World's Data Centres Are Guzzling Power, Water. Does India Have Solutions?

As artificial intelligence fuels a data centre explosion, electricity and water demand are surging. Can India bridge the widening gap between digital ambition and climate action, or is low-carbon nuclear power the only way out?

AI is power hungry and how the data centre boom is testing India's climate resolve, is it really a battle between compute vs climate in this new energy tug of war, one may witness soon in India.

The world is in the grip of artificial intelligence fever. From generative chat bots and image engines to automated logistics and predictive healthcare, AI is rapidly becoming the backbone of the modern economy. But beneath the excitement lies an uncomfortable truth: artificial intelligence runs on vast, invisible infrastructures that guzzle electricity and water at an unprecedented scale. At the heart of this infrastructure are data centres, warehouse-sized clusters of servers that must run 24/7, cooled relentlessly, powered continuously.

Nowhere is this tension more visible than in India, one of the fastest-growing data centre markets in the world. As the country races to become a global digital and AI hub, a fundamental question is emerging: can India pursue AI-led growth without derailing its climate commitments, can the 2070 Net Zero target set by India be ever met as it was set in the pre-AI era

Speaking to NDTV at the AI Impact Summit in New Delhi, Dr Arunabha Ghosh, and CEO of the Council on Energy, Environment and Water (CEEW), New Delhi framed the challenge bluntly. "There is no question there is going to be a lot of demand coming from data centres, both for energy as well as for water," he said. "In India alone, by 2030, what is today about half a gigawatt of electricity demand from data centres could rise to six and a half or seven gigawatts." 

That single statistic captures the scale of the disruption ahead.

Data Centres: The Hidden Engines Of AI

India's installed data centre capacity has already nearly tripled, from about 520 megawatts in 2020 to almost 1.5 gigawatts by mid-2025. Projections suggest capacity could reach between 4.5 and 6.5 gigawatts by 2030, driven largely by AI workloads, cloud computing, and data localisation requirements. 

Investments committed between 2019 and 2025 stand at roughly USD 95 billion, with more than $100 billion expected by 2027. Incidentally, India generates a lot of renewable energy but most of it is intermittent.

These facilities are not just power-hungry; they are water-intensive too. Cooling servers requires enormous quantities of water, especially in tropical climates. In 2024, India's data centres consumed an estimated 150 billion litres of water. Both electricity and water use are projected to more than double within this decade.

"Equally, there's a significant increase in the water demand to cool these data centres," Dr Ghosh warned. "The question before us is not an either-or."

The False Binary Of AI Vs Climate Action

That "either-or" framing, either to pursue AI or protect the climate, is precisely what Dr Ghosh rejects. "I would urge everybody to say this is a false binary," he told NDTV. "That either you do climate action or you do AI and data centres. We have to bring these two revolutions, the decarbonisation revolution and the digital revolution, closer together, like a double helix."

Yet the contradiction is hard to ignore. India has pledged to achieve net-zero emissions by 2070 and is pushing aggressively to expand renewable energy. At the same time, AI-driven data centres demand round-the-clock, reliable power, often in urban, water-stressed regions like Mumbai, Chennai, Hyderabad, and Bengaluru, which together host the bulk of India's data centre capacity.

Without careful planning, AI risks becoming counterproductive to climate goals, locking India into decades of higher emissions and resource stress.

Measuring What We Manage

One of the biggest gaps today, Dr Ghosh argues, is transparency. "Number one, we have to become much more transparent about the environmental, energy, and water footprint of artificial intelligence, and report it," he said.

Unlike household appliances or vehicles, data centres and Graphics Processing Units or GPUs operate largely outside public scrutiny. There are no mandatory disclosures on how much water a data centre consumes, how carbon-intensive its electricity is, or how efficiently its compute power is used.

Dr Ghosh proposes a radical but simple solution: star ratings. "We should have star ratings for energy and water for the data centre, for GPUs," he said. "Once we establish parameters on water use efficiency, power use efficiency, and carbon efficiency, innovation will follow."

Dr Ghosh suggests India's Bureau of Energy Efficiency (BEE) should immediately set standards for data centres. Such ratings, he argues, would reshape incentives across the entire technology stack, from chip design and rack architecture to cooling systems and end-use applications. "Today, the incentive structure is just not there," he said.

Sitting With Foresight

Where data centres are built may matter as much as how they are built. "Where we deploy our data centre capacity should be determined by foresight," Dr Ghosh said. "Can we co-locate it with renewable energy? If we need cooling, what kind of cooling technologies are we using?"

In water-stressed regions, he suggests, air-cooled systems, though more energy-intensive, may be unavoidable. In water-abundant regions, water-based cooling could reduce carbon footprints. These are trade-offs that demand granular, location-specific planning rather than blanket policies.

Yet India's policy framework remains fragmented. Fifteen states have notified data centre policies or included them within IT or industrial policies, but only five currently embed sustainability parameters. There is no binding national framework that aligns data centre growth with climate and water realities Dr Ghosh says.

The Compute Arms Race

Another problem, Dr Ghosh argues, is the global obsession with raw compute power. "Today, there's a competition for compute power, whoever has more compute power, that's considered a great thing," he said. "But the really great thing is: to what purpose did you use it?"

He calls for a shift away from brute-force scaling of large language models toward optimised, fit-for-purpose AI applications. "The goalpost cannot only be who has more compute power," he said. "The goalpost has to be who deployed the technology to have maximum impact for people, planet, and prosperity."

Without such a shift, AI risks becoming an arms race that rewards inefficiency, amplifying energy demand without commensurate social benefit.

Is Nuclear Energy The Only Answer?

As renewable energy struggles with intermittency, nuclear power is increasingly being discussed as a potential solution for powering AI and data centres.

"We at CEEW are technology-agnostic," he said. "We are seeing extensive improvement in renewable energy and battery storage. Solar plus storage is now under three rupees per kilowatt hour."

At the same time, CEEW's modelling suggests India may need over 200 gigawatts of nuclear capacity by 2070 to meet its net-zero target, far above today's sub-10 gigawatts. Nuclear, he argues, will likely play a role in providing low-carbon baseload power, especially as AI-driven demand grows.

But nuclear energy is not a silver bullet. "The question is how do we solve for the timescale for deployment? How do we get the social licence to operate?" Dr Ghosh asked. "Any technology comes with risks. We have to assess costs, unit economics, and the signals needed to bring down those costs."

In his view, the future lies in a carefully optimised mix: renewables plus storage for affordability and speed, nuclear for stability and scale, and AI itself to manage and optimise increasingly complex power systems.

Can Dichotomy Be Bridged?

So is there a paradox between AI deployment and climate mitigation? Dr Ghosh's answer is conditional optimism. "There is no paradox," he said, "if we have an intentional approach towards measuring what we manage, sitting with foresight, and innovating for applied use cases."

India's challenge is not whether to build data centres or embrace AI. Data sovereignty, economic competitiveness, and digital public infrastructure make that inevitable. The real question is whether India can steer this growth intelligently, embedding efficiency, transparency, and sustainability before today's decisions harden into tomorrow's constraints.

If it fails, AI could become one of the biggest obstacles to climate action. If it succeeds, AI and data centres could instead become tools to accelerate decarbonisation itself.

As India stands at the crossroads of digital ambition and climate responsibility, the choices it makes now will determine whether artificial intelligence becomes a climate liability, or a climate ally.