Tata Consultancy Services has announced a large new infrastructure project through its subsidiary HyperVault AI Data Center Limited: a planned artificial-intelligence data center campus in Hyderabad, India, with capacity of up to one gigawatt. The company said on September 5, 2026, that HyperVault had secured 264 acres for the campus in Telangana.
The size of the headline is striking, but the qualifiers matter. The proposed one-gigawatt capacity is an upper limit, not a facility already operating at that scale. The investment—described in reporting as up to ₹700 billion, or roughly $7.4 billion at the exchange rate used by Reuters—is expected to come from HyperVault and partners and to be deployed in phases. Actual construction and capacity additions will depend on customer demand, technology requirements, power availability and execution.
Key takeaways
- HyperVault, a TCS subsidiary, says it has secured 264 acres in Hyderabad for an AI-focused data center campus.
- The planned campus could reach up to 1GW; that is a long-range capacity ceiling, not current operational capacity.
- HyperVault and partners may invest up to ₹700 billion through phased development.
- TCS says the campus is intended for high-density AI training, inference and advanced-computing workloads.
- Delivery, power sourcing, cooling design, customer commitments and phase-by-phase dates remain the key questions.
What TCS and HyperVault announced
The primary announcement came from TCS. It said HyperVault had secured 264 acres in Hyderabad to develop a large-scale AI data center campus with capacity of up to 1GW. Independent reports from Reuters, Financial Express and Mint repeated the core details and identified the project as a major addition to India’s emerging AI infrastructure market.
The project is designed around the demands of modern accelerators rather than conventional enterprise hosting alone. AI training and inference clusters concentrate enormous computing loads in a relatively small footprint. That changes the design problem: operators need dense electrical distribution, resilient networking, high-throughput cooling and a supply chain capable of deploying large numbers of accelerators without long interruptions.
TCS also describes HyperVault as a platform that can evolve from AI-ready physical infrastructure toward compute-led services. That direction is strategically relevant because it would place the group closer to the infrastructure layer on which model developers, cloud providers and enterprises run AI workloads. It does not, however, prove that every planned service, tenant or capacity block has been contracted.
The confirmed numbers—and the words around them
Three figures anchor the announcement: 264 acres, up to 1GW of planned capacity and up to ₹700 billion of potential investment by HyperVault and partners. Each should be read with care.
“Secured 264 acres” indicates control of a development site, but it is not the same as a completed campus. “Up to 1GW” describes the potential end-state capacity. A gigawatt is 1,000 megawatts, yet data center projects normally energize capacity in smaller stages as substations, buildings, cooling systems and customer demand come online. Finally, “up to ₹700 billion” is a ceiling across a multi-party, phased program rather than a single day-one expenditure.
The practical takeaway is that readers should separate site control, announced ambition, funded phases, capacity under construction and live capacity. Those milestones can occur years apart.
Why a 1GW AI campus matters
If delivered at scale, the project would be significant because AI infrastructure is increasingly constrained by physical systems. Advanced chips receive most of the attention, but compute cannot be deployed without power, transformers, switchgear, cooling equipment, fiber routes, land, permits and skilled operators. A campus planned around one gigawatt signals an attempt to coordinate those dependencies rather than add isolated server rooms.
Hyderabad already has a large technology-services base, cloud demand and engineering talent. A major AI data center campus could strengthen the local ecosystem by attracting network providers, equipment suppliers, operations specialists and customers that need proximity to high-density compute. The extent of that effect will depend on procurement, hiring and tenant decisions that have not yet been fully disclosed.
The SCALE framework: how to evaluate the announcement
A useful way to assess a very large data center announcement is SCALE. It separates tangible progress from promotional ambition.
S — Site control
Has the developer identified and secured land, and is the site suitable for grid interconnection, water strategy, transport and fiber? HyperVault’s 264-acre disclosure is a concrete starting point, although detailed permitting and interconnection milestones still matter.
C — Capacity in stages
What is the first energized phase, and when could it enter service? The one-gigawatt figure should be treated as the potential full build. Investors, customers and policymakers should look for megawatts under construction and commissioned, not only the eventual ceiling.
A — Agreements and demand
Are there named anchor customers, contracted capacity or partnership commitments? The announcement describes target users and workloads, but public evidence of tenant commitments is limited. That is normal at an early stage, yet it remains a key execution indicator.
L — Load, cooling and resilience
AI racks can require far more power and cooling than traditional enterprise racks. Future technical disclosures should clarify power density, liquid-cooling architecture, redundancy, backup strategy and how the campus will handle maintenance without disrupting workloads.
E — Environmental and ecosystem delivery
TCS has referred to green-energy and water-neutral ambitions. Those are objectives that require measurable plans: contracted clean power, water-use effectiveness, heat-rejection design, reporting boundaries and local supplier development. The strongest evidence will be audited operating data once facilities are live.
Power, cooling and water are central—not side issues
A one-gigawatt campus is ultimately an energy-infrastructure project as much as a computing project. The full capacity would be comparable to the output of a large power station, although actual consumption would vary by utilization and efficiency. The developer will need staged grid connections, on-site electrical systems and credible resilience arrangements before each capacity block can operate.
Cooling is equally consequential. High-density accelerator systems increasingly use direct liquid cooling, but the exact architecture varies. The environmental impact depends on climate, water source, cooling method, power mix and how efficiency is measured. Until HyperVault publishes phase-specific engineering information, claims about efficiency should be treated as goals rather than verified operating performance.
What the project could mean for India’s AI ecosystem
The announcement fits a broader effort to expand domestic computing capacity. More locally available AI infrastructure can reduce latency for regional users, support workloads with data-location requirements and give enterprises another option beyond capacity located overseas. It may also create demand for specialized skills in electrical engineering, cooling, networking, security and data center operations.
The project could also deepen TCS’s relationship with customers moving from AI experiments to production. A services provider that can connect consulting, software, operations and infrastructure may offer a more integrated route to deployment. That is an analytical possibility, not a confirmed commercial outcome; customers will still compare cost, performance, portability, security and contractual terms.
What remains unknown
- The size, cost and target date of the first construction phase.
- The grid-connection timetable and sources of electricity for each phase.
- Which partners will provide capital, equipment, cloud platforms or accelerator systems.
- Whether any anchor tenants have signed capacity commitments.
- The planned power-usage and water-usage efficiency targets, measurement boundaries and reporting schedule.
- How much of the stated investment will be borne directly by HyperVault versus partners.
Those gaps do not invalidate the announcement. They define the evidence readers should seek as the project moves from land and planning into construction, commissioning and operation.
FAQ
Is the Hyderabad AI data center already operating at 1GW?
No. One gigawatt is the planned maximum capacity for a phased campus. The announcement did not say that 1GW is currently built or energized.
How much is HyperVault investing?
Reuters and Indian business publications reported potential investment of up to ₹700 billion by HyperVault and partners. “Up to” and “with partners” are important qualifications; the final amount and timing will depend on the buildout.
Where will the campus be located?
TCS said HyperVault secured 264 acres in Hyderabad, Telangana. Independent reporting places the project in the wider Hyderabad development area.
Why are AI data centers different from conventional facilities?
AI clusters often combine high-density accelerators, fast interconnects and intensive cooling. They can require more power per rack and more specialized thermal and electrical design than many conventional enterprise environments.
Does this guarantee jobs or economic benefits?
No. Large construction and operating projects can create direct and indirect work, but the number, duration and local share of jobs depend on procurement, automation, construction phases and final operations. Forecasts should not be treated as delivered results.
