AI campuses need firm megawatts on calendars the grid often cannot meet. Through early–mid 2026, enhanced geothermal systems (EGS) moved from niche clean-energy story to a named option on hyperscaler power frameworks—multi-gigawatt development agreements, first-power targets at utility-scale plants, and pilots that wire geothermal output toward GPU compute.
Unlike intermittent renewables alone, geothermal targets high capacity-factor baseload. Unlike SMRs, lead times can be shorter where drilling and offtake close—but geology, transmission, and first-of-a-kind execution risk remain real.
Why geothermal is on the AI agenda now
- Interconnection pain — Multi-year queue delays push buyers toward resources that can co-locate or contract firm clean power.
- 24/7 carbon-free demand — Training and inference loads that run continuously fit baseload profiles better than solar-only stacks.
- Commercial templates — Framework agreements that standardize offtake make follow-on deals faster than one-off PPAs.
What still gates scale
Reservoir performance varies by site. Transmission rights in the western U.S. and similar basins can cap how much contracted capacity actually delivers. Capital intensity sits in drilling; faster wells help economics, but they do not erase the need for turbines, interconnection, and offtaker credit.
Behind-the-meter pairings with compute campuses are attractive where the grid is congested—yet they concentrate project risk on a single industrial customer and a single subsurface asset.
What to watch next
- First commercial power dates at flagship EGS plants versus slipped schedules.
- Whether Google-scale framework models spread to other hyperscalers and industrials.
- Transmission and behind-the-meter workarounds when contracted megawatts exceed available wires.
SMR and battery stories already dominate AI energy headlines. Enhanced geothermal is the newer firm-power chapter: clean baseload that only counts when wells, wires, and compute campuses commission on the same calendar.