Modern cooling technologies for AI data centers, such as closed-loop liquid and immersion cooling, reduce on-site water evaporation. However, these systems increase electricity demand by 6–35% during peak operations compared to traditional evaporative cooling methods.
The Lawrence Berkeley National Laboratory estimated that US data centers directly consumed 17.5 billion gallons of water in 2023. Indirect water consumption, linked to the electricity production required to power these facilities, was approximately 12 times higher than direct usage.
AI infrastructure significantly increases power requirements, with server racks now exceeding 100 kW compared to 6–10 kW in older facilities. Projections suggest that electricity-related water use will account for 50–72% of total consumption by 2030.
Industry metrics like Water Usage Effectiveness (WUE) track on-site consumption but exclude water used at power plants. AWS reported a WUE of 0.12 liters per kWh, while Microsoft expects to reach 0.27 liters per kWh by fiscal 2025.
