Schneider says hotter coolant can make AI datacenters less thirsty

AI data center water usage can be cut by more than half through higher-temperature liquid cooling.

Schneider Electric has published findings suggesting that advanced liquid-cooling systems can significantly reduce the water consumption of AI data centers. Their modeling indicates that operating data centers with liquid coolant supplied at higher temperatures, specifically 45°C (113°F) instead of 32°C (90°F), can cut water usage by at least 50 percent. This is attributed to the extended periods where external air can be used for cooling without mechanical chilling. The analysis compared four 100 MW data center designs, from traditional air-cooled to optimized liquid-cooled systems, considering weather patterns in Paris and Dallas. Schneider Electric has a commercial interest, as it holds a significant stake in liquid-cooling specialist Motivair and plans to acquire the remainder.

The study highlights that the choice of cooling architecture and operating temperatures are critical factors in water consumption, especially as AI demands increase rack density and power requirements. Schneider Electric's research suggests that liquid cooling, particularly when coolant is supplied at higher temperatures, offers a substantial reduction in water needed for heat rejection compared to traditional air-cooled methods. This finding is especially relevant given the growing global concern over data center water footprints and local community opposition to new facilities.

This research from Schneider Electric, a company with a vested interest in liquid cooling solutions, positions them to capitalize on the growing need for sustainable AI infrastructure. The findings will likely influence future data center designs, pushing operators to reconsider traditional cooling methods and explore liquid-cooling options to mitigate water scarcity issues and operational costs. The focus on evaluating water consumption early in the design process implies a shift towards more integrated and environmentally conscious data center planning.