How Epic Systems stores data center heat underground for winter

Epic Systems’ campus in Verona, Wisconsin, stores heat from a data center underground during summer and draws on it in winter to warm buildings, according to an explanation published October 9 by geoscientist Moones Alamooti. The example addresses a basic problem for reusing data center heat: servers produce it year-round, while demand for building heat rises in colder months.

Alamooti’s article, published by The Conversation and republished by Phys.org, describes an existing system, rather than a newly completed project or a new performance test. Epic has developed its geothermal system over roughly two decades. The account offers a practical example of seasonal heat storage, while leaving its cost and performance at much larger data centers unresolved.

How Epic Systems stores summer heat for winter

Epic has drilled roughly 6,100 boreholes, 300 to 800 feet deep, that serve about 40 campus buildings, Alamooti writes. The system uses heat from a 3.5-megawatt data center. In summer, warmed water circulates through pipes underground and transfers heat to the surrounding rock. In winter, water moving through the pipes picks up some of that stored heat and brings it back to the surface.

The heat recovered this way is suited to building uses, including office heating, hot water for dining halls and melting snow from sidewalks, according to Alamooti. It is not hot enough to drive a turbine to generate electricity. The distinction matters because storing a server’s waste heat for later use is different from drilling for naturally occurring geothermal heat to produce power.

What the Wisconsin example shows about scale

A U.S. Department of Energy case study of Epic’s geothermal heat-pump system reports that its campus buildings use about 25% less energy than comparable buildings in the same climate. That is a comparison for those buildings, not a measure of savings available to data centers generally. Alamooti says Epic has not published the cost of its system, so the account does not establish what a similar installation would cost elsewhere.

The boreholes also need a workable balance between heat stored and heat removed. Alamooti explains that if the ground receives more heat each summer than the campus extracts each winter, the rock around the boreholes gradually warms. Eventually, that could reduce the field’s ability to accept heat at the temperature for which it was designed. As Epic expanded, the company drilled additional boreholes to increase the system’s capacity.

Space and proximity present further limits. Epic’s data center is rated at 3.5 megawatts; Alamooti contrasts it with AI campuses planned at 100 to 1,000 megawatts. He says no one has yet stored a data center’s heat underground at that scale. Epic’s borefield serves a campus, rather than demonstrating that every large data center could find nearby buildings and enough suitable ground to use its waste heat.

Other examples and unresolved questions

Alamooti points to a different form of underground storage in Bonn, Germany, where an aquifer system has served a complex containing offices, a hotel, a medical center and a data center since 2009. He says that system supplies 60% to 80% of the complex’s winter heat demand. Its storage medium and setting differ from Epic’s boreholes, so those figures are context rather than an expected result for the Wisconsin campus.

Alamooti also describes research into using idle oil and gas wells for thermal storage, including modeling work in California. He says most states lack a regulatory category for operating such wells as heat-storage sites, and that he and a colleague have proposed a streamlined permitting route. That proposal has not been reported as adopted policy in his account.

The question has wider relevance as data center electricity demand grows. Alamooti cites a Lawrence Berkeley National Laboratory estimate that data centers used about 4.7% of U.S. electricity in 2024, alongside a projection of nearly 12% by 2030. He also cites cooling as a substantial share of data center electricity use. Neither those figures nor Epic’s example establish that seasonal storage has cut local bills, water use or emissions. The published account provides no independently measured figures for those effects or for the system’s seasonal storage efficiency.