India is working on a dedicated data centre policy as it looks to expand its AI infrastructure while addressing concerns over power availability, water use and the cost of running large computing facilities.
Junior IT Minister Jitin Prasada announced the proposed policy on October 14 at an event organised by the Central Electricity Authority (CEA) in Delhi.
“India is a very fertile ground for hyperscale data centres,” Prasada said, pointing to the country’s potential to host large computing facilities.
However, he warned that data centres should not take up electricity needed by households and other businesses. The policy push comes amid opposition from residents in Mumbai and Visakhapatnam to data centre projects planned by Amazon and Google, with concerns centred on their potential impact on local water supplies and electricity availability.
The government expects investment in the sector to reach around $200 billion, but the scale of the opportunity also raises questions about electricity supply, water availability, land use and the benefits that local economies will receive.
India’s data centre capacity has grown from approximately 375 MW in 2020 to 1.57 GW by August. The sector is expected to see further expansion, with installed capacity forecast to reach nearly 8 GW by 2030. The CEA, under the Ministry of Power, projects that the sector’s electricity demand could rise to 17 GW by 2031–32.
What Industry Wants From the Policy
Sunil Gupta, Co-Founder, CEO and MD of Yotta Data Services, tells AIM that the policy should include measures to ensure dependable, scalable power; faster grid connectivity and project approvals; a predictable tax framework; and long-tenure financing suited to the investment cycle.
Similarly, AS Rajgopal, MD and CEO of NxtGen Cloud Technologies, says AI data centres should be treated as essential infrastructure, with assured 24×7 power and direct access to renewable energy at industrial tariffs.
He also calls for extending the proposed long-term tax holiday to Indian neocloud and GPU-as-a-service providers, rationalising GST and customs duties on GPUs, servers, networking and cooling equipment, and providing low-cost, long-tenure infrastructure financing.
Swastik Chakraborty, Vice President of Technology at Netweb, tells AIM that as GPU deployments expand, power availability and cooling requirements are becoming central to the economics of data centres.
He notes that India cannot assume that every location will have access to the hundreds of megawatts required for large AI clusters, let alone 500MW or 1GW at a single site.
Chakraborty advocates a federated data centre model, in which computing infrastructure is distributed across multiple locations and connected to operate as a cohesive system.
Some locations may support 100MW or 250MW, while others may be limited to 30MW, 40MW or 50MW. Rather than concentrating all the infrastructure in one location, operators could deploy multiple availability zones and coordinate them across regions.
Rahul Dhar, President of Global Data Centre Operations, Enterprise Colo Delivery, Programme Management and Energy R&D at CtrlS Datacentres, tells AIM that the government should expedite approvals for transmission lines and fibre connectivity to support the growing demand for data centre capacity.
“Power availability has never been a problem. The problem is the time to deliver, or the speed to deliver,” says Dhar.
Data centres currently account for around 1% of India’s total power consumption, according to Dhar, who expects the share to reach 3% by 2030. He says better coordination between operators and government departments would help ensure that power infrastructure is ready when new facilities come online.
He suggests reducing the power forecasting period from seven to eight years to three to four years, allowing utilities to respond more effectively to changing requirements.
Dhar also calls for faster right-of-way approvals to lay fibre networks and transmission lines, particularly when operators acquire land on city outskirts.
“If we have purchased a new land parcel on the outskirts of a city, getting the right of way to lay a new transmission line from a power distribution point of view takes a lot of time,” Dhar says. He recommends streamlining these approvals and offering incentives to providers undertaking data centre projects.
Renewable Energy Could Reduce Dependence on the Grid
Chakraborty also calls for renewable energy to supplement conventional electricity supplies for power-intensive data centres. He suggests that solar and wind power could meet around 30–50% of a facility’s power requirements. He also identifies hydrogen-powered generators as a potential alternative to diesel generators.
The objective is to reduce dependence on conventional electricity sources while improving the sustainability of data centre operations.
He argues that renewable energy should serve as a secondary source of electricity rather than relying solely on additional grid connections.
These considerations should form part of the planning for new data centre capacity, alongside the cost of running and cooling GPUs.
Similarly, Dhar says CtrlS is investing in renewable energy and water-recycling measures to reduce the environmental impact of its operations.
The company owns and operates a captive solar farm with nearly 100 MW of capacity in Nagpur, Maharashtra. The electricity generated is fed into the grid to offset power consumption at its Mumbai data centres. CtrlS is looking to replicate the approach in other locations, including Hyderabad.
The company also operates at a power usage effectiveness (PUE) of around 1.4, Dhar says.
PUE measures the total energy consumed by a data centre against the energy used by its IT equipment; a lower figure indicates greater energy efficiency.
At one of its Mumbai facilities, CtrlS uses treated sewage water in its chiller systems to reduce dependence on freshwater. The company also uses closed-loop chillers, which limit water losses from evaporation.
Government as Customer of Compute Capacity
Gupta suggests the government should become a major consumer of AI and aggregate demand through long-term GPU offtake contracts, allowing Indian providers to compete for limited global GPU capacity.
“The government should become a prime consumer of AI and a demand aggregator,” Gupta says. “Multi-year commitments make the business case more viable, allowing providers to secure financing, place equipment orders and build capacity against visible demand.”
He suggests that the government’s move towards capability-based empanelment, followed by five-year contracts and direct orders for GPU capacity, was a step in the right direction. Such measures, he says, would give providers greater clarity on requirements and help translate empanelment into actual deployments.
“Predictable demand is essential if providers are to invest ahead of need and make capacity available in India,” Gupta says. “We have built data centre, cloud and AI compute capacity in India and serve government customers at scale. The policy challenge is to make it commercially and financially feasible to offer that capacity to Indian startups, enterprises and public- sector users, while continuing to compete for global demand.”
Rajgopal says the government should help Indian companies finance sovereign GPU infrastructure through viability gap funding or subsidised interest rates.
“Sovereign AI compute should be treated as strategic national infrastructure. Indian neoclouds are building capacity that keeps critical data and intelligence within the country, and policy should ensure that capacity is both affordable to build and easy to consume.”
