The largest deal in a still-young technology category
Google Energy LLC and Fervo Energy's Cape Generating Station 6 LLC signed a long-term power purchase agreement on August 26 covering 396 megawatts of enhanced geothermal capacity, with a contractual option to expand by roughly another 600 megawatts, bringing total potential capacity to nearly a gigawatt by June 2030. Industry trackers describe it as the largest enhanced geothermal system deal signed to date, a notable superlative for a power generation technology that has only recently moved from pilot projects to commercial scale.
Enhanced geothermal systems differ from traditional geothermal in that they do not require naturally occurring hot water reservoirs near the surface. Instead, they use techniques borrowed from oil and gas drilling, hydraulic fracturing among them, to create permeable rock formations deep underground that can circulate fluid and generate steady heat almost anywhere with sufficiently hot bedrock, dramatically expanding where geothermal power can realistically be sited.
Where the power comes from and where it goes
The capacity is sourced from Fervo's Cape Station EGS GeoCluster near Milford, in Beaver County, Utah, a facility the company has been developing since construction began in 2023. The power is intended to support Google's data center expansion in the region, though the precise location of the future low-carbon data center or centers that will draw on this capacity has not been disclosed publicly.
The deal structure rolls out in phases rather than arriving all at once. Phase I brings approximately 100 megawatts into operation by early 2027, with initial power expected to start flowing as early as the fourth quarter of 2026. Phase II adds 400 megawatts entering commercial operation in 2028, with the remaining expansion capacity available as an option through 2030 depending on demand.
Why baseload power is the actual selling point
The specific advantage enhanced geothermal offers over solar or wind is dispatchability: geothermal plants generate power continuously regardless of weather, time of day, or season, functioning as true baseload capacity in the same category as nuclear or natural gas rather than as intermittent renewable generation that requires storage or backup to guarantee uptime. For AI data centers running training and inference workloads around the clock, that continuity matters more than the marginal cost per kilowatt-hour.
This is the same underlying logic that has pushed hyperscalers toward nuclear power deals over the past two years. Geothermal offers a comparable reliability profile without the regulatory and public perception hurdles nuclear carries, and with a substantially shorter development timeline than building new reactor capacity, which is likely why Google is willing to sign a deal of this size with a technology still scaling its commercial track record.
The cost curve that makes this economically credible
Fervo's disclosed development costs show a steep declining trajectory: approximately 7,000 dollars per kilowatt for Phase I, dropping to roughly 5,500 dollars per kilowatt for Phase II, with a long-term target near 3,000 dollars per kilowatt as the company scales drilling techniques and standardizes its GeoCluster design across sites. That cost curve, if it holds, would put enhanced geothermal within striking distance of other firm power sources on a levelized cost basis within the next several years.
For Google, locking in a large capacity commitment now, while the technology is still on the steep part of its cost-reduction curve, functions similarly to an early-mover bet in any capital-intensive infrastructure category: the buyer accepts some execution risk on a newer technology in exchange for capacity commitments and pricing that may look increasingly favorable as the sector matures and other hyperscalers compete for the same limited pool of developed geothermal sites.
What this means for the broader power-for-AI race
Hyperscalers have spent the past two years signing an escalating series of power deals, spanning nuclear, natural gas, and now enhanced geothermal, all chasing the same underlying constraint: AI data center growth is increasingly gated by available firm power rather than by chip supply or capital availability alone. A deal of this scale for a technology still building its commercial track record signals how acute that power constraint has become.
For enterprises and utilities watching this space, the practical takeaway is that geothermal has moved from a niche renewable technology to a credible, competitively bid option for large-scale firm power procurement. Utilities and grid operators in geothermally favorable regions should expect more hyperscaler interest in similar deals, and enterprises evaluating their own data center siting decisions should factor geothermal availability into location strategy alongside the traditional considerations of fiber connectivity and tax incentives.
The execution risk still on the table
None of the phased timeline is guaranteed. Enhanced geothermal at this scale remains an emerging technology, and Fervo's ability to hit its Phase I and Phase II online dates depends on drilling execution, subsurface performance matching pre-drilling models, and continued progress on the cost curve the company has publicized. Delays or underperformance at Cape Station would be a meaningful setback for the broader case that enhanced geothermal can scale fast enough to matter for the current AI buildout.
Still, the size of this commitment from a buyer as sophisticated as Google, backed by a phased structure that lets both parties validate performance before the largest capacity tranches come online, suggests Google's own diligence found the execution risk acceptable relative to the alternative of competing for scarce natural gas turbines or waiting years longer for new nuclear capacity to clear regulatory review.
