When Sam Altman visited India for the AI Impact Summit, he compared training AI models to raising a human being. “People talk about how much energy it takes to train an AI model, but it also takes a lot of energy to train a human,” the OpenAI chief said. “It takes about 20 years of life and all the food you consume during that time before you become smart.”
The remarks come at a moment when India is betting big on AI infrastructure. At the summit, held in Delhi between February 16-20, government officials reiterated plans to scale up domestic computing capacity under the IndiaAI Mission, including support for large data centre parks and high-performance GPU clusters to power generative AI systems. The pitch was for India to build the backbone to compete globally.
But the environmental costs of that backbone remained under-examined.
According to the International Energy Agency, data centres accounted for roughly 1.5% of global electricity consumption in 2024. That share is projected to grow by about 15% annually until 2030 — more than four times faster than overall electricity demand.
Electricity is only part of the story. Water is the other invisible input.
A small 1 megawatt (MW) data centre is estimated to consume about 26 million litres of water annually. At that rate, a 30 MW facility—roughly the current IT load capacity at the Yotta Data Centre Park, a hyperscale campus coming up in Greater Noida—would use nearly 780 million litres of water each year, or around 2.2 lakh litres every day. That single day’s consumption could meet the domestic water needs of more than 15,000 urban residents, based on the government norm of 135 litres per person per day.
And expansion is already underway. Once all six facilities at the Yotta park become operational, the projected IT load capacity is expected to rise to 160 MW, translating to an estimated 4.2 billion litres of water annually—comparable to the yearly consumption of a town of 85,000 people.
Yotta is not alone. Global operators such as Microsoft, Amazon, NTT Global Data Centers, ST Telemedia Global Data Centres, Iron Mountain and Equinix are expanding in India. Domestic players include CtrlS, Pi Datacenters, Web Werks, Nxtra Data, AdaniConnex and Sify Technologies. New collaborations involving Google and AdaniConnex, as well as interest from Tata Consultancy Services and Reliance Industries, signal further growth.
Yet public disclosures remain patchy. Companies rarely reveal how much water they draw, whether it is groundwater or municipal supply, or how operations affect already-stressed aquifers.
Some operators do publish ESG reports highlighting Water Usage Effectiveness (WUE), a metric meant to capture water efficiency. But WUE does not account for peak demand, seasonal variation, or local water stress—key factors in assessing real-world impact.
As India accelerates its AI ambitions, a core question remains: do existing environmental regulations adequately account for the scale, resource intensity and cumulative impact of the country’s coming data centre boom?
Here are edited excerpts from our interview with Advocate Yashasvi Rathore, who specialises in environmental and technology law and discusses whether India’s current legal framework is equipped to ensure environmental accountability amid the rapid expansion of the country’s data centre sector. Rathore is a legal consultant at Inductus Group, a law and policy advisory firm and an official member of NASSCOM, with expertise in technology, infrastructure, and regulatory compliance.
If a coal plant of comparable energy demand were proposed, would it face stricter scrutiny than a data centre? Why the difference?
A fairly unambiguous yes. The difference is where the governance crisis lives.
India has a reasonably well-developed environmental clearance architecture for the industrial era it was designed for. The EIA Notification, 2006, under the Environment Protection Act, 1986, is the primary instrument; it mandates prior environmental clearance for scheduled project categories before any construction or expansion can begin. The problem is that 'data centre' appears nowhere in that schedule. The notification was drafted when the digital economy was a fraction of its current scale.
A coal thermal power plant faces Category A scrutiny under the EIA Notification, national-level appraisal,mandatory public hearings, a full environmental impact study covering air emissions, ash management, water abstraction, and long-term operational footprint.
The only route through which a data centre routinely triggers environmental clearance is the building construction category, which kicks in above 20,000 square metres of built-up area. If it crosses 20,000 square metres of built-up area, it needs a building construction clearance, but that process asks about dust, drainage, and landscaping. It doesn't ask how much power this facility will draw from a coal-heavy grid for the next twenty years, how many millions of litres it will consume cooling its servers in a water-stressed city, or what its carbon trajectory looks like.
Are there any existing laws or guidelines in India that currently govern the environmental impact of data centres?
There are adjacent regulatory touchpoints. State Pollution Control Boards (SPCB) issue ‘Consent to Establish’ and ‘Consent to Operate’ under the water and air pollution control acts. But these are mere compliance frameworks, not upstream impact assessment frameworks. They ask 'does this facility meet discharge and emission standards?' not 'should this facility exist at this location at this scale?'
Back in 2023 MeitY's Green Data Centre guidelines set Power Usage Effectiveness benchmarks and nudged operators toward efficiency. But these benchmarks in advisory guidelines are just that, advisory guidelines, not enforceable law. They carry no penal consequence under the Environment Protection Act, they give no standing before the National Green Tribunal, and SPCBs have no mandate to enforce them.
What this means in practice is that a 300 MW hyperscale facility can be planned, approved, built, and operated without its full environmental footprint; energy, carbon, water, e-waste ever being formally assessed under any binding legal instrument. From a systems perspective, that's an extraordinary gap.
Where do you see the biggest gaps in legal coverage for data centre environmental concerns (energy, carbon emissions, water use)?
All three, and I'd add a fourth that cuts across all of them: disclosure and transparency. But let me work through each.
On energy: There are no binding rules requiring data centres to use renewable power or meet energy-efficiency standards. MeitY’s benchmarks apply only to public AI infrastructure, leaving the largely private sector unregulated. With coal still dominant in India’s power mix, large data centres effectively run on coal without any disclosure or obligation to change.
On carbon: India has no carbon emissions framework specific to data centres. The carbon trading scheme does not cover the sector, and existing ESG disclosures are inconsistent and lack facility-level detail, making meaningful oversight impossible.
On water: Water use is the most serious and least regulated gap. Data centres consume massive volumes of water, often in already water-stressed cities, yet laws regulate discharge, not consumption. There is also no mandatory reporting of water use.
On enforcement: Even existing standards suffer from weak enforcement. State Pollution Control Boards lack the capacity to monitor and regulate large, complex data centre operations at scale.
What risks does India face if it delays a national data centre policy that integrates environmental sustainability?
We need a national data centre policy urgently, and the cost of delay is already building up. India circulated a draft National Data Centre Policy in 2020, but it was never finalised. A revised draft released in 2025 is more ambitious and encourages renewable energy and efficiency, but it still treats sustainability as an incentive, not a requirement.
State-level policies largely focus on attracting investment, not enforcing environmental compliance. Even the more progressive states reward “green” facilities with extra incentives instead of making sustainability a basic condition for operating.
What’s missing is a national framework that ties environmental responsibility to approvals. Large data centres can currently be planned and built without a full assessment of their long-term impact on water, energy, or emissions.
If regulation is delayed further, the risks are clear. India is set to nearly triple its data centre capacity by 2030. These facilities are long-term assets, and the choices being made now will lock in environmental impacts for decades. Retrofitting sustainability later will be far more costly than building it in from the start.
Water makes this especially urgent. Most data centres are coming up in already water-stressed cities, yet there is no binding framework to govern their water use. Allowing a water-intensive sector to expand without clear rules is a political choice with serious consequences for urban water security.
India already has the building blocks—draft policies, AI mission benchmarks, and energy efficiency work. What’s missing is the will to turn guidelines into enforceable law.
