A Geothermal Plant in Rural Utah Just Made History. Google's Near-Gigawatt Bet on It Is for a Data Center That Doesn't Exist Yet.

On September 24th, a wellhead buried in the high desert scrubland outside Milford, Utah synced to the regional power grid for the first time. Six days later, Fervo Energy declared the unit commercially operational — one day ahead of its own contractual deadline. By the company's account, it's the first time anywhere in the world that a next-generation, engineered geothermal plant has reached that stage: drilled, built, connected, and now selling power under contract. It produces 33 megawatts. That's a small number next to the headlines this project has generated. It is also, strictly speaking, not the part of the story Google is paying for.

The plant is called Cape Station, built on technology Fervo borrowed almost wholesale from the shale drilling industry: wells bored horizontally thousands of feet down into hot, dry rock, with engineered fractures that let water circulate through the heat and return to the surface as steam, in places where the rock never had the natural plumbing old-style geothermal needs. The GeoBlock that just went live is one of three making up Cape Station's first phase — a combined roughly 100 megawatts. The other two are still being commissioned and are expected to reach commercial operation by January 1st, 2027.

Reaching commercial operations at Cape Station is both a huge milestone for Fervo, and a turning point for the entire energy industry. No team has ever built a project like this anywhere in the world, and we did it ahead of schedule.Tim Latimer, CEO and co-founder, Fervo Energy

Google's money is riding on what comes next. A separate, much larger agreement, entered on August 26th and announced the following week, has Google buying 396 megawatts of Cape Station's output over a 15-year term, delivered in four 99-megawatt tranches targeted to start reaching commercial operation in the third quarter of 2028. That figure lines up almost exactly with Cape Station's second phase — an additional 400 megawatts Fervo says is already under construction. Google also holds an option to add roughly 600 megawatts more by June 2030, which would push its total commitment toward a full gigawatt — sources differ on the precise ceiling, citing figures between 950 megawatts and 'nearly 1 GW' — but either way, it's the largest contract ever signed for enhanced geothermal power, by a wide margin.

Here's the part that doesn't get top billing in either company's announcement: nobody has said where the data center this power is supposed to run is actually going to be built, or whether it will be built at all. Google's own language on the project is carefully conditional — plans remain subject to engineering feasibility, state and local approvals, and commercial conditions, the kind of sentence that covers a company whether it breaks ground next year or never. Google bought land near Eagle Mountain, Utah back in 2021 for a data center and has never built on it. Whether that site, some other Utah parcel, or nowhere at all ends up drawing on Cape Station's power is, as of this writing, genuinely unknown.

It's an inversion of the story we usually tell on this blog. Normally the data center is the fixed point — a real building, a real address, a company asking a utility whether it can get power there — and the power is the open question still climbing GridPermit's evidence ladder toward something contractually real. Here the power is already the most advanced part of the story: a 33-megawatt plant selling electricity under contract today, a 396-megawatt commitment with a named delivery schedule and a target date. The site that's supposed to consume any of it hasn't climbed past the bottom rung. I think that's worth sitting with before anyone treats 'Google signs record geothermal deal' as the same kind of fact as 'Google is building a data center.' One of those sentences describes an executed contract. The other describes a maybe.

What Milford got first

Long before any of this made national headlines, Cape Station was already reshaping its host county. Beaver County has roughly 1,400 residents, and when KUER visited Milford in October 2024 as construction ramped up, County Commissioner Brandon Yardley didn't undersell what the project meant for a town that size — 'in our little community of 1,400 people, this is a big deal for us,' he said — while also noting the more mundane cost of hosting it: he raises cattle near land where Fervo's crews had already started drilling, and told KUER plainly, 'it impacts my operations a little bit.' Mayor Nolan Davis called the project a tremendous asset to the Milford Valley; Scotty's Diner owner Melissa Wunderlich said she'd had to double her staff to keep up with the construction crews passing through.

The jobs numbers are a good example of why a single figure never tells the whole story. At groundbreaking in 2023, officials projected something like 6,600 temporary construction jobs and 161 permanent operational ones across the full build-out. Fervo's own count of its eventual permanent headcount is far smaller — up to about 30 full-time positions, three of them filled by local hires as of late 2024. Both numbers can be true: a project this size genuinely employs thousands of people while it's being built and a few dozen once it's running. The gap just means whichever number shows up in a headline is doing a lot of work to describe a very different kind of local benefit than the other one.

A new kind of wire, if regulators allow it

Fervo says its approach to Cape Station — drilling what it calls a 'GeoCluster' rather than a single well field — could let it deliver power straight to a customer's load under Utah's SB 132, instead of selling everything into the wholesale market the way generators normally have to. That direct-to-load pathway, if regulators approve it, is part of what Fervo is pitching as proof this deal costs existing Utah ratepayers nothing: the power is sold to Google specifically, not folded into the rates everyone else pays. That's Fervo's characterization of its own deal, not an independent finding, and the regulatory sign-off it depends on hasn't happened yet. It's the same claim nearly every one of these large-load power deals makes in its first press release — Michigan's DTE/Google contract and the Black Hills/Google deal in Wyoming both opened the same way — and it's worth checking against the actual tariff filing once one exists, not the announcement.

What actually happened this month is specific and real: a geothermal technology that didn't have a single commercial plant running a year ago now has one selling power under contract, on schedule, in a county of 1,400 people that's already felt the difference. What Google is betting on is also real, in the narrow sense that the money and the delivery schedule are signed. Whether either of those things ends up inside a server hall anyone can point to is still, for now, the only part of this story nobody can answer yet.