Bring Your Own Power: The Great Basin's 10-Megawatt Window

For two years the AI buildout was a chip story. It isn't anymore. Satya Nadella said the quiet part out loud last fall — the constraint isn't silicon, it's "warm shells," powered buildings to plug the silicon into. Jensen Huang, asked whether energy was the real bottleneck, used three words: "It's the bottleneck." When the company selling the GPUs and the company buying the most of them agree the scarce resource is electricity, the argument is over.

Most of the coverage that followed went straight to the biggest possible number — gigawatt gas plants, nuclear-powered "data cities," Stargate. It makes for great headlines and it's mostly irrelevant to the operator trying to stand up a 10-to-50-megawatt facility this year. We spend our days at that smaller, faster-moving end of the market — sourcing the power, the turbines, and the GPUs that have to land in the same window — and from where we sit, the most underpriced power opportunity in the country right now isn't a gigawatt anything. It's a well or two in the Nevada desert.

The grid in Nevada is already full

Start with why this is even a conversation. Northern Nevada — Reno, Sparks, the Tahoe–Reno industrial corridor out by the gigafactory — has become one of the densest data-center build zones in the West, and the local grid can't keep up. The honest read from inside the utility planning cycle is that meaningful new capacity is roughly four years out — call it 2030 before the supply picture loosens. In the meantime the available power goes to whoever can pay, and the entity buying it has money: data centers outbid residential load every time. Around Tahoe the supplier is an out-of-state utility subsidiary, and the optics of routing the region's power to hyperscalers while locals wait are exactly as bad as they sound.

That's the macro story everywhere — grid interconnection queues now run past four years nationally (Berkeley Lab, Queued Up, 2025). But Nevada has a wrinkle the national pieces miss. Under existing Nevada law, a genuinely self-sustained generator can bypass the utility entirely. Build your own power on-site, behind the meter, serving only your load, and you're not in the queue, not paying the tariff, and not waiting on 2030. That legal seam is the whole game. The utilities know it and don't love it; it's the kind of thing that gets fought over with lobbyists and lawmakers. But the door is open — and in the Great Basin there's a particularly good thing to put behind it.

What's actually under the Great Basin

Nevada sits on one of the best geothermal resources in the United States. The USGS has estimated that enhanced geothermal in the Great Basin alone could eventually supply on the order of 10% of the country's electricity (USGS, 2025). The reason is geological and a little counterintuitive: the basin is being pulled apart — extending on the order of 16 millimeters a year — and it's seismically restless. There are earthquake swarms in the Silver Springs area that would make a homeowner nervous and make a geothermal driller smile, because every fracture is a fresh path for heat to move toward the surface. In this business broken rock is the asset. A solid block of hot granite is useless; a fractured fault zone is a straw into a milkshake.

For a data center, that subsurface heat does two distinct jobs, and conflating them is the most common mistake people make.

Power. Where heat rises along fractured fault zones, you drill a well or two and run a turbine. Independent, smaller-scale projects in this region deliver on the order of 10 to 20 megawatts off one or two wells. That's not a gigawatt. It's also exactly the size most real facilities actually need, and it fits on a footprint that surprises people — a well pad about the size of a tennis court, a turbine package roughly a hundred feet by fifty, ten to twenty acres all-in with room to breathe. It is not loud. Stand next to it and you get a low electrical hum, not a jet engine.

Cooling. This is the underrated half. Further from the heat, where the ground sits at a stable ~57°F year-round, you run a closed-loop fluid through it and into a heat exchanger against the data hall's cooling system. It generates no power; it removes heat — meaningfully cutting the cooling load and, with it, the water draw that is the single loudest community objection to data centers in the desert (LBNL, 2025). Critically, a closed loop pulls nothing out of the aquifer — no steam tower, no drawdown. It's the opposite of the boomer-era geothermal people picture.

