dgavin, 11 days ago

@paninid @aral @gerrymcgovern About a million times more water is lost by leaky water pipes. A few data centres don’t really add up, if you look at the numbers…

aral, 11 days ago

@dgavin @paninid @gerrymcgovern How about we fix leaky pipes AND stop wasting water on AI? We can do two things at once. One doesn’t stop the other also being a problem.

dgavin, 11 days ago

@aral @paninid @gerrymcgovern Of course, but our energy would be better used first fixing the real problems and then going after the crumbles 🙂

aral, 11 days ago

@dgavin @paninid @gerrymcgovern That’s the thing, though, it’s not crumbs, it’s significant: https://theconversation.com/ais-excessive-water-consumption-threatens-to-drown-out-its-environmental-contributions-225854

gerrymcgovern, 11 days ago

@dgavin
In 2024, China Water Risk estimated that China's data centers would use about 1.3 trillion liters of water annually, which is sufficient for 26 million people's residential needs. The amount would surpass 3 trillion liters a year by 2030.

That's a lot of water, and data center water use is growing exponentially.
https://www.techtimes.com/articles/303669/20240417/ai-data-centers-strain-chinas-water-supply-claims-new-study.htm

@paninid @aral

airwhale, 11 days ago

@gerrymcgovern @dgavin @paninid @aral

I just can’t accept the conventional wisdom that liquid cooling using water by default needs to contaminate that water.

Sure, use the water for cooling, but keep it in a closed and clean system so it can be recirculated or reused. This seems like a problem current engineering should be able to solve (supported by AI, naturally.)