AI data centers consume enormous amounts of water primarily for cooling. The GPUs and other chips used to train and run AI models generate intense heat, and the most efficient way to manage that heat at scale is evaporative cooling — essentially, running hot air or water through cooling towers where evaporation carries the heat away. Thousands of gallons of water can evaporate per hour in a large facility.

This is distinct from electricity consumption, which gets more attention. Water usage effectiveness (WUE) is the metric used to measure this, but many companies don't disclose it publicly. As AI workloads grow denser and more continuous, water consumption grows with them — often in regions already facing water stress.