- TCS subsidiary HyperVault will invest up to ₹70,000 crore (approximately $8.2 billion) with partners to build a 1GW AI data center campus across 264 acres in Hyderabad’s Bharat Future City, targeting inauguration by June 2, 2028.
- The facility will use direct-to-chip liquid cooling for high-density GPU racks, designed to handle frontier AI training and inference workloads at scale.
- TCS and TPG had already jointly committed up to ₹18,000 crore to HyperVault ahead of the Hyderabad announcement, with TPG’s own stake capped at ₹8,820 crore, giving the project a concrete financial anchor before a single rack is deployed.
On September 5, 2026, Tata Consultancy Services subsidiary HyperVault announced a 1GW AI data center campus spanning 264 acres in Hyderabad’s Bharat Future City, backed by up to ₹70,000 crore from HyperVault and its partners. TCS and TPG had already jointly committed up to ₹18,000 crore before the announcement was made, with TPG’s own contribution capped at ₹8,820 crore for a 27.5-49% equity stake. The campus is purpose-built for frontier AI companies and hyperscalers running dense GPU workloads that most facilities cannot cool.
Hyderabad’s Infrastructure Case
The Telangana government reportedly saw off competition from five other states to secure this project, and it has been aggressive on incentives. The competition is real. Microsoft already has a multi-billion dollar data center hub in the region, CtrlS is building a 612 MW park, and NTT Data, Neysa Networks and the Telangana government are developing a 400 MW AI-focused cluster. Google is putting $15 billion into an AI hub in nearby Visakhapatnam, Andhra Pradesh. HyperVault’s 1GW campus enters a market that is already crowded at the top end, mirroring the hyperscale buildout happening across the broader region.
Cooling Architecture
The engineering story is the cooling. Modern AI training clusters running large GPU arrays generate heat densities that air cooling cannot handle at scale. HyperVault’s campus will use direct-to-chip liquid cooling throughout, pulling heat away at the source rather than relying on raised-floor airflow. That lets you pack more compute into a given footprint and keep rack densities high enough for frontier model training.
Deepesh Nanda, CEO and MD of HyperVault AI Data Center Limited, described the design philosophy as building infrastructure for where AI is going: higher density, liquid cooling, larger power blocks and faster deployment. For customers running large language models or other compute-heavy workloads, that means access to dense, high-power infrastructure where the difference between days and weeks on a training job often comes down to cooling capacity. AI workload growth is pushing power and cooling requirements well beyond what most legacy facilities were designed to handle, which is the gap HyperVault is targeting.
Jobs, Capital and State Strategy
The campus is projected to create around 7,000 direct and indirect jobs across technical, construction, engineering and operations roles. TCS CEO K Krithivasan said HyperVault “advances TCS’ Infrastructure-to-Intelligence strategy,” connecting AI-ready infrastructure to the company’s existing cloud, engineering and enterprise transformation work. The joint TCS-TPG commitment of up to ₹18,000 crore, secured before the Hyderabad announcement, gives the project concrete financial backing rather than just headline ambition. Private capital is increasingly the anchor for facilities of this scale, where the gap between announcement and first power-on is measured in years.
Timeline and Green Energy
Telangana Chief Minister A. Revanth Reddy has set June 2, 2028 as the inauguration target, timed to coincide with Telangana Formation Day. The build will proceed in phases aligned with customer demand and technology requirements. HyperVault has committed to green energy sourcing and water-neutral design. At gigawatt scale, those commitments carry real weight: the grid and water demands of a facility this size will face scrutiny as construction progresses. How phased power procurement actually works in practice is not yet detailed in public disclosures.
