement as a signal that Indian AI infrastructure is scaling, not as a reason to skip diligence. Shortlist workloads by latency, data-location, power-density and portability needs. Ask suppliers for phase-specific evidence, test a representative workload before committing volume and retain an exit path if delivery or operating performance diverges from plan.

ELYMENT AI's analyses of Together AI's India AI factory, Europe's LUMI-AI build and NVIDIA's capacity cycle show why infrastructure strategy must connect supply announcements to usable workloads and commercial proof. ELYMENT AI can help operators turn those questions into a measurable procurement and acceptance framework.

Sources

• TCS: HyperVault to establish large-scale AI data-centre campus in Telangana (5 September 2026) - Primary announcement covering the 264-acre site, phased plan, up-to-1-GW capacity, investment ceiling, liquid cooling and stated design principles.
• HyperVault: AI-led data-centre infrastructure (Accessed 6 September 2026) - Official product information on direct-to-chip liquid cooling, high rack density, power, fibre and HyperVault's broader India capacity plan.
• Reuters: TCS unit plans Hyderabad AI data-centre campus (5 September 2026) - Independent reporting confirming the announcement, phased structure and contemporaneous US-dollar conversion of the investment ceiling.
• TNW: Why AI data-centre sustainability claims need measurable detail (5 September 2026) - Independent analysis distinguishing announced sustainability design principles from disclosed operating metrics and measurement boundaries.

Continue learning

• Together AI's India AI Factory: What 10,000 GPUs Mean for Business AI
• Europe's LUMI-AI Supercomputer: What Businesses Should Prepare For
• NVIDIA's Record AI Revenue: What Businesses Should Budget For

Related analysis

• JERA's Chiba AI Infrastructure: Buy Capacity as a System

  JERA, Dell and RHAELM plan a 400 MW AI site in Chiba for 2028. Learn how to test power, cooling, compute, financing and recovery as one system.

• Lambda’s Oklahoma AI Data Centre: Put Community Costs in the Contract

  Lambda plans an Oklahoma AI data centre under new ratepayer protections. Learn which power, water and community obligations buyers should verify.

• Stargate UAE Rethink: Why AI Capacity Needs Failure Domains

  The reported Stargate UAE redesign shows why AI data-centre strategy needs distributed failure domains, tested recovery and resilient supplier contracts.

Frequently asked questions