Why 10–20 MW is the sweet spot, not the consolation prize

Here's the contrarian part, and it's the thing I'd bet on. The market's attention is stuck at the extremes — the 500-megawatt EGS flagships (Fervo's Cape Station in Utah, ~$1.6B, Bill Gates money) and the gigawatt nuclear fantasies. Those are real, and they're someone else's problem. The independent 10-to-20-megawatt self-sustained project gets treated as the small-fry option, and it's actually the one that pencils for the operator who needs power now:

• It's competitive per megawatt. All-in, a self-sustained geothermal block in this range lands in the neighborhood of what operators are paying to buy gas turbines outright on the open market — and turbines are a sold-out, multi-year-lead-time market right now (Utility Dive, 2025). Geothermal's cost is front-loaded into drilling; after that the fuel is free and the maintenance is close to nothing.
• There's no utility tariff, because you've bypassed the utility. Stack the available clean-energy credits on top and the operating math gets hard to argue with.
• It's geology, not acreage. This trips everyone up: a precise ten-acre parcel over the right three-dimensional hydrothermal structure is worth more than six hundred acres over cold rock. Output is set by where you drill, not how much land you hold. The winners aren't whoever buys the most dirt — they're whoever has the geological read and the lease in the right spot.
• The public-image math is lopsided. Compare it honestly to nuclear, the other "firm clean power" answer everyone reaches for. Nuclear gives you more power, faster, with a fifty-year track record — and a waste liability and a NIMBY problem that's brutal next to a data center. Small modular reactors are the shiny new thing and they inherit the same fight. Geothermal asks for high upfront capital and then goes quiet for decades: no waste stream, no fuel-price exposure, and a story communities will actually accept.

The catch — and it's a real one — is that the binding constraints are drill-location intelligence and the capital to secure a lease, not land or technology. Geothermal rights in Nevada are state-owned fluid minerals, auctioned by the BLM roughly once a year (BLM geothermal program), with parcels running from about $50K to north of $500K. You don't buy land and drill; you win the right lease and prove the resource. That's a specialist's game — which is why this market is relationship-gated, not capital-gated the way gas is.

What we're seeing, and where we sit

We source across the full stack — the gas turbines that bridge the gap, the GPUs that fill the racks, and increasingly the harder thing: connecting a site to power that isn't on anyone's grid. A lot of the inbound we field is some version of "find me cheap land and turbines to put on top." That instinct is fine. But in the Great Basin specifically, it leaves the better option on the table. For a 10-to-20-megawatt facility within a half hour to an hour of Reno–Tahoe, self-sustained geothermal is a genuinely competitive, genuinely cleaner answer — and one most buyers don't even know to ask for.

We're plugged into that ecosystem here — the developers who hold the geological expertise and the government relationships, the people who know which leases are coming up and where the heat actually is. We don't own megawatts; nobody hands those out. What we can do is connect a serious 10+ MW project to the people who can put steel in the ground, and run the procurement around it — power, cooling, and compute as one stack instead of three disconnected scrambles. (The compute half of this story — what the GPU resale market is actually telling buyers — is here.)

And there are eyes on this region for exactly this reason. The land is being mapped, the leases are being watched, and the operators who move in the next twelve to twenty-four months — before the grid loosens in 2030 and before the geothermal lease map gets picked over — will lock in an advantage that's very hard to replicate later. The window is open because most of the market is staring at the wrong end of it. (Why that twelve-to-twenty-four-month window is the only one that matters — and why the grid and the turbine line are both closed to you — is the third piece in this series, on time-to-energized as the real constraint.)

If you're sizing a site in northern Nevada and the grid answer is "2030," that's not a no. It might be the best thing that ever happened to your power bill. Talk to us — or browse the stack we put around projects like this.

---

Pantheon Research is our series on the infrastructure behind AI: power, turbines, cooling, and the procurement reality that decides who actually ships compute. Field notes from the deal flow, not the keynote.