First power proves the drill; the 900MW is still a claim

The first of Cape Station's three 33-megawatt GeoBlocks has pushed power onto the grid in Beaver County, Utah, and Fervo Energy describes the achievement at full size: in the announcement's framing, the first utility-scale enhanced geothermal project anywhere to reach first power. What the company has banked is narrower. Fervo says it will keep ramping the first GeoBlock as it brings the full well system online, so the milestone is a commissioning event, and the block's rated 33MW is still being worked up; measured against the roughly 900MW Fervo claims at full build, the delivered share is a sliver of the nameplate.

The construction calendar behind the headline is easier to underwrite than the headline: Phase I is designed for 100MW across three 33MW GeoBlocks, with the two that have not yet produced expected to reach commercial operations in January 2027, and the 400MW stage after that is already under construction and expected to be completed in 2028. Every one of those dates is a claim about the late 2020s made by a company that has just delivered power from one machine in a field of three.

Google signed for the output before any of it existed. The long-term PPA covering 396MW from Cape Station was executed earlier this year, which means the anchor offtake for a project that reached first power this month was contracted against a resource still being proven at commercial scale. The announcement does not say whether the rest of the capacity is spoken for, and the arithmetic is unforgiving on that point: 396MW against a claimed 900MW puts 44 percent of the project's eventual output under a contract the announcement names, leaving the larger share of the megawatts outside the disclosed deal. Load contracted first, capacity following behind it, delivery date carrying the credit—that sequence is this buildout's shape in a single agreement.

Google's route to that agreement is a five-year ladder. It first backed Fervo in 2021 on a small-scale pilot in Nevada that came online in 2023, then signed a 115MW PPA with NV Energy as utility partner to move geothermal power into the Nevada grid; Google moved from customer to shareholder next, leading Fervo's Series B last year in a $462 million raise, and in May Fervo completed its IPO at a $10 billion valuation. On the day of that listing, the flagship project had no commercial power on the grid.

That is the part of this story that deserves more scrutiny than the drilling. Ten billion dollars of equity value rests on 900MW that is claimed but not yet built, with the first tranche sitting behind a January 2027 date nobody has tested. Set against the roughly 900MW claimed at Cape Station, the May valuation works out to about $11 million for each megawatt of that one project—arithmetic that only holds if the remaining checkpoints land on schedule. The capital hierarchy at work here is familiar: anchor-contracted digital assets get infrastructure pricing while everything else asks the equity market to underwrite the blank, and it is easiest to see at exactly this point in a company's life—contract signed, first electrons moving, delivery record still an intention.

Google's willingness to sign ahead of proof is not specific to geothermal. The company has funded a virtual power plant in PG&E's territory that coordinates hardware already sitting in homes, and it has been staking an option in Lea County before committing capital. What Cape Station adds to that portfolio is the one product those arrangements cannot manufacture on demand: firm baseload with a 396MW contract attached, which is the commodity an AI load forecast is short of.

The step in scale is worth naming: the Nevada contract was 115MW, and Cape Station claims roughly 900MW at full build, close to eight times the previous commitment and the reason the delivery schedule carries as much weight as the resource does. A developer that has proved a pilot and contracted a mid-sized project has proved something narrower than what Fervo's valuation now assumes, and the equity raised in May is likely doing its work on the 400MW stage already under construction.

First power retires one question and leaves the harder one standing. When analysts read the Fervo-Google agreement earlier this year, the doubt they recorded put the risk in the delivery chain, not the resource; a producing first GeoBlock is the strongest available answer to the resource question, and it does nothing at all for the schedule. That distinction carries more weight here than it would at a single-build plant, because Cape Station is being delivered in thirds: each GeoBlock is separately drilled, completed, and commissioned, each one its own encounter with the rock, and the financing of the 400MW stage likely gets decided on a record of three blocks rather than one.

Jack Norbeck, Fervo's chief technology officer and co-founder, put the achievement in engineering terms. "Cape Station works because we treated the subsurface like an engineering challenge," he said, crediting years of drilling, completion design, subsurface modeling, and flow testing, and calling first power the validation that matters most and proof that the science works. It is the right claim to make, and deliberately the smaller of the two on offer: proof that the science works is not proof that the third GeoBlock lands on time, or that completion techniques borrowed from the shale patch keep behaving at the deeper, hotter rock the company is targeting.

The mechanical case for enhanced geothermal is simple to state and tedious to execute: drill past the depths conventional geothermal reaches, use the fracking toolkit to open flow paths in the hot rock that traditional methods cannot reach, and convert previously stranded heat into firm capacity on a wire. Fervo is selling into a market with other sellers; Meta has signed offtake agreements with two enhanced geothermal developers in the US, including a 150MW deal with Sage Geosystems for a project under development east of the Rocky Mountains, and the load growth behind those contracts is the same AI demand curve that has hyperscalers optioning power years ahead of the steel.

The second-order effect belongs to every other EGS developer. A contract written against a resource class with no utility-scale precedent carries a technology discount, and first power at Cape Station is the precedent that shortens the diligence file for developers holding Meta agreements and Google options. Phasing a build into 33MW blocks has its own capital logic, the geothermal analogue of the modular data center: small enough to commission, small enough to finance, small enough to fail without taking the project down with it. One block has now moved from claim to output, which is the whole of what a reference case needs.

None of those agreements carries a price, and Fervo's is only partly the exception. It names the offtaker and the megawatt figure, more disclosure than a project milestone usually gets, but the terms stop there, and Meta's geothermal contracts are cut in the same shape. Milestone announcements that omit owners, offtake terms, or prices push merchant risk back onto developers and turn completion into a financing event rather than an infrastructure proof; Fervo's structure sits at the better end of that spectrum and still leaves the economics of 396MW invisible to anyone outside the two counterparties.

The grid side is where Cape Station is genuinely ahead. It has a route to interconnection at utility scale and a named offtaker committed to take 396MW, and grid permission is the underwriting asset in this cycle—the variable that sits ahead of generation technology in the order that decides which projects get energized. Fervo has that in hand. The megawatt-hours are the open item.

Two GeoBlocks and a 400MW stage stand between this announcement and 2028, and the delivery record built across them is what will decide whether May's $10 billion was a valuation or an option. The second GeoBlock's first power is the number that moves it.