The real cost of the ’80s AI trend — Water crisis: The wider industry impact

AI trends are fun. The water crisis isn’t.

A 2026 report by the United Nations University Institute for Water, Environment and Health estimates that generating a standard-resolution AI image uses about 2.9 watt-hours of electricity, with an associated water footprint of roughly 29 millilitres, or about two tablespoons. The report estimates that producing 300 million AI images a day could result in an annual water footprint of more than 3 billion litres. Bengaluru is already a major data-centre hub, with 31 of Karnataka’s 32 operational data centres located in the city. Reportedly, a one-megawatt data centre requires 68,500 litres of water a day. Hydrologist Shashank Palur of WELL Labs estimates that Bengaluru’s existing data centres consume around 20 million litres of water a day, with a significant share coming from groundwater.

Ten more are planned, according to IT minister Priyank Kharge. Imagine a group of looters attacking your neighbourhood, and instead of stopping them, you encourage them and appreciate the way they are distracting you while they loot. Yeah, that’s exactly how absurd it is when we hop on meaningless trends for a few moments of joy and completely forget how artificial intelligence platforms and their data centres are consuming resources we desperately need.

Leave a Reply

Your email address will not be published. Required fields are marked *