Driven largely by the increased computing needs of artificial intelligence (AI), data centers are running out of power to grow. In the U.S., data centers are on track to consume as much as 12 percent of total U.S. electricity by 2028. Thirty percent of a data center’s total power draw comes from cooling the infrastructure. A new study looks at a way to ease that strain: pairing data center cooling systems with an emerging geothermal technology called Cold Underground Thermal Energy Storage (Cold UTES).
Cold UTES stores cold water underground during off peak hours and draws it back up to meet cooling demand when it’s needed most. Unlike a conventional battery, it can hold that stored cooling capacity for weeks or even an entire season. To test the idea, Andrew Chien, the William Eckhardt Distinguished Service Professor of Computer Science at the University of Chicago, and his University of Chicago co-author Wedan Emmanuel Gnigba modeled Cold UTES equipped data centers in three major U.S. markets: Virginia, Arizona, and Texas—regions chosen for their differing climates and grid conditions. Across the three sites, Cold UTES increased available IT power capacity by 9.8 percent. The biggest gains were seen in Arizona’s hot, dry climate, where capacity rose by 11 percent.
“The key benefit of Cold UTES is that is enables data centers to make better use of their grid interconnection capacity, reducing pressure on the grid for power, and enabling AI companies to run more GPUs,” says Chien of the broader research effort behind the technology. “Given the more than $10 billion price tags on AI data centers, the 11 percent increase we have demonstrated could mean savings of billions of dollars. Cold UTES ability to create flexible cooling is a powerful new way to make and operate data centers.”
The technology also improved cooling efficiency and eased pressure on space constrained equipment like rooftop dry coolers, the study finds, extending how long a facility can keep scaling before it outgrows its physical footprint.
The research is part of a broader Cold UTES collaboration funded by the U.S. Department of Energy’s Geothermal Technologies Office, which also includes the National Laboratory of the Rockies, Lawrence Berkeley National Laboratory, and Princeton University.
“With data center operators and grid planners racing to keep pace with AI-driven power demand, the findings point to underground cold storage as one option to accelerate AI capacity at lower cost and reducing the need to build new data centers and new power plants,” Chien says.
In The News
