Valar Atomics proposes 456-reactor SMR data center campus in Utah

Valar Atomics has put a number on the table that the small modular reactor business has so far only put on paper: 456. The proposal, called Project Beehive and first reported by NPR and the Salt Lake Tribune, would site those reactors, a fuel production facility, and nuclear waste storage on more than 9,000 acres of federally held land near Price, in Carbon County, Utah, 199 miles southeast of Salt Lake City.

According to the account of the company's filing with federal regulators that NPR reviewed, the fleet could total 9.6GW of electrical capacity, with thermal heat produced for reuse, and construction could reportedly start as soon as the end of this year. The first reactors are targeted for 2028 and full buildout for 2032, with about 650 acres of state land whose terms appeared in a land-use agreement discussed at a recent meeting of the Utah Trust Lands Administration.

Two facts frame the schedule. No company has yet put a small modular reactor into commercial operation, and the sponsor proposing 456 of them was founded in 2023 and is developing a Generation IV high-temperature gas reactor whose 5MW Ward 250 design only completed a zero-power fueled criticality demonstration at the Utah San Rafael Energy Lab in Emery County in June; in July, the company powered up the test reactor and ran a single Nvidia GPU off a simple thermoelectric system.

That record sits a long way from 456 reactors and a 2032 buildout, and getting from a critical 5MW design to a fleet measured in gigawatts is a manufacturing and licensing program, not a final design exercise—none of which the filing's account describes, from factory to fuel supply to construction sequence. None of that makes the proposal unserious, just early.

The demonstration's audience is easy to read, because running an Nvidia GPU off the test reactor ties the proof point to the load Project Beehive says it wants to serve—compute—rather than to the grid in the abstract. If the argument to a prospective data center tenant is that these reactors can power a rack, July was the smallest version of that argument the company could stage, and the filing is the largest.

The proposal's arithmetic is looser than a construction schedule usually runs, because Valar aims to deploy 25MW reactors at Project Beehive and 456 of those would sum to roughly 11.4GW, more than the 9.6GW the filing assigns to the fleet—while the outlet's headline uses a third figure, 9.4GW. The reporting does not reconcile those numbers, and nothing outside it can; a reactor count and a capacity total that move independently are what an engineering study exists to settle, or an early sign that the figures came from different drafts of the plan.

The larger omission is commercial: the campus would include data centers, in the filing's description, but the account names no operator, no tenant, and no buyer for the power, and it carries no price. For a project measured in gigawatts and spread across a decade, that leaves nothing on the other side of the electricity and no figure against which to underwrite the build—no offtake, no lease, no disclosed capital cost.

The proposal fits a familiar unpriced energy project, in which a milestone is announced without an owner, offtake terms, or price, pushing merchant risk onto the developer and turning completion into a financing event rather than evidence that the asset performs. Nuclear power adds a second layer of uncertainty, because a tenant asked to commit to first-of-a-kind reactors would be underwriting technology risk alongside power price—and no such tenant has been named.

Sizing the financing from the outside is guesswork because the filing carries no capital cost, and what can be described is what the proposal asks of its sponsor: a fuel production facility, waste storage, 456 reactors, and a decade of construction, none of it yet matched by an executed contract. A project of that shape tends to be assembled in stages, each contingent on the last, and the first stage is a permit.

The scope of the filing is itself worth noting, because Project Beehive as described is a fuel-cycle proposal as much as a power proposal, bundling a facility to produce nuclear fuel and others for storing nuclear waste onto the same 9,000 acres as the reactors. Fuel production and waste storage are separate regulatory questions from reactor licensing, and folding them onto one site gives the approvals process more to weigh than a power plant would.

Carbon County is where the proposal meets a competing idea: Altigen Energy, which develops natural gas plants for data centers, is targeting an area east of Wellington for a 1,200-acre project it calls 'Jurassic Spark' that would open with 250MW of gas capacity and eventually scale to 3GW, less than a third of Project Beehive's stated 9.6GW and built on natural gas rather than an unproven reactor design. Wellington sits about six miles south of Price, and Altigen filed its application with the county in March.

The two projects took different doors, with Project Beehive running through federal land and the National Environmental Policy Act—plus a state land-use agreement for the roughly 650 state acres—while Jurassic Spark went to Carbon County, and a county application and a federal one move on different clocks. That difference matters when the federal sponsor is also the one reported to be weighing construction by the end of this year.

One detail in the reporting is worth isolating: construction could start as soon as the end of the year, the coverage says, but the Bureau of Land Management's Utah office described its posture to NPR as 'We are currently reviewing the application for completeness.' An application still being checked for completeness on October 1 has not cleared the environmental review that follows it, and Valar does not claim otherwise, telling NPR, 'No construction begins before permits and approvals, and the public process leads along our internal targets set the pace.' The reported year-end start and the company's own caveat point in different directions—and the caveat is the one with the company's name on it.

What is documented and what is proposed are worth separating: the project is known through NPR's account of a filing it reviewed, the state-land terms through a land-use agreement aired at a Utah Trust Lands Administration meeting, and the federal posture through the BLM's confirmation that it holds the application—but no power contract, financing, or construction permit appears in that list.

The nearer gate is the one the bureau controls, and as the application moves through the environmental review, the BLM said it would 'make sure the public has meaningful opportunities to provide input'—the ordinary sequence for a project on federal land, and also the schedule. Consent has become a pre-construction currency, with permit conditions rather than lease terms pricing data center risk, and Project Beehive concentrates reactors, fuel production, and waste storage on land that is partly federal and partly state, a lot for a single environmental record to carry.

Set beside each other, the two proposals describe a fork: data centers are shopping for round-the-clock power, and Carbon County now hosts a gas bid and a nuclear bid at once, with the reactor project larger and more speculative and the gas project smaller and more buildable on today's terms. Dispatchable generation and grid access have been repricing while renewable pipelines still announce without buyers, and whichever project gets its permits and a signed customer first will say more about the market than either capacity figure does.

The BLM's completeness determination is the next checkpoint, and only the first; the 2028 first reactor, the 2032 buildout, and the 9.6GW fleet all sit downstream of it, and on the record so far none of them has a customer.