Your '80s AI image could run your water tap dry

Your Instagram feed is a time machine. Big hair, aviator sunglasses, denim jackets, the grainy warmth of a Bollywood studio portrait from 1984. Your friends are posting them. Your favourite cricketers are posting them. Even Union ministers have joined the party, turning a harmless nostalgia trip into a full-blown political meme war.

All you do is upload a selfie, type a prompt, and the AI does the rest. From becoming Bachchan of Naseeb (1981) or Sridevi of Chandni (1989) to turning into Rishi Kapoor of Karz (1980), AI can do wonders. But wonders are asking for a price. And we are paying it without question or thought.

It looks like magic. It is not.

Every AI-generated image you create involves real computing power, which means real electricity and a water footprint. The United Nations University estimates that the electricity-associated water footprint of a single AI image is around 29 millilitres, roughly two tablespoons. That sounds negligible. Now multiply it by the millions of Indians who are generating not one, but dozens of versions, trying to get the perfect retro look. When Indian users generated over a billion AI images on ChatGPT in less than a month earlier this year, as claimed by Sam Altman, the arithmetic stopped being trivial.

Here is the uncomfortable part - AI infrastructure consumes freshwater primarily through evaporative cooling in data centres, indirect water use in electricity generation, and water-intensive semiconductor manufacturing. In the United States, about two-thirds of data centres built or in development since 2022 are located in areas experiencing high water stress.

More than half of India's data centres sit in areas already facing water stress, according to the World Resources Institute. A typical 100 MW data centre can consume around two million litres of water daily, although the actual figure varies depending on factors such as cooling technology, climate and facility design. That is equivalent to the daily water use of thousands of people. India's data-centre water consumption is estimated at about 150 billion litres a year and could reach roughly 358 billion litres by 2030.

A country already under pressure

Now place that against the country you actually live in. India is home to about 18 per cent of the world's people but holds roughly four per cent of its fresh water. Average annual water availability has fallen below 1,700 cubic metres per person. India crossed this threshold for water-stressed conditions in 2011. This June, India received about a third less rain than normal. Fourteen of the 24 states with major reservoirs were below their own ten-year averages. West Bengal's reservoirs were 60 per cent below normal. Karnataka's 40 per cent. Mumbai has made low-pressure rationing a recurring feature of its summers. Pune has moved to an alternate-day water supply.

This is not a distant, hypothetical crisis. It is happening now, in the same cities where data centres are being built and where water stress is already putting pressure on local supplies.

The problem is not AI. It is how we use it.

The problem is not that AI exists. The problem is how we are choosing to use it. The same technology that could help optimise irrigation, forecast droughts, or design water-efficient systems is being deployed to generate nostalgia portraits for a social media trend that will be forgotten by next week. The United Nations University warns of a ‘rebound effect’. As AI becomes cheaper and more efficient, increased usage can offset some of those efficiency gains, ultimately putting more pressure on total resource consumption. Efficiency without restraint is just a faster road to depletion.

UN has shown the way

The UN's Sustainable Development Goal 12 is unambiguous. It calls for responsible consumption and production, i.e., doing "more and better with less", recognising that sustainable development cannot ignore the environmental consequences of consumption. Conscious consumption is not a slogan. It is a survival strategy.

What we spend without seeing

Tomorrow's water is being spent on today's trends. And most of us cannot see it. The image appears on your screen in seconds. The water footprint behind it may be spread across electricity generation, data-centre cooling and other parts of the technology supply chain, far from the screen in front of you, and often in places you will never visit. The 80s are back. But the future is running dry.