Water quality considerations for sustainable data center cooling: treatment needs, reuse potential and environmental trade-offs

Data centers already use about 560 billion liters of water a year for cooling, and the build-out for artificial intelligence is expected to push that past 1,200 billion by 2030. The industry measures this with one number, liters per kilowatt-hour, which is blind to quality. A liter of treated wastewater, a liter of drinking water and a liter of chemical-laden blowdown going out the drain all count the same.

This review assesses water quality across the full range of data center cooling technologies. Four things stood out. Scaling, corrosion and microbial growth in the cooling tower set hard limits on how many times water can be recycled before it has to be dumped. Several of the largest operators have used reclaimed municipal wastewater for up to 17 years, and there is no peer-reviewed study of how it performs. The blowdown they discharge carries 1,200 to 6,000 milligrams per liter of dissolved solids plus biocides and scale inhibitors, and there is no federal effluent guideline for it. Every commercially available two-phase immersion cooling fluid is a per- and polyfluoroalkyl substance (PFAS), with the degradable ones breaking down to trifluoroacetic acid in the atmosphere. We propose a companion metric, Water Quality Effectiveness, that accounts for source, discharge and receiving water.

Data centers are being sited at speed in places that share their water with public utilities, and the metric everyone reports cannot tell a well-run facility from a harmful one. The review gives regulators and operators the chemistry, the treatment options and the gaps in one place. It also connects to a question I want to pursue at Chicago: whether this new upstream demand can be modelled at watershed scale and folded into drinking water risk frameworks that were never designed for it.