Google’s Cloud on Nevada Hot Rock
The most interesting thing about Google’s data centers outside Las Vegas is increasingly what’s happening under the ground and not inside the racks.
In a corner of Nevada’s geothermal country, a small but important power plant called Project Red has started delivering electricity into the local grid that serves Google’s facilities. It is only 3.5 megawatts today, enough to power a few thousand homes, but it represents something genuinely new: the first commercial enhanced geothermal project in the world, co‑developed with a hyper-scale cloud provider that wants firm, carbon‑free power every hour of every day.
This project, and the much larger Corsac Station that will follow it, are a practical test of an idea from my earlier piece that in current political setup, focused on reliable 24/7 baseload, geothermal actually becomes an unassuming winner. Now we get to see what that looks like when it stops being a policy abstraction and becomes the power backbone for a data‑center cluster.
Why wind and solar weren’t enough for Google
For a decade, the corporate clean‑energy story has been simple: sign big wind and solar PPAs, get lots of low‑cost megawatt‑hours, and declare yourself “100% renewable” on an annual basis. That works surprisingly well if all you care about is matching your yearly electricity consumption with an equal number of green megawatt‑hours somewhere on the grid.
It breaks down if you care about when the clean power flows. A hyperscale data center does not stop because the wind drops at 2 a.m. or a storm covers a solar array at 3 p.m. Server demand is stubbornly 24/7, while wind and solar are not. In the conventional model, the gap is filled by whatever the grid has handy, often gas or coal units that ramp to cover the difference.
A few years ago, Google decided it wanted to go further than this accounting trick. The company set a goal to run all its data centers and offices on 24/7 carbon‑free electricity by 2030, explicitly defined as “clean energy to meet our electricity needs, every hour of every day, on every grid where we operate.” That requires not just lots of cheap renewables, but a portfolio that includes firm resources, things that look, from a grid operator’s perspective, a lot like baseload plants.
Enhanced geothermal is one of the very few technologies that can check that box without relying on a fossil power plant. That’s why Google was willing to pay more for it.
Project Red: enhanced geothermal for the cloud
Geothermal itself is not new. For decades, plants have tapped naturally permeable hot water reservoirs in places like Iceland or California’s Geysers field. What is new is the ability to create a geothermal reservoir almost anywhere with hot rock, using techniques borrowed from unconventionals: directional drilling, multi‑stage stimulation, and engineered fluid circulation paths. University of Michigan Center for Sustainable Systems gives a good overview of this technology.
Project Red, developed by Fervo Energy in northern Nevada, is the first commercial‑scale proof that this enhanced geothermal approach can work for the grid. Fervo drilled a horizontal well pair with laterals over 3,000 feet long into hot rock around 191 °C, stimulated the rock to create a controlled fracture network, and then circulated fluid to bring heat back to the surface. A 30‑day well test showed a flow of 63 liters per second, enough to support about 3.5 megawatts of electric generation.
Those are record‑setting numbers for enhanced geothermal. More importantly, they translated into a stable power plant. Since late 2023, Project Red has been sending carbon‑free electricity to NV Energy, the local utility, which in turn supplies the same grid that feeds Google’s nearby data centers. This is not an offset or a distant REC purchase but a physical power on the same network where the load sits.
Get Isle of Tech’s stories in your inbox
Join Medium for free to get updates from this writer.
From a system perspective, the plant behaves like a small baseload unit. Capacity factors can exceed 90%, there is no fuel price to worry about, and output does not care about weather patterns. That is exactly the behavior a data‑center operator wants but has historically only been able to buy from coal, gas, or nuclear plants.
The Clean Transition Tariff: rate for firm clean power
If Project Red is the technology proof, Corsac Station is the business model test. Fervo is now developing a 115‑megawatt enhanced geothermal project in Nevada that will supply Google’s data centers under a new regulatory construct called the Clean Transition Tariff (CTT), recently approved by the Public Utilities Commission of Nevada.
The structure looks different from a classic renewable PPA. Under the CTT, NV Energy signs a long‑term contract to buy power from Fervo’s Corsac Station, then resells that electricity to Google under a dedicated tariff. Google agrees to pay a premium over standard grid electricity, covering the difference between the cost of the new geothermal resource and what the utility would otherwise have procured under least‑cost planning. In exchange, Google gets fixed‑price power and clear ownership of the environmental attributes.
There are two important consequences here. First, the extra cost of commercializing an emerging technology is not socialized onto other ratepayers; it is carried explicitly by the customer that wants the premium product. Second, the tariff is designed to enable true 24/7 matching: Google can claim that a specific geothermal resource is providing a specific block of continuous clean power into the same grid where its data centers pull load.
A hyperscaler is voluntarily paying above‑market rates to create a bankable, firm clean‑energy asset that behaves like infrastructure rather than like a bookkeeping exercise for annual renewable credits. It is using its balance sheet to drag an expensive technology down the cost curve in exchange for reliability and decarbonization.
It’s a bit misleading to call this “radical” or “innovative”, as many sources describe it. The truly strange innovation was the old model: buying credits and pretending they were sustainability. That was the accounting trick. What Google is doing with Fervo is just how energy should work if you’re serious about decarbonization and reliability.
The first hyperscale cluster built on enhanced geothermal
At the scale of actual data centers, Nevada is the first real example of a region being built around enhanced geothermal as a key power source. Today, Project Red’s 3.5 MW is a small part of NV Energy’s mix. In the second half of this decade, Corsac Station’s 115 MW will be large enough to matter for the region’s power planning and for where Google builds its cloud infrastructure.
Rough order‑of‑magnitude numbers help here. A 115‑megawatt geothermal plant running at high capacity factor can deliver on the order of a billion kilowatt‑hours over a decade, and hundreds of gigawatt‑hours per year, enough to power tens of thousands of homes, or a significant share of a data‑center cluster’s continuous demand. Estimates suggest annual emissions reductions in the range of six figures worth of tons of CO₂ compared with a fossil‑heavy baseline.
More importantly from an infrastructure standpoint, Google is no longer just “greening” its consumption with certificates; it is shaping the underlying generation fleet. Choices about where to site data centers, how to contract for power, and which technologies to underwrite begin to blur into grid planning decisions. In that sense, enhanced geothermal in Nevada is an experiment in how digital and physical infrastructure co‑evolve.
What this signals for AI-era energy systems
The Nevada experiment is drawing attention from other hyperscalers that are faced with the same problem: AI and cloud loads are climbing, traditional grid build‑outs are slow, and the easiest renewables are no longer enough to guarantee 24/7 low‑carbon operation. Fervo’s CEO has already described dozens of similar conversations as companies look for ways to anchor data‑center growth in firm clean supply rather than in ever more complex offsets.
If it works, the template is straightforward: pair a high‑growth cloud region with an enhanced geothermal resource, wrap it in a tariff or PPA structure where the customer pays the innovation premium, and use that to both decarbonize and stabilize part of the local grid. Over time, as drilling costs fall and reservoir designs improve, the premium shrinks and geothermal begins to compete head‑to‑head with gas on cost as well as on reliability.
In my earlier article, I argued that under a policy focused on energy dominance and round‑the‑clock baseload, geothermal was an odd but logical beneficiary, something that looked like a fossil plant to grid operators but did not burn fuel. The Nevada story makes that thesis concrete. A hyperscale cloud cluster leans on engineered system in hot rock to keep its servers running and in the process nudges the grid a little closer to a firm, clean future.
