Most data center environmental impact discussions focus on water consumption and electricity draw. Neither is visible to the people living next to the facility. The heat is.
Researchers at Arizona State University spent four months, June 18 through October 25, 2025, measuring the thermal plume from four Phoenix-area data centers. The study, led by David Sailor, director of ASU's School of Geographical Sciences and Urban Planning, was funded by the U.S. Department of Energy. It is published in the Journal of Engineering for Sustainable Buildings and Cities. The co-authors are Soroush Samareh Abolhassani and Eli Martin.
The facilities studied ranged from a 36-megawatt single-building site in Mesa to a 169-megawatt colocation campus in Chandler. All were primarily air-cooled. The cooling systems discharged air heated 14 to 25°F above the surrounding ambient temperature. The measurement methodology compared upwind and downwind readings at increasing distances from the facility perimeter.
Average downwind temperature increase: 1.3 to 1.6°F warmer than upwind readings across measurement periods. Peak increase: up to 4°F. The thermal effect was detectable at one-third of a mile from the facility boundary, approximately five city blocks. The waste heat output from a single facility in the study was equivalent to the combined thermal output of 40,000 households.
Phoenix is already among the hottest major metros in the United States, running an urban heat island effect that has extended the number of extreme heat days by several weeks over the past two decades. Adding a measurable thermal plume that carries five blocks in a city where heat-related mortality is already a documented public health problem is a different kind of facility impact than a cooling tower's water draw. The water draw affects the utility bill and the reservoir level. The heat discharge affects the air temperature in someone's backyard.
All four facilities in the study were primarily air-cooled. This is the critical technical detail for operators reading this. Air cooling rejects heat to the local atmosphere directly and continuously. Every watt dissipated by CRAC units, cooling towers in economizer mode, and rooftop air handlers exits the facility as heated air discharged at some elevation and velocity into the surrounding environment. The facility does not absorb the heat. It transfers it to the neighborhood.
Liquid cooling changes the rejection architecture. A facility running direct-to-chip cold plates with a closed-loop CDU still needs to reject heat at the facility boundary, but the mechanism is different. Dry coolers and adiabatic coolers concentrate the rejection at a smaller number of points with defined discharge angles. More importantly, liquid cooling enables higher-temperature heat rejection, which allows the facility to transfer thermal load to district heating systems, industrial processes, or other beneficial reuse applications rather than exhausting it at ambient temperature into the local air mass.
The ASU study was funded by the Department of Energy and published in a peer-reviewed journal. That combination means it will be citable in permitting proceedings and environmental impact reviews. Regulatory staff and community groups that opposed data centers previously on water or electricity grounds now have quantified, DOE-funded thermal impact data to include in their filings.
U.S. data center capacity is projected to double by 2030. Hundreds of megawatts are already operating in Phoenix, Chandler, and Mesa. Thousands more are proposed. The thermal discharge question is not hypothetical in Phoenix. It is already present, and now it is measured. Operators proposing air-cooled facilities in dense urban markets should expect this study to appear in their next public hearing.