Every state wants the next giant data centre. Few have asked whether their grid can power it, whether their water can cool it, or whether the jobs and taxes will justify the concessions handed out to land it. NITI Aayog’s new report, Data Centre Evaluation Framework to Help States Plan for Sustainable Growth, puts those questions on the table.
The timing matters. Generative AI, GPU-heavy computing and cloud migration are pushing demand for large, power-dense facilities. Reporting around the release cited NITI Aayog member Abhay Karandikar’s estimate that data centres use about 0.5% of India’s electricity today and could reach 1 to 2% within a decade. Those are reported estimates, not location-specific forecasts. But the share hides the real pressure, because a handful of large campuses can strain a single substation or transmission corridor.
Power comes first
The report’s sharpest message is that proposed capacity is not deliverable capacity. It wants developers, DISCOMs and state transmission utilities to jointly validate a power readiness and ramp-up plan before a project is treated as real. That plan would cover total campus demand, sanctioned megawatts, commissioning dates, grid upgrades and who pays for them. Expansion should come in phases matched to the electrical infrastructure actually available. Renewable energy claims would count only when supply, connectivity and metering conditions are met.
Water could decide where data centres go
As AI raises rack densities, cooling becomes a siting issue. The framework prescribes no single technology. It asks for assessments that reflect local climate, peak summer demand and competing users such as households and farms. It calls for a 20 to 30 year seasonal water balance, and says allocations should not cut dependable supplies for existing users. Where water stress rules out more freshwater abstraction, it says desalination should be required where feasible. A cheap plot with good fibre may prove costly if the water isn’t there.
Incentives would have to be earned
States currently compete with land concessions, infrastructure support and tax benefits. NITI Aayog proposes measuring a project’s net fiscal and economic contribution after those benefits, and comparing like with like. A colocation facility, a cloud provider and a GPU-as-a-service business have different footprints in jobs, procurement and strategic value. Each incentive would carry a defined KPI, a measurement method, a reporting schedule and a suspension trigger. This is outcome-based industrial policy: support stays, but it has to buy something verifiable.
Ready-made zones and a single window
The report also urges states to publish data-centre-ready zones with verified information on land, power, water, fibre and expansion headroom of 10 to 15 years, and to mark high-risk zones where development should be limited. It backs a time-bound digital single-window approval system, but a portal alone won’t build a substation. That takes named officials, published timelines and tracking after approval.
The caveats
This is guidance, not a binding national rule. States decide whether and how to adopt it, and several questions remain open. Will states use common standards? How will “net economic contribution” be calculated? Who pays for grid expansion? What happens when a project misses its commitments? And how will transparency be balanced against security? These are questions whose answers are still not clear
The framework’s underlying argument is hard to dodge. The measure of success is no longer how many projects a state announces, but how many it can power, cool and operate sustainably, and whether they deliver lasting economic and technological value.