More computing doesn't have to mean more water

For an enterprise-owned data center, water use can affect operating cost and expansion decisions. A campus adding AI capacity may need to support denser computing while working within existing cooling infrastructure and local water constraints.

Liquid cooling can sound like a commitment to consume more water. But circulating coolant through servers and consuming water to reject heat are different things. A closed coolant loop carries heat away from the equipment. The outdoor cooling system determines whether removing that heat requires evaporation, dry cooling, or a combination.^{1, 2}

Warmer coolant can make dry heat rejection practical for more hours of the year, depending on outdoor conditions and equipment performance. That can reduce reliance on water-assisted cooling. Avoiding mechanical chilling can also save energy, but the water and energy results depend on the complete design.^{1, 3}

Schneider Electric’s white paper of 21 September 2026 illustrates that potential. In modeled 100 MW designs using adiabatic heat rejection, an optimized liquid-cooled AI architecture reduced annual on-site cooling-water use by 48% in Dallas (from about 382,000 to 197,000 cubic meters) and by 53% in Paris (from about 108,000 to 51,000), compared with an air-cooled AI design. These are vendor-modeled results, not measurements from operating enterprise facilities or savings forecasts for your site.⁴

The lesson is that cooling architecture, operating temperature and climate need to be evaluated together.

A warmer technology-cooling loop is not simply a higher setpoint for an existing chilled-water system. The technology loop supplies coolant to the IT equipment; a coolant distribution unit (CDU) may transfer its heat into a separate facility-water loop. Each side has its own temperature, flow and equipment requirements. ASHRAE recognizes a W45 liquid-cooling class with an upper temperature limit of 45°C (113°F), but the selected hardware’s approved conditions still govern the design.^{2, 5}

Before choosing an architecture, bring IT and facilities together around five decisions:

• Equipment requirements: what coolant temperatures, flow rates, fluid quality and warranty conditions does the selected platform require?
• Heat rejection: can the existing plant support the load, including remaining air-cooled equipment, or will dry, adiabatic or evaporative equipment need to be added?
• Local conditions: how do summer design temperatures, seasonal weather, water availability and utility charges affect the options?
• Reliability and operations: what happens during a cooling failure or maintenance event, and how will controls, leaks, coolant quality and recovery be managed?
• Whole-project economics: what are the capital cost, annual water and energy use, maintenance needs and expansion implications of each workable design?^{2, 3}

For an existing facility, a dedicated AI zone may provide a practical way to introduce a different cooling architecture while retaining air cooling elsewhere. Its water benefit will depend on how heat leaves the site.²

← All briefs