Google signed a 396 megawatt power purchase agreement with Fervo Energy on September 1, the largest enhanced-geothermal deal on record. The contract carries an option to expand toward one gigawatt by June 2030. Its stock rose as much as 26% the day the news broke.
Data center operators are now contracting firm, around-the-clock power years before a facility is built, turning geothermal into a bankable baseload asset rather than a pilot program.
The next leg of the sector, superhot rock, is already being financed and will decide whether geothermal stays a niche or becomes a grid-scale answer to AI's power hunger.
Enhanced geothermal systems (EGS) drill into hot, dry rock that conventional geothermal cannot reach, circulate water through engineered fractures, and produce power around the clock instead of following the sun or wind. The whole technology class still generates just 0.4% of U.S. electricity. That share is now the point.
Firm clean power Google is buying from Cape Station
The Utah GeoCluster supplies Google around-the-clock, carbon-free electrons for a potential data center, with power expected online in 2028. · Fervo Energy, September 2026
The deal, and the signal behind the numbers
The 396 MW PPA for Cape Station was signed on September 1 and covers the project in Beaver County, Utah. The electricity is earmarked for what Google describes as a potential data center, with final plans still subject to engineering feasibility, state and local approvals, and commercial conditions. Google has bought the power, not the building.
The anchor offtake is roughly $7,000 per kilowatt for Cape Station's first 100 MW phase and $5,500 per kilowatt for the second phase, according to Fervo chief operating officer Sarah Jewett. The first phase is a modular stack of three 33 MW GeoBlocks, with the first block in commissioning and a commercial operation date targeted for the fourth quarter of 2026. The full site is planned at 500 MW, with more than $2 billion going into it.
Google's potential total geothermal offtake
The option to expand by roughly 600 MW, on top of the 396 MW base, plus a separate framework agreement covering up to 3 GW through 2033. · Fervo Energy, September 2026
This agreement reinforces that EGS is ready to power the next generation of computing infrastructure.— Tim Latimer, CEO and co-founder, Fervo Energy
Read the PPA against the company's capital story and it reads differently from a headline deal. Fervo went public on Nasdaq in May at $27 a share, raising about $2.2 billion in an upsized offering that valued the company near $10 billion. The stock popped on its first day of trading, then shed those gains and more over the summer before the announcement drove it up about 30% in a session, per TechCrunch.
This is an offtake deal for firm power, not a subsidy or a speculative land grab. A data center operator committing to buy baseload from an unproven technology class is the mechanism that turns EGS from a pilot into a financeable asset. Google was early to the pattern: it backed Fervo's Project Red, a 3.5 MW Nevada pilot that came online in 2023, then signed a 115 MW agreement through NV Energy in June 2024 under the Clean Transition Tariff framework.
Why Google buys power before it has a data center
Confirmation criteria: a formal Utah data center announcement, or the 600 MW option being exercised.
Unit economics decide whether this scales
The case for EGS rests on repeatability. Fervo builds standardized 50 MW GeoBlocks and groups them into GeoClusters, borrowing horizontal drilling and fiber-optic sensing from oil and gas. The cost trajectory is the tell: $7,000 per kilowatt in phase one, $5,500 per kilowatt in phase two. If that curve holds, geothermal undercuts the gas-turbine bridge deals that hyperscalers have signed in bulk this year, without the emissions.
The balance sheet already carries real weight. Fervo closed $421 million in non-recourse project financing in March, secured a multi-year tubular supply agreement with Vallourec, and locked turbine capacity for 750 MW of its pipeline. Its Cottonwood observation well in Utah reached 555°F at 11,200 feet, the hottest well in the company's history, a data point on reservoir quality rather than a promise.
The risks are equally concrete, and the market has been pricing them since May. First-of-a-kind execution and grid interconnection queues sit outside any developer's control. Without a clear buyer, a first-of-a-kind facility struggles to attract financing at all, as Canary Media put it. One geothermal play clearing this gauntlet does not make a sector.
Scale or nothing.
The superhot frontier is already being financed
The next asymmetry sits below the current wells. Superhot rock, reservoirs above 375°C at pressures above 22 megapascals, could deliver five to ten times the energy per well at a modeled cost of $20 to $35 per megawatt-hour, competitive with natural gas. The U.S. Department of Energy's Advanced Research Projects Agency-Energy (ARPA-E) opened the SUPERHOT program with $30 million in Small Business Innovation Research (SBIR) and Small Business Technology Transfer (STTR) funding for well construction and heat extraction research.
Companies are raising against that curve. Quaise closed a $134 million Series B in July to build its 250 MW Project Obsidian near Oregon's Newberry Volcano, with a hyperscaler already signed for the first 50 MW and millimeter-wave drilling now past 500 meters depth at its Texas test site. Sage Geosystems brought its first next-generation plant online, and Zanskar raised $115 million to move from AI-driven exploration into construction. The Department of Energy added $171.5 million for field-scale tests of these technologies.
Watch the drilling economics rather than the press releases. Conventional EGS proves the offtake model. Superhot decides whether that model scales beyond the few sites hot enough to make today's numbers work.
What happens to next-generation geothermal in the next three years?
Probability: 60% — three hyperscalers have already signed hundred-megawatt geothermal PPAs, and ARPA-E and DOE funding has turned the drilling bottleneck into a targeted engineering problem with clear owners.
✅ Arguments for
Confirmation criteria: Cape Station phase one hits its fourth-quarter 2026 commercial operation date, and Quaise completes its first flow test at Project Obsidian by the end of 2026.
❌ Arguments against
Disconfirmation criteria: Cape Station slips more than two quarters, or Google fails to exercise the 600 MW option by 2030.
Cape Station phase one first power, scheduled for the fourth quarter of 2026, and the commissioning of its first 33 MW GeoBlock.
Whether Google exercises the roughly 600 MW expansion option before June 2030.
Utah's SB132 direct-to-load pathway and how the interconnection queue absorbs new baseload capacity.
Quaise's first flow test at Project Obsidian and superhot drilling depth milestones through 2027.
Development scenarios
🟢 Optimistic scenario (30%)
Implications: EGS becomes a standard line in corporate firm-power procurement, and the capital stack deepens beyond the current pioneers.
🟡 Base-case scenario (50%)
Implications: geothermal compounds as a slow, financeable infrastructure asset while AI power demand keeps tightening the market.
🔴 Pessimistic scenario (20%)
Implications: the sector consolidates around conventional hydrothermal sites, and this PPA reads as a footnote rather than a turning point.
As we wrote in June, when Fervo priced its IPO and became the first pure-play enhanced-geothermal developer on a U.S. exchange, the question was never whether the rock was hot enough. It was whether the offtake would be there to pay for drilling it. Google's 396 MW contract is the first clean answer at scale, and the option language attached to it suggests the buyer intends to keep drilling.