Fervo Energy’s 396-megawatt power purchase agreement with Google, expandable to nearly 1 gigawatt by June 2030, is described as the world’s largest enhanced geothermal deal to date. It is being signed against a Cape Station project whose first and far smaller 100-megawatt phase has not yet delivered a single megawatt: the first of three GeoBlocks making up that phase is targeted for the fourth quarter of 2026, with the remaining two following in early 2027.
Fervo’s agreement with Google, signed August 26 and announced September 1, commits the company to supply 396 megawatts from an enhanced geothermal GeoCluster at Cape Station near Milford, Utah, by 2028, with an option for Google to add roughly 600 megawatts more by June 2030, bringing total contracted capacity to close to 1 gigawatt. That figure represents nearly all of the 1.1 gigawatt development target Fervo has set for the entire Cape Station site by 2030, a target the company raised alongside its second quarter 2026 results, meaning Google’s agreement accounts for most of what Fervo currently expects to build at the site over the same period, not a modest slice of a far larger independent buildout. The site itself is permitted for up to 2 gigawatts and has drawn more than 2 billion dollars in investment, according to Fervo’s own disclosures, giving the company room to expand further if the Google relationship and its own construction execution both continue on pace.
The construction pace behind that 2028 target is worth examining on its own terms. Cape Station’s first phase, roughly 100 megawatts built across three 33 megawatt GeoBlocks, has not yet come online. Fervo’s own guidance puts first power from the initial GeoBlock in the fourth quarter of 2026, with the other two following in early 2027, meaning the Google agreement’s 2028 delivery date for 396 megawatts sits well ahead of when even this much smaller first phase will have completed a full year of operation. Fervo’s disclosed capital costs also remain a work in progress rather than an achieved outcome: the company reported an all-in cost of 7,000 dollars per kilowatt for Phase 1, with a target of 5,500 dollars per kilowatt for the second 400 megawatt phase and a longer-term goal of 3,000 dollars per kilowatt, cost reductions the company is counting on delivering through its standardized GeoBlock approach rather than costs it has already demonstrated at scale. Those figures sit against the Department of Energy’s Enhanced Geothermal Shot target of cutting the levelized cost of enhanced geothermal to 45 dollars per megawatt-hour by 2035, a benchmark that assumes exactly the kind of capital cost decline Fervo’s roadmap projects but has not yet reached.
The political backdrop the source material describes as “bipartisan backlash” understates how sharply Utah’s tolerance for large data center infrastructure has shifted in the months surrounding this announcement. Utah Senate President Stuart Adams, who had served more than 20 years in the legislature and chaired the Military Installation Development Authority that fast-tracked approval for the Stratos data center project in Box Elder County, lost his Republican primary on June 23, 2026, the first sitting Utah Senate president to lose a primary in the state’s modern political history. Two Box Elder County commissioners who had voted to advance the same project, Lee Perry and Boyd Bingham, lost their primaries the same night. Polling tracked the shift in real time: a Deseret News-Hinckley Institute of Politics survey conducted May 15 to 18 found 53% of Utah voters opposed the project at its original scale, and a follow-up survey conducted June 16 to 22 found opposition had climbed to 60%, with Republican opposition specifically rising from 36% to 49% in the span of five weeks. Nearly seven in ten respondents in the May poll said the economic benefits of new data centers did not outweigh the costs to natural resources.
That framing, resource costs rather than opposition to data centers or artificial intelligence in the abstract, is what makes Logan Mitchell’s comment about geothermal potentially receiving “a different reception” a genuinely relevant distinction rather than a generic reassurance. Enhanced geothermal’s smaller land footprint and lack of on-site combustion emissions separate it clearly from the natural gas generators and self-contained power plants that have driven most of the recent opposition, and its round-the-clock output gives it an operating profile closer to nuclear than to wind or solar, a comparison Mitchell drew directly. What remains less established in available reporting is how enhanced geothermal’s own resource footprint, particularly the water required for the hydraulic stimulation process central to opening fractures in hot rock at depth, compares to the water and land concerns that fueled the Stratos backlash, a question Utah voters who have just demonstrated they will unseat a sitting Senate president over resource-intensive data center infrastructure are likely to ask of any large new power project sited in the state, geothermal included.
Google’s willingness to commit to Fervo at this scale before Cape Station has completed even its first phase reflects how compressed the alternatives have become. Gas turbine orders now take at least five years from order to delivery, and small modular reactors, the technology Governor Spencer Cox has promoted to meet the state’s projected demand growth, require roughly seven years to permit, site, and construct. Against those timelines, Fervo’s 2028 target for 396 megawatts is genuinely faster, and the company’s March 2026 nonbinding framework agreement with Google, which contemplated up to 3 gigawatts by the end of 2033, suggests both companies were already planning around Fervo’s construction pace before this binding agreement converted part of that framework into a firm commitment. But speed to power, the quality Mitchell called “a precious commodity” in the data center world, is only demonstrated once Cape Station’s GeoBlocks are actually delivering electricity on schedule, and as of this agreement’s signing, that demonstration remains pending for a project now being asked to scale to nearly ten times its currently under-construction first phase within the same general timeframe.
