India’s ambition to become a global hub for data centres and artificial intelligence makes economic sense. Businesses increasingly depend on cloud computing, and the infrastructure that stores and processes their data has become essential to economic activity. But attracting investment also means deciding who will pay for the electricity infrastructure it requires.
NITI Aayog’s Data Centre Evaluation Framework, released on October 14, brings that question into project appraisal. It asks states to assess infrastructure readiness and resource availability alongside investment proposals. India should use this opportunity to require developers to help expand the clean power supply their facilities need. Otherwise, incentives intended to build a technology industry could leave households and other businesses paying for its expansion.
READ | India’s data centre ambitions hinge on availability of water, power
Data centres require reliable electricity throughout the day, including when solar and wind generation is unavailable. Their servers generate heat, while cooling adds to electricity consumption and, depending on the technology, water demand. A relatively compact facility can therefore place substantial demands on the infrastructure around it.
Data centre electricity demand is rising
India’s data centre capacity increased from about 375 MW in 2020 to around 1,500 MW by 2025, according to the Ministry of Electronics and Information Technology. Further expansion will require power planning to begin well before facilities seek a grid connection.
NITI Aayog’s buildings-sector modelling projects 16 GW of installed IT load capacity by 2030. With its assumed utilisation of 40% and power usage effectiveness of 1.6, that translates into an average electricity load of roughly 10 GW, including supporting infrastructure. The same modelling puts electricity load at around 45 GW by 2050. These are projections dependent on assumptions about growth and efficiency, rather than committed demand. Installed IT capacity and actual electricity consumption should not be treated as interchangeable measures.
Much of this demand requires uninterrupted service. Financial transactions, digital payments and enterprise applications cannot simply stop when electricity becomes scarce. Some computing tasks, including certain AI training workloads, can be rescheduled or moved between locations, although the scope depends on the operator’s control over those workloads and the cost of delaying them. Planning should recognise this flexibility without assuming that every facility can offer it.
The geographical concentration of investment adds another constraint. Mumbai, Chennai, Hyderabad and Bengaluru are major data centre locations. Additional facilities in these hubs can strain particular substations and water systems even when national electricity supply appears adequate. A CEEW–Systemiq study cites estimates suggesting that data centres accounted for around 0.5% of India’s electricity consumption in 2024, with electricity and water requirements expected to more than double by 2030. The national share gives little indication of the pressure a cluster can exert locally.
READ | AI data centres: The new backbone of national competitiveness
Renewable contracts cannot guarantee hourly supply
India has expanded the generation capacity needed to meet new demand. By January 31, 2026, non-fossil sources accounted for 52.3% of installed electricity capacity, according to the Ministry of Power.
Installed capacity, however, does not establish how much electricity is available at a particular hour or location. Solar output falls after sunset, wind generation varies, and transmission constraints can prevent power from reaching consumers.
A developer’s solar power purchase agreement can support renewable investment and reduce emissions. It does not, by itself, provide uninterrupted clean electricity. Matching annual consumption with renewable purchases can conceal continued dependence on fossil generation during hours when contracted renewable output is insufficient.
India’s digital infrastructure strategy must therefore account for the cost of dependable supply. Renewable generation will need storage and adequate transmission, with other reliable low-carbon sources contributing where feasible. Batteries can shift daytime solar electricity into evening hours, but their contribution depends on storage duration and the demand they must serve.
Developers should contract for additional clean generation and contribute to the infrastructure required to deliver it. Disclosure should distinguish annual renewable procurement from the electricity actually supplying a facility over time. Such requirements would give regulators a better basis for assessing emissions and grid dependence than a headline renewable-energy percentage.
READ | Data centre future hinges on policy stability, not tax breaks
Incentives need an economic justification
The Union Budget 2026–27 proposed a tax holiday until 2047 for eligible foreign cloud service providers using Indian data centres for global operations. The proposal offers long-term tax certainty, subject to specified conditions; it is not a blanket exemption for every domestic data centre business.
There is a reasonable case for public support. Data centre investment can strengthen domestic computing infrastructure and support businesses that depend on digital services. Predictable approvals can also help investors commit capital to projects with long operating lives.
But the value of that support must be assessed against its cost. State concessions on land or electricity duties do not become economically justified merely because a project announces a large investment. Governments should establish what domestic economic benefits the facility will generate and how much public expenditure or forgone revenue is required to secure them.
Cost allocation also requires care. A substation built exclusively for a data centre is different from a transmission upgrade that serves a growing industrial district. Making ordinary consumers pay for dedicated infrastructure would subsidise the developer; charging the developer for every shared improvement could discourage investment that benefits other users.
Approvals should follow infrastructure readiness
NITI Aayog’s framework provides a useful basis for making these distinctions. It recommends jointly validated power-readiness plans involving developers and electricity utilities. Project-exclusive assets may be applicant-funded, while shared network costs should be allocated according to beneficiaries and use. Additional load should be approved after the necessary assets are commissioned and agreed milestones are met. These remain recommendations whose implementation will depend on state authorities and regulators.
Water appraisal deserves similar discipline. Describing every data centre as a water guzzler obscures differences in cooling design and local conditions. Approval should establish the proposed water source and the consequences for existing users, especially during shortages. Developers need clear requirements before construction so that resource constraints enter their investment calculations.
States courting data centres should make those conditions part of the initial offer. A project seeking public concessions should demonstrate how it will secure dependable electricity and meet the infrastructure costs attributable to it. If the investment is commercially attractive only when other consumers absorb those costs, the subsidy should be disclosed and justified before approval.