🗞️ Why in News On 5 September 2026, HyperVault, a subsidiary of Tata Consultancy Services, announced plans to invest up to ₹70,000 crore with its partners to develop a 1 gigawatt AI data centre campus on 264 acres in Bharat Future City, Hyderabad. Telangana Chief Minister A. Revanth Reddy described the project as historic, and it is projected to create about 7,000 direct and indirect jobs.
The Announcement
| Parameter | Detail |
|---|---|
| Developer | HyperVault, a TCS subsidiary, with partners |
| Investment | up to ₹70,000 crore |
| Capacity | 1 gigawatt (GW) |
| Land | 264 acres in Bharat Future City, Hyderabad |
| Employment | about 7,000 direct and indirect jobs |
| Cooling | liquid-cooled infrastructure, with green energy sourcing |
| Execution | phased, in line with customer demand and technology requirements |
TCS CEO and Managing Director K. Krithivasan framed the investment in categorical terms, saying AI is “not a technology disruption but a civilisational shift”, that “it will change everything we do today”, and that societies which make the transition will be at the forefront.
The Chief Minister used the announcement to make a continuity commitment to investors: “Political and ideological differences will not affect the investment policy.” He set a target of inaugurating the campus on 2 June 2028, Telangana Formation Day.
What 1 Gigawatt Actually Means
This is the number that deserves attention, and it is an electricity number, not a computing number.
A data centre’s capacity is quoted in IT load, the power its servers can draw. One gigawatt is the scale of a large thermal power station. For comparison, a conventional enterprise data centre operates in the range of tens of megawatts. A 1 GW campus is therefore not an incrementally larger facility; it is a power-sector project that happens to compute.
Why AI changed the arithmetic. Traditional server racks drew perhaps 5 to 10 kilowatts each. Racks packed with AI accelerators for training and inference draw many times that, and the density is still rising. The constraint on building AI infrastructure in India is therefore shifting from land and capital to grid connectivity, generation capacity and cooling.
Two Metrics to Know
| Metric | Meaning | Why it matters |
|---|---|---|
| PUE, Power Usage Effectiveness | Total facility energy divided by IT equipment energy | A PUE of 1.0 is the theoretical ideal; overhead is mostly cooling |
| WUE, Water Usage Effectiveness | Litres of water consumed per kilowatt-hour of IT energy | The metric that decides whether a data centre competes with a city for water |
The Cooling Question, and Why It Was in the News
The same day’s reporting carried a second data-centre story that frames the first. Amid concern over the water consumption of the Google data centre coming up in Visakhapatnam, Member of Parliament M. Sribharat said the facility would use air-cooling rather than water, and that there was therefore “no question of impact to water supply”.
The HyperVault campus is reported to use liquid cooling.
These are not contradictory choices; they are different points on a real trade-off, and understanding it is the examinable content.
| Approach | Water | Energy | Suitability |
|---|---|---|---|
| Air cooling | Minimal | Higher, since chillers and fans work harder | Lower-density racks; water-stressed regions |
| Evaporative cooling | High, water is consumed by evaporation | Lower | Water-abundant regions; increasingly contested in India |
| Liquid or direct-to-chip cooling | Can operate in a closed loop with low consumptive use | Efficient at high density | Essentially required for dense AI accelerator racks |
The key nuance is that “liquid cooling” does not automatically mean high water consumption. A closed-loop liquid system circulates coolant and rejects heat through dry coolers, consuming little water; an open evaporative system consumes a great deal. The distinction between water withdrawn and water consumed is where public debate usually goes wrong.
For an aspirant the wider point is this: as AI capacity concentrates in a few Indian cities, data centres become a new claimant on two contested state resources, firm power and water. Telangana and Andhra Pradesh are competing for these investments, and both are in a region with real summer water stress. The governance question is whether siting, cooling and sourcing conditions are written into approvals in advance, or negotiated after a campus is built.
