Sequoia led a $1 billion round for Valar Atomics at a reported $6 billion valuation, weeks after its Ward 250 test reactor reached criticality in Utah and powered an Nvidia Blackwell system at roughly 100 kilowatts thermal. The announced AI factory is 30 MW. The binding constraint is not physics but licensing — and Valar is a plaintiff in a live suit against its own regulator, now stayed pending settlement talks.
On 3 August, Sequoia Capital led a $1 billion equity round for Valar Atomics, with partner Shaun Maguire joining the board, alongside a $200 million credit line from Erebor and other lenders. Bloomberg reported the valuation at $6 billion post-money. Valar did not disclose a valuation itself, and the figure should be carried as Bloomberg's reporting rather than as a company statement — though TechCrunch had reported the round in talks at the same $6 billion in mid-July, which makes the number well-sourced.
The launch narrative more or less writes itself: factory-built nuclear for AI data centres, Sequoia leads, Nvidia partnership, energy abundance. All of it is real. None of it is the interesting part.
The interesting part is the distance between what has been demonstrated and what has been valued. That distance is not a criticism — it is the entire investment thesis, and a reader evaluating this deserves to see it measured rather than gestured at.
On 18 June 2026, Valar's Ward 250 reactor achieved self-sustaining criticality at the San Rafael Energy Research Center in Emery County, Utah. It was the second advanced reactor to reach criticality under the Department of Energy pilot programme established by Executive Order 14301. About a week later, at a demonstration, output was converted to electricity through a thermal-electric generator and used to power an Nvidia Blackwell system — reportedly the first time an advanced reactor has directly powered AI hardware. The system briefly hosted a website.
This should be stated clearly rather than sneered at: taking a novel reactor design from nothing to sustained fission in a working test facility is genuinely hard, and doing it inside a compressed timeline is harder. Valar's own account is that it took two years to complete the NOVA core and seven months to take Ward 250 critical. That is the company's characterisation of its own progress, and worth attributing as such — but the criticality event itself was independently confirmed through the DOE programme, which makes it the one fully external fact in this story.
Now the scale. Ward 250 is a roughly 100-kilowatt-thermal high-temperature gas-cooled reactor — TRISO fuel, helium coolant, graphite moderator. At the demonstration it was running at approximately 100 kW of thermal output. Converted to electricity at the efficiency a thermal-electric generator achieves, the delivered power was a small fraction of even that.
One hundred kilowatts of heat is roughly the output of a few dozen domestic space heaters. It is a real reactor producing real fission power, and it is a trickle of electricity.
The announced Nvidia AI factory is 30 MW and waterless. The Ward 250 design is intended to scale to 5 MWe per unit. So the ladder priced into $6 billion runs roughly: 100 kW thermal demonstrated, to 5 MWe per commercial unit, to a 30 MW installation, to the thousands of factory-produced reactors the business model requires.
From 100 kW thermal to a 30 MW facility is about two and a half orders of magnitude. Every intervening step is a manufacturing and licensing problem, not a physics problem.
That framing cuts both ways, and both readings are legitimate. The optimistic one is that physics risk is the risk that kills nuclear startups, and Valar has retired it — fission works, the core works, the fuel works, and what remains is engineering and paperwork, both of which respond to capital. The pessimistic one is that nuclear's graveyard is full of companies that proved the physics and then spent a decade failing to manufacture and licence at cost. Manufacturing and licensing are not the easy part. They are where the money goes and where the schedule dies.
The raise plus credit facility gives Valar about $1.2 billion to attempt that transition. Whether that is generous or thin depends entirely on the next section.
Ward 250 operated under a DOE-associated pathway, not an NRC commercial licence. Those are different regulatory objects, and the gap between them is where the commercial risk sits.
For reference: NuScale's VOYGR is the only small modular reactor the NRC has fully licensed, and it required more than a decade of review — for a light-water pressurised design with more operating precedent behind it than any other reactor type in existence. Valar's HTGR is a non-light-water design with far less regulatory precedent. Nothing about that guarantees a longer review, but nothing about it suggests a shorter one either.
