India's Largest IT Services Firm Bets ₹12,000 Crore on Domestic AI Data Centres

One of the country's biggest technology services companies unveiled a multi-year plan to build a network of artificial-intelligence data centres across three states, committing roughly ₹12,000 crore to capacity it says clients increasingly want hosted on Indian soil.

The push reflects a broader shift among enterprises that want to train and run AI models closer to home, both for latency and to satisfy tightening rules on where sensitive data can live. The company said the first facilities would come online next year, powered in part by long-term renewable energy contracts, and would be built to house the dense clusters of specialised chips that modern AI workloads demand.

Betting on sovereign AI demand

Executives framed the investment as a response to demand from banks, hospitals and government departments that are wary of sending data abroad. Many of these customers operate under regulations that require certain records to remain within national borders, and the arrival of a domestic data-protection regime has sharpened those concerns.

The plan pairs raw computing capacity with a software layer intended to let customers fine-tune open models on their own records without exposing that data to third parties. That, the company hopes, will let it move up the value chain from selling head-count-based services to selling outcomes — a transition the entire industry has been chasing for years with mixed success.

Industry watchers note that the economics of AI infrastructure are unlike anything the sector has built before. A single rack of AI accelerators can draw as much power as dozens of conventional servers, which is why access to cheap, reliable electricity — and increasingly to water for cooling — has become as important as access to the chips themselves.

A crowded, capital-hungry race

Rivals have announced similar ambitions, and the scramble for graphics processors and power-hungry real estate has become a defining feature of the sector. Global cloud providers are expanding their own regions in the country, telecom operators are converting exchanges into edge sites, and a clutch of specialist operators is racing to break ground.

Analysts cautioned that returns will depend on how quickly enterprise AI budgets translate into signed contracts rather than pilots. Much of the current spending is experimental, and there is a real risk of overbuilding if demand does not scale as fast as the capital being deployed. The winners, they said, will be those who can keep utilisation high and power costs low.

Reshaping the workforce

The firm also said it would retrain thousands of engineers for AI infrastructure and platform roles, part of an industry-wide effort to reshape a workforce built for an earlier era of outsourcing. For a sector that has long grown by adding people, the pivot to a more capital-intensive, automation-heavy model marks a significant strategic bet — one whose payoff will not be clear for several years.