If you follow power deals at all, you already know the shape of these. A tech company buys a few hundred megawatts of something clean, both sides put out a press release, and the developer has a good day on the market. Google did exactly that on Tuesday, signing up for 396 megawatts of geothermal power out of Beaver County, Utah.<\/p>\n
Fervo Energy’s shares jumped 28.41 percent and closed at $19.75. On August 12, three weeks earlier, they closed at $20.16. So the largest contract enhanced geothermal has ever signed did not get the company back to where it stood before its last earnings call.<\/p>\n
I’ll get to that call.<\/p>\n
Fervo filed the terms with the SEC, and the filing is far more specific than the press release. A subsidiary called Cape Generating Station 6 LLC signed with Google Energy LLC on August 26. The power arrives in four successive 99-megawatt tranches starting in the third quarter of 2028, and the delivery term runs 15 years. Fervo guaranteed its subsidiary’s obligations and Alphabet guaranteed Google’s.<\/p>\n
Then there’s the option. Fervo has agreed to offer Google roughly 600 megawatts more at Cape Station, taking the total to no less than about 950 megawatts, operating by June 2030 at the latest. Google does not have to take it.<\/p>\n
One clause in there is genuinely odd, and I read it twice to be sure. If Fervo never makes that expansion offer, it owes Google a refund on what the 8-K<\/a> calls “deemed delivered energy,” meaning power Google was billed for but never physically got. Fervo says it cannot currently estimate that payment, because the size depends on how much deemed energy Google has already paid for and where the market price sits against the contract price when the bill comes due.<\/p>\n Neither side disclosed the price. That is normal for these deals and still annoying, because the price per megawatt-hour is the one number that tells you whether geothermal is winning on merit or on a climate budget.<\/p>\n Cape Station sits in Beaver County, about 12 miles northeast of Milford, a town of 1,431 people at the 2020 census. Fervo is building it in stages toward 500 megawatts, with permits to go considerably bigger on the same ground.<\/p>\n Enhanced geothermal is simpler than the name suggests, and if you have read anything about fracking you already know most of it. A conventional plant needs a natural reservoir, hot water already sitting in permeable rock, which is why nearly all US geothermal capacity ended up in California and Nevada. Fervo builds the reservoir instead of hunting for one. Crews drill about two miles straight down, kick the bit over and run another 7,500 feet sideways, then frack the rock the way a shale crew would. Water goes down one hole, comes back up another at more than 400 degrees, and a turbine on the surface handles the rest.<\/p>\n The rigs, the crews and the pumping spreads all come out of the oil patch. There is no oil under Cape Station.<\/p>\n The wells have gotten pretty quick. In July, Fervo drilled Sawtooth 7 to 19,448 feet of measured depth in rock at 460 degrees, and it took 21 days from spud to total depth. We covered that well when it landed<\/a>, because 21 days was exactly what the company’s previous fastest well took, and that one was shallower, cooler and had a shorter lateral. Drilling normally gets slower the deeper you go, so holding the clock flat while going deeper and hotter suggests a crew that has stopped experimenting and started repeating. A single Cape Station well had already cleared 10 megawatts on its own<\/a> back in June.<\/p>\n You are probably wondering about earthquakes, because everybody does. Fervo monitors Cape Station with fiber optic cable cemented in behind the well casing, publishes its full seismic catalog, and says three years of operations have produced no events that posed a safety or community risk. That is the company grading its own data. It did put the Phase I dataset in front of outside researchers at a Stanford workshop in July, and nine of them sent back written reviews.<\/p>\n Here’s where I’d push back on the “geothermal has arrived” framing you’re going to read everywhere this week. Phase I of Cape Station lands at about $7,000 per kilowatt of installed capacity and Phase II should come in at $5,500, chief operating officer Sarah Jewett told Canary Media<\/a>. Fervo’s long-term goal is $3,000. BloombergNEF put the average build cost of a combined-cycle gas plant at $2,157 per kilowatt in 2025.