And here is the fact that makes this a business story rather than a press release: Valar is a plaintiff against the agency it will need a licence from.
In April 2025, Valar joined a federal suit alongside the states of Texas, Utah, Louisiana, Arizona and Florida, and fellow reactor startups Last Energy and Deep Fission, challenging the NRC's licensing authority over certain small reactors. The suit contests the NRC's interpretation of the Atomic Energy Act — particularly its definition of a utilization facility — and seeks declaratory and injunctive relief requiring the agency to revise its licensing framework and undertake new rulemaking. If it succeeded fully, licensing authority for small-scale reactors could shift substantially to individual states.
The suit is live and, notably, is not being fought hard right now. In June 2026 the NRC joined the plaintiffs in asking the court to extend a stay to 29 September 2026 in order to pursue what the filing described as a mutually agreeable resolution that could avoid or limit further litigation. The US District Court for the Eastern District of Texas granted that motion on 30 June. Both sides have repeatedly paused the proceedings, which is the posture of parties negotiating rather than litigating.
Hold those two facts together without editorialising. A company whose entire scale-up depends on a regulatory pathway is simultaneously a party to litigation over the scope of that regulator's authority — litigation that appears to be heading toward settlement. That is not an accusation of anything. It is the actual shape of the risk, and it is what a reader evaluating $6 billion should understand.
The counter-case deserves equal weight, and it is strong. The litigation and the state-level pathways exist precisely because reform advocates argue that NRC review is the bottleneck strangling nuclear deployment in the United States — that the agency's framework was built for gigawatt light-water plants and applies poorly to reactors three orders of magnitude smaller. If they are right, and if the settlement talks produce a workable alternative pathway, the constraint could loosen far faster than the NuScale precedent implies. A decade-long review is the base case under the current framework. The bet here is partly that the framework changes.
Zoom out and the sector context clarifies what this raise does and does not signal. Antares raised $470 million in late July to build reactors for the US military. X-energy raised roughly $1 billion through an IPO in April. Deep Fission is public. Advanced nuclear is not short of money.
It is short of three other things. High-assay low-enriched uranium fuel supply, where Western capacity remains thin and the TRISO fabrication base is small. Factory throughput, because the entire premise of factory-built nuclear is a production line that has not been built and whose unit economics are therefore projected rather than observed. And regulatory clearance, discussed above.
Money solves the first two eventually. It does not solve the third, which is why the September court date matters more to this company's trajectory than the next funding round will.
Almost every capability and timeline claim in the Valar story is Valar's account of Valar, and it should be read that way. The valuation is Bloomberg's reporting. The criticality event is externally confirmed and genuinely impressive. If you want one question to track from here, it is not how much power the reactor is producing — it is what comes out of the NRC settlement talks by the end of September. That determines whether the ladder from 100 kilowatts to 30 megawatts has rungs on it.
Ward 250 is a roughly 100-kilowatt-thermal reactor, and at the Nvidia demonstration it was producing on the order of 100 kW of thermal output, converted to a much smaller amount of electricity through a thermal-electric generator. The design is intended to scale to 5 MWe per unit, and the announced AI factory is 30 MW — about two and a half orders of magnitude above what has been demonstrated.
DrafterDaily separates what a company has demonstrated from what it has been valued at — across AI, energy and deep tech.
Subscribe freeThe headline was the 94% drawdown from a $3.44 billion peak. The more useful arithmetic is the cash sitting inside the thing being bought — and the fact that everyone except the angels was paid in another private company's paper.
The biography is the least interesting part. What the board actually did was unbundle the CEO job — keeping Tim Cook on with a defined policymaker-engagement remit — and that separation says more about Apple's next decade than the appointment does.
Every outlet reported the size of Meta's teen-safety settlement and the screen-time limits. Almost nobody read the payment structure — where roughly $5.3bn is released only if two competitors act.