The Policy Backdrop
India’s data-centre build-out sits within a broader push on compute sovereignty. The IndiaAI Mission, approved by the Union Cabinet in March 2024 with an outlay of over ₹10,000 crore, has as one of its pillars the creation of a national GPU compute capacity accessible to startups, researchers and government. Several states, Telangana among them, run dedicated data centre policies offering land, power tariff and clearance incentives.
The strategic argument, stated fairly. Compute is becoming an input to national capability in the way that steel or refining capacity once was. A country that trains and serves its models on infrastructure located abroad exports both the value added and a measure of control over availability and data residency. That is the case for building domestic capacity. The counter-argument is equally serious: these are electricity-intensive assets whose output is largely commercial, they compete for grid capacity in a country still expanding per-capita consumption, and the employment they generate is modest relative to the capital deployed. About 7,000 direct and indirect jobs against up to ₹70,000 crore is a high capital cost per job by the standards of manufacturing. Both propositions are true, and a good answer holds them together.
UPSC Relevance
GS Paper 3. Infrastructure: energy and investment models; awareness in the field of IT and computers; conservation and environmental impact; effects of liberalisation on the economy.
GS Paper 2 angle. Centre-state dynamics in investment attraction, and the competitive federalism visible in states bidding for the same projects with land and tariff concessions.
The Mains framing. Treat this as infrastructure policy rather than technology news. The three questions worth writing about are: whether India’s grid can supply firm, round-the-clock power to gigawatt-scale loads without displacing other consumers; whether cooling and water conditions are being set at the approval stage; and whether the promised sovereignty benefit is realised when the campus is built by a domestic firm but the accelerators, the models and often the customers are foreign.
A question worth preparing. “Large AI data centre investments are being positioned as national digital infrastructure. Examine their implications for India’s energy and water security, and suggest a regulatory approach. (250 words)”
📌 Facts Corner — Knowledgepedia
Prelims, statement-ready facts:
- HyperVault, a subsidiary of Tata Consultancy Services, announced on 5 September 2026 an investment of up to ₹70,000 crore with partners in a 1 GW AI data centre campus in Hyderabad.
- The campus will occupy 264 acres in Bharat Future City, Telangana, and is projected to create about 7,000 direct and indirect jobs.
- The project will use liquid-cooled infrastructure and green energy, and will be executed in phases.
- K. Krithivasan is the CEO and Managing Director of TCS; he described AI as “not a technology disruption but a civilisational shift”.
- Telangana Chief Minister A. Revanth Reddy set a target of inaugurating the campus on 2 June 2028, Telangana Formation Day.
- Telangana Formation Day is observed on 2 June, marking the state’s creation in 2014.
- Power Usage Effectiveness (PUE) is total facility energy divided by IT equipment energy; Water Usage Effectiveness (WUE) measures litres of water per kilowatt-hour of IT energy.
- The Google data centre being built at Visakhapatnam is to use air-cooling rather than water cooling, according to MP M. Sribharat.
- The IndiaAI Mission was approved by the Union Cabinet in March 2024 with an outlay of over ₹10,000 crore, including a national GPU compute capacity pillar.
Prelims, the traps:
- A data centre’s “1 GW” is its electrical capacity, not a measure of computing power.
- Liquid cooling does not necessarily mean high water consumption. A closed-loop liquid system can consume very little water, whereas open evaporative cooling consumes a great deal. Water withdrawn and water consumed are different quantities.
- Bharat Future City is in Telangana; the Google data centre discussed alongside it is at Visakhapatnam in Andhra Pradesh. The two projects and states should not be conflated.
- HyperVault is a TCS subsidiary, not TCS itself, and the investment is stated as being made with partners.
Source: TCS Subsidiary HyperVault to Build a 1 GW AI Data Centre Campus in Hyderabad — Ujiyari.com | Free UPSC & State PCS Current Affairs
