The Need

AI data centers are being built faster than the grid can keep up. In some regions, electricity demands are projected to grow from about 3 GW to 100 GW within a few years. New generation and transmission take much longer to plan, permit and build. AI training loads also ramp steeply and are sensitive to voltage disturbances, which creates reliability risks. Cooling adds significant water and energy use. Conventional siting practices rarely account for these power, thermal and grid constraints together.

The Technology

OSU engineers have developed a "ground-to-grid" (G2G) framework that pairs AI data centers with subsurface energy resources on the same site, such as geothermal generation and underground thermal storage. Coordinated, workload-aware controls manage how power and heat move across the site, so the facility acts as a flexible grid asset instead of a passive load. A companion siting method helps developers find locations where subsurface, grid and community factors line up for faster, more resilient deployment.

Benefits/Advantages

• Lower water use and lower grid energy demand, by recovering and storing waste heat and cooling capacity
• Smoother load ramps and voltage support, which ease interconnection and help put off grid upgrades
• More viable sites, beyond regions limited by grid capacity or water supply
• Greater resilience, through microgrid islanding that reduces reliance on diesel backup and can support community services such as district heating