India is grappling with an enormous spectre of drought. Monsoon this year ended with a rainfall deficit of roughly 13%. Karnataka, Maharashtra and Andhra Pradesh officially declared drought, and Telangana, Rajasthan and Gujarat have since followed. For farming families, water is not just imperative for drinking, cooking and cleaning. It is the crop and the income. When it fails, the household fails with it.
Today, artificial intelligence is so integrated into our everyday lives that it is almost impossible to be alienated from it. It has transformed how we work, communicate and access information. Behind it sit AI data centres, hailed as a gigantic technological leap. They are massive digital warehouses of thousands of super-fast computers that store, process and distribute large volumes of digital information. They run day and night, and hot servers are cooled with local water. India’s data centres used an estimated 150 billion litres in 2024-25, and this is projected to more than double by 2030. A single 100 MW facility using evaporative cooling can consume around 800,000 litres a day.
The problem is where this thirst sits. More than 65% of India’s data centre facilities are in Mumbai, Chennai, Hyderabad, Bengaluru and Noida, and many of them are already water deficient. When one city is asked to supply both its people and machines, someone’s share shrinks. It is usually the one with less bargaining power. In Andhra Pradesh, residents objected to Google’s largest data centre project outside the United States, citing that “people cannot drink data.” The concerns become significant in a state that has declared drought. Google has since said the Visakhapatnam centre will use air cooling, as its own assessment showed pressure on the local watershed. It has also pledged to replenish 120% of the freshwater it uses across its operations by 2030. That is welcome. But it is a company’s promise and without mandatory reporting no one can verify it.
In fairness, data centres are not India’s biggest water users. Agriculture and thermal power also consume a substantial amount. So the argument is not against AI per se. It is that we cannot weigh a claim we cannot see. India does not have a national framework that mandates data centres to disclose their water use and most figures are estimates. Without knowing the amount and source of water, it becomes difficult to ascertain the pressure these facilities place on local resources. And without this information, it becomes equally difficult to determine if competing water needs are being justly accounted for.
This is the gap between policy intent and implementation. The Ministry of Jal Shakti says the industry is adopting advanced cooling. That is intent. Without disclosure, nobody can verify implementation. And efficiency alone does not settle who comes first. A more efficient data centre in a drought district still draws from the same shrinking source. The Economic Survey 2025-26 itself warned of groundwater stress from scaling large frontier models and suggested smaller, frugal models aligned with local priorities. The challenge is to balance AI infrastructure’s growth with the needs of local communities.
A fair answer needs rules before crisis, not after. First, mandatory public reporting of water use, including the source, by every data centre. Transparency is the starting point for gauging how these facilities affect local water resources. Second, data centres should switch to recycled water when a district declares drought to counter asymmetrical bargaining power. Third, clustering needs water-linked siting. Clearance should require a local water assessment with fresh water permitted only where supply is proven. Before approving new water drawing capacity, we should ask whether shared compute and smaller, task-specific models, both backed by the Survey, can do the job. None of this needs a new agency. State pollution control boards, which already grant consent to operate, could collect and publish the data. The Central Ground Water Authority could oversee groundwater use.
India stands at a crossroads amid the challenges posed by the rise of data centres. Whose thirst should be quenched: machines or people? The two can be reconciled only if India reckons before committing. Until then, the answer will go to whoever has the better pipeline. Progress that leaves farmers thirsty is hard to call a better humanity.
AI disclosure: The use of Generative AI has been limited to editing and research purposes.
Discussion (0)
No comments yet. Be the first to start the discussion.