India is building data centres faster than it can count them. Savills India reports that 258 MW of new capacity came online in the first half of 2026, up 59% from 162 MW a year earlier, taking operational capacity to about 1.8 GW IT. MediaNama's 16 October coverage of NITI Aayog's new framework notes that MeitY keeps no official record of capacity. The framework is a response to that gap, and on balance a sensible one.
What NITI Aayog actually proposed
NITI Aayog's Frontier Tech Hub released the Data Centre Evaluation Framework on 14 October 2026, developed with the Ministry of Power and MeitY. All India Radio's report describes it as a "practical decision support tool" that states can adapt, and says it adds no new layer of regulation. It is guidance for state governments, not law.
As reported by MediaNama and Outlook, the framework has several main elements:
- A centralised audit mechanism, with MeitY as nodal ministry, to periodically check operators' compliance and commitments.
- State approval matrices and a time-bound digital single window. Each permission would list its authority, required documents, processing time, dependencies, fees and escalation route.
- Post-commissioning monitoring of energy and water efficiency. MediaNama reports this as measured PUE, WUE and CUE reporting. Outlook's account describes it only in general terms, and I could not open the framework PDF to confirm the metric names.
- Siting rules. States would designate data-centre-ready zones and flag high-risk areas, with water stress, hazard exposure, land-use conflicts or weak infrastructure as grounds for restriction.
- Proportionate community impact assessments for large campuses, covering water, traffic, emissions, noise, waste and employment.
- Utility-facing conditions. Outlook reports a recommended 20–30-year seasonal catchment water balance. Facilities above 20 MW would need demand-response readiness and telemetry links to State Load Dispatch Centres.
The strongest case for tighter oversight
The case for scrutiny is real. Data centres draw concentrated power and, depending on the cooling design, large volumes of water. Outlook cites one estimate of up to 2 million litres a day for a single AI data centre, and notes proposed AI campuses near Visakhapatnam totalling about 6,500 MW. If operators self-declare efficiency and water-sourcing claims and nobody checks them, those claims are marketing. Communities and grid operators then carry the risk of the gap between promise and reality. An auditor with the power to verify commitments addresses a genuine accountability problem.
Where the framework is strongest
The most valuable ideas are the least glamorous. A single digital window with a published approval matrix replaces the guesswork of dozens of unsequenced clearances with predictable timelines. Investors can price that predictability. It also makes delay visible: a matrix that names each authority and processing time shows who is slow.
Linking load approvals to infrastructure milestones is likewise proportionate. Outlook reports that extra power load would be approved only after the required infrastructure is commissioned. That ties the obligation to a physical fact and avoids a blanket moratorium.
Measured reporting after commissioning is also the right order of operations. A design-stage promise of a low PUE tells you little. A measured figure from a running facility tells you a lot. It also gives India the data to benchmark itself, which it currently lacks.
Where it needs discipline
Three risks deserve attention before states adopt it.
First, the central audit could become a second clearance layer. The framework says it creates no new regulation, yet a MeitY-run audit of operator "claims" has no defined statutory basis in what has been reported. If the audit's scope, standards and appeal route are vague, it will function as a de facto licence. Operators would then face a state matrix and a central audit that can each stall a project. If the audit goes ahead, it should be narrow and rules-based. It should check declared metrics against measured ones and publish its methodology, rather than assess strategic merit.
Second, red zones need transparent criteria. Excluding genuinely water-stressed or hazard-prone sites is defensible. But the framework leaves classification to states, and an opaque map can freeze investment on discretion rather than data. Red-zone status should rest on published catchment data, be reviewable, and allow conditional approval where an operator can show closed-loop or contracted recycled-water cooling. Outlook reports that recycled water would count only when contracted, connected and available, which is a good standard. The same evidentiary approach should apply to the zones themselves.
Third, the data foundation is thin. The framework asks operators for measurements while the government has no baseline of its own. Capacity estimates differ across sources: Savills reports 1.8 GW, while other brokers put the end-2026 total near 1.7 GW. Reporting rules should be paired with a public registry of capacity, load and water use, so that the state is held to the same transparency standard as operators.
A reasonable path for states
Because the document is advisory, states can adopt the useful parts first. The single window, approval matrix and post-commissioning reporting carry low regulatory risk and clear investor benefit. The audit, red-zone and community-assessment provisions should follow only with published criteria, proportionality thresholds (the 20 MW trigger for grid telemetry is a model of a threshold tied to a measurable fact) and sunset or review clauses.
The framework's instinct is right: India's AI and cloud ambitions depend on physical infrastructure, and that infrastructure should be sited, measured and held to what it promised. The test is whether states use it to make clearances faster and more predictable, or to add checkpoints. Tying every new obligation to a published metric is the best guard against the second outcome.