Underground cold storage could cut a 1 GW Virginia data center's cooling energy bill by about two-thirds, national lab model finds

Key points

• A National Laboratory of the Rockies model cut annual cooling energy costs for a 1 GW data center from about $30 million to about $10 million in Virginia, and from about $22 million to about $9 million in Arizona.
• The modeled system stores cold water in deep brackish or saline formations, charged mostly in cooler seasons with dry coolers.
• Peak cooling load fell by about 200 MW at the Virginia site.
• The report says the technology is less proven for systems of 100 MW and up.

Storing cold in deep underground formations could cut annual cooling energy costs for a 1 GW data center in Loudoun County, Va., from about $30 million to about $10 million, according to a report from the National Laboratory of the Rockies. For a site in Maricopa County, Ariz., the modeled cost fell from about $22 million to about $9 million.

The lab first published the report in June 2026 and revised it in October to add two coauthors, correct terminology in one figure and add a sentence each to two sections. Josh McTigue, a senior researcher at the lab, led the technical work, and the lab worked with Lawrence Berkeley National Laboratory, the University of Chicago and Princeton University, the lab said in an Aug. 11 release. The report lists 22 authors.

How the modeled system works

The report focuses on reservoir thermal energy storage, an open-loop system that stores cold water in deep brackish or saline formations. The modeled data center draws 1 GW for IT and auxiliary load, uses direct-to-chip liquid cooling, and rejects heat through dry coolers with no evaporative cooling. The cooling supply temperature is 30 degrees C in Virginia and 40 degrees C in Arizona.

The storage is charged during off-peak periods, mostly in cooler seasons, using spare cooling capacity and dry coolers. During peak hours, the stored cold is released so chillers can be turned off, the lab said.

What the model found

The report's main results:

• Peak electrical load for cooling fell by about 200 MW at the Virginia site.
• Peak partial PUE fell from 1.22 to 1.07 in Virginia and from 1.17 to 1.07 in Arizona.
• The levelized cost of cooling fell 71% in Virginia and 66% in Arizona.
• Two separate methods estimated that the storage allows 9% to 20% more computing capacity.
• Avoided grid infrastructure and fuel costs were valued at about $90 million to $390 million for a 1 GW facility.

The results cover 12 case studies built on the one 1 GW reference facility at the two sites. The report says cold storage "can decrease the overall cooling infrastructure size and associated capital expenditures," but the sections reviewed give no capital cost or payback period.

The report says the system has "no water consumption when paired with dry-cooling technology." It also says dry cooling raises electrical load, which "can be as much as 30%–40% of the IT load" for direct-to-chip liquid cooling.

Arizona's climate limits charging. "There are limited hours to charge the UTES with dry coolers in hot environments, such as Arizona," the authors wrote. Charging with mechanical chillers instead "does not reduce peak loads as significantly," they wrote.

The report also notes that the technology is less proven at scale: "For larger systems (100–500 MW+), the TRL is lower due to scaling risks that have not yet been fully validated." TRL is technology readiness level.

Next phase

The Department of Energy's Office of Geothermal launched an Underground Thermal Energy Storage Collaboration that includes four national laboratories, and Guangdong Zhu, the lab's principal investigator for cold storage research, will direct the three-year project, the lab said. "The energy involved in large cooling systems is a pressing issue faced by developers of facilities like data centers," Zhu said.

---

Sources

• Project Phase 1 Report: Reducing Data Center Peak Cooling Demand and Energy Costs With Cold Underground Thermal Energy Storage (Cold UTES), NLR/TP-5700-100780, Josh McTigue, Guangdong Zhu and others, National Laboratory of the Rockies, June 2026, revised October 2026
• Computing Facilities Can Save Big and Keep Cool by Looking Underground, National Laboratory of the Rockies, Aug. 11, 2026