<\/p>\n Which means the $3,000 Fervo hopes to reach years from now would still sit about 39 percent above the $2,157 a gas plant costs today. Against nuclear the comparison flips. Fervo’s IPO filing cites a Department of Energy range of $6,000 to $10,000 per kilowatt for conventional and small modular reactors, and Phase I already undercuts the middle of that.<\/p>\n The industry is also smaller than the headlines make it sound. Geothermal plants in seven states produced about 0.4 percent of US utility-scale electricity in 2025, according to the EIA<\/a>. Fervo’s own pilot in Nevada, Project Red, has been running since 2023 at 3.5 megawatts, and it is still the largest operating enhanced geothermal system in the country. That is the state of the art you are hearing described as ready to power the AI boom.<\/p>\n Fervo reported its second quarter that morning. A $55.9 million net loss barely registered, and neither did the $2.1 billion of cash sitting on the balance sheet after May’s IPO. What moved the stock was 2027 revenue guidance of $60 million to $80 million, and management’s explanation that the shortfall comes from curtailment on the transmission lines running from Cape Station to its customers.<\/p>\n Curtailment is when the plant runs fine and the electricity cannot get out, so nobody pays for it. Shares fell 16.59 percent that day to $20.16, against a May IPO price of $27.00. Management said the cause is transmission operator work on the line serving Cape Station rather than a problem with the wells or the plant, and that it should be specific to 2027.<\/p>\n Nineteen days later, Fervo laid out the situation in more detail on its own site. The Utah delivery point that anchored the approved transmission path into California is being retired in 2027, and capacity on the remaining paths into the state is limited. Roughly 400 megawatts of Cape Station output is contracted for delivery into CAISO, the California grid operator. In the same August 31 post<\/a>, Fervo wrote that it was “actively exploring contracts with prospective data center customers” to meet Cape’s near-term timeline.<\/p>\n Google’s 396 megawatts showed up the next morning. Are those the same megawatts that lost their route to California? Fervo has not said, and I would not assume it, because the company’s contracted book went from 658 megawatts to roughly 1,050 with this deal instead of staying flat.<\/p>\n One more thing from that week. Shareholder firm Hagens Berman said on August 31 that it had opened an investigation into whether Fervo disclosed the curtailment risk adequately in its IPO documents. That is an investigation notice soliciting investors, not a lawsuit, and no court or regulator has found anything.<\/p>\n Lucia Tian, Google’s director of advanced energy technologies, told The Wall Street Journal that the company does not yet have clarity on how this power will serve a data center, but that it will be a foundational building block for a data center presence in Utah. That is a candid thing to say on announcement day. Both companies said final plans still depend on engineering feasibility, state and local approvals and commercial conditions.<\/p>\n Utah has a law that makes this easier than it would be almost anywhere else. SB 132, signed by Governor Spencer Cox in March 2025, lets electricity users above 100 megawatts buy power outside Rocky Mountain Power’s normal process, either by direct contract or through what the statute calls closed private generation, while requiring that existing homes and businesses not pick up the cost. Fervo says that pathway, subject to regulatory approval, would let it sell straight to a data center campus next door and skip the interconnection line, which it puts at one to two years instead of five to seven.<\/p>\n That is a route around the transmission problem, and Fervo is not pretending otherwise.<\/p>\n Before any of the Google megawatts matter, though, Fervo has to get GeoBlock 1 turning. The company said on August 12 that the first 33-megawatt unit should begin producing test power in the fourth quarter of 2026 and reach full production by the end of the year, with GeoBlocks 2 and 3 hitting initial power in early 2027. The first 99 megawatts owed to Google are not due until the third quarter of 2028.<\/p>\n","protected":false},"excerpt":{"rendered":" If you follow power deals at all, you already know the shape of these. A tech company buys a few … <\/p>\nSo what is Cape Station?<\/h2>\n
Geothermal is still expensive<\/h2>\n
So what happened on August 12?<\/h2>\n
Google has not picked a site<\/h2>\n