Cooling Design Drives Water Use

Data centers generate heat that must ultimately be rejected to the environment. The equipment inside the facility may be cooled by air, chilled water, direct-to-chip liquid cooling, or other approaches, but those internal cooling methods do not by themselves determine how much water the facility consumes.

The key distinction is the final heat-rejection system. Dry or air-cooled heat rejection transfers heat to ambient air without intentionally evaporating water for cooling. An internal closed liquid loop can therefore operate with little or no on-site water consumption for heat rejection when it is paired with dry coolers or other air-cooled equipment.

Evaporative systems work differently. Cooling towers remove heat by evaporating a portion of the circulating water. Makeup water is required to replace evaporation and blowdown. DOE describes cooling-tower water demand as being driven primarily by evaporation, with additional water discharged as blowdown to control dissolved minerals.

Hybrid systems can combine dry and evaporative heat rejection and change operating mode with weather or load. Direct liquid cooling at the servers can improve heat-transfer efficiency, but its water footprint still depends on what happens downstream – whether the captured heat is rejected through dry equipment, evaporative cooling, or a hybrid arrangement.

Berkeley Lab's review of data-center water use identifies cooling-system type and climate among the important determinants of water consumption and finds very large variation across workloads and facilities. That is why project-specific cooling design matters more than a generic label such as 'liquid cooled' or 'closed loop.'