AI data centers can use up to 10 million gallons of water a day
Why it matters: After voters backed limits on data centers, water use and cooling methods are becoming a central local permitting fight.
Large AI data centers planned across the US can consume enormous amounts of water, sharpening a backlash that helped ballot measures restricting such projects pass in 40 of 41 votes this week. Shaolei Ren, a University of California, Riverside researcher, said 1-gigawatt facilities built by companies such as Meta, Google, Amazon and SpaceX can use as much as 10 million gallons a day, though the total depends on hardware, location and cooling systems. Modern AI servers packed with GPUs often need liquid cooling because air alone is far less effective. Even when operators use closed-loop systems inside buildings, overall water demand can still hinge on how heat is expelled outside the facility.