.svg)
Two miles under a stretch of Utah desert better known for cattle than computing, a drill bit is chewing through rock. Google is betting nearly half a gigawatt of its future data center power on what it finds down there.
Traditional geothermal power only works where nature already put hot water close to the surface, mostly places like Iceland or pockets of the Pacific Ring of Fire. Next-generation geothermal doesn't wait for nature to cooperate. It drills the plumbing itself.
The technique borrows directly from oil and gas fracking, drilling horizontally into hot dry rock, creating pathways for water to flow through, and pumping it back up as steam or hot fluid to spin a turbine. Because it doesn't depend on a naturally occurring hot spring, the theory goes, it can work almost anywhere the rock underground is hot enough, turning geothermal from a geography-limited niche into something closer to a repeatable industrial process.
AI data centers don't run on a schedule the weather can interrupt. They need power every hour of every day, what the industry calls firm power, and solar and wind simply can't promise that on their own.
Heat flowing from thousands of feet underground doesn't disappear at sunset or go quiet when the wind stops. That's the specific appeal enhanced geothermal is selling to companies like Google, a source of round-the-clock, carbon-free power that behaves more like a nuclear plant than a solar farm, without nuclear's construction timelines or its own regulatory hurdles.
On August 26, 2026, Fervo Energy signed a 396 megawatt power purchase agreement with Google, the largest enhanced geothermal contract ever signed, with an option for Google to add roughly 600 more megawatts through June 2030.
The power will come from Fervo's Cape Station project in Beaver County, Utah, where the company is investing more than $2 billion on a site permitted for up to 2 gigawatts total. The contract itself runs between two subsidiaries, Cape Generating Station 6 LLC on Fervo's side and Google Energy LLC on Google's, backed by parent guarantees from both Fervo and Alphabet. Fervo's stock, which had gone public just months earlier, jumped 15 percent the morning the deal was announced.
Initial contract: 396 MW. Option to expand: roughly 600 MW more by June 2030, for a potential total near 1 GW. Cape Station investment: more than $2 billion. Site permit ceiling: 2 GW.
Fervo's own numbers put the first phase of Cape Station at around 7,000 dollars per kilowatt to build. The company says the second phase should come in near 5,500 dollars, with a long-term goal of 3,000 dollars.
Even at the improved price, some reporting puts Cape Station's power at roughly twice the cost of a new natural gas plant, though Fervo's own filings argue it will beat the cost of newer nuclear projects instead. Fervo's chief operating officer has called the Google deal validation that geothermal can meet growing power demand, and specifically framed it as proof of a repeatable commercial model rather than a one-off. Google isn't alone in the bet either. Meta has separately signed its own geothermal deals with two other developers, XGS Energy and Sage Geothermal Systems, suggesting hyperscalers see this as a genuine category, not just one company's experiment.
Before Cape Station, Fervo already had a 3.5 megawatt pilot called Project Red running in Nevada. Cape Station's first phase alone is more than twenty five times that size, with much more still to come.
Cape Station's build-out is happening in stages on purpose. A 33 megawatt unit was expected to start generating test power by the end of 2026, ahead of the full 100 megawatt first phase, with the 400 megawatt second phase that anchors the Google deal following in 2028. Fervo raised its own development target to 1.1 gigawatts by 2030 after its most recent quarterly results, a sign the company itself is betting the model scales faster than a single flagship project.
Project Red pilot: 3.5 MW, Nevada. Cape Station Phase 1: 33 MW test unit expected online late 2026, scaling to 100 MW. Phase 2: 400 MW, expected 2028. Company-wide development target: 1.1 GW by 2030.
Geothermal still supplies less than half a percent of America's electricity today. This deal is a bet that enhanced geothermal can grow from a niche technology into something utilities and tech companies plan their futures around, not proof that it already has.
Analysts tracking the deal have flagged real risks alongside the enthusiasm, first-of-a-kind execution problems, interconnection bottlenecks, the sheer capital intensity of drilling at this scale, and competition from other firm-power options entirely. That list matters more given what's happened to older-style geothermal projects elsewhere, where a single endangered species finding can freeze a fully permitted plant for years. Enhanced geothermal's drilling-anywhere pitch sidesteps some of that fragility, since it isn't tied to a single natural hot spring an ecosystem depends on, but it hasn't yet been tested at the scale Google is now paying for.