Valar Atomics' SMR Reaches Criticality in Utah

On June 18, 2026, in the Utah desert, a reactor called Ward 250 achieved criticality. According to IE Explains, this made Valar Atomics the first startup ever to switch on a self-sustaining nuclear chain reaction. The machine ran at around 100 kilowatts — a fraction of its 5-megawatt design rating — but that is not quite the point. The point is that it ran at all, and that it ran as a small modular reactor, the class of machines the nuclear industry has been promising would change everything for the better part of a decade.

Whether Ward 250 represents a genuine inflection point or another well-funded chapter in nuclear energy's long history of almost-theres is a question worth sitting with carefully.

The problem that hasn't changed since 1979

To understand what Valar is attempting, you have to start with the physics that nobody in nuclear power gets to ignore. As IE Explains puts it: "Switch off a nuclear reactor and it does not stop making heat. Fission stops, but the fragments left behind keep decaying, still giving off 6 to 7% of full power the moment you shut down."

On a large core, that residual heat load can reach 200 megawatts. This is not a design flaw — it is an inescapable consequence of fission. It is also precisely what destroyed Three Mile Island in 1979 and Fukushima in 2011. Both reactors had already scrammed. Both still melted down, because the cooling systems meant to manage decay heat failed under conditions nobody had adequately planned for.

Every generation of reactor designers since has been, in one way or another, trying to solve this. The conventional light-water reactor answer involves enormous engineering margins: thick steel pressure vessels, reinforced containment structures, redundant emergency injection systems, and pumps that must run continuously for roughly a day after shutdown. As IE Explains frames it, "the safety case for water is that you need the pumps to never fail." That is a strong safety case until it isn't.

Generation 4 reactor design starts from the opposite premise: make pumps unnecessary in the first place. The TRISO fuel technology at the heart of several Gen 4 approaches — including Valar's — is central to this ambition.

What Ward 250 actually is

Ward 250 is a prismatic gas reactor. Hexagonal graphite blocks hold compacts of TRISO fuel — tristructural isotropic particles, uranium kernels under half a millimeter across encased in four layers of ceramic and carbon — while helium flows through the block structure carrying heat away. Control rods run on chain drives. A water-jacketed cooling system outside the vessel operates passively, boiling and condensing without any external pumping.

The design is engineered so that if everything goes wrong — power cut, operator error, mechanical failure — the physics of the core itself prevent a meltdown. As temperatures rise in a graphite-moderated, helium-cooled core, the reaction rate naturally decreases. The core cannot run away. It simply cools down.

This is not a theoretical claim. China's HTR-PM at Shidaowan, commercially operational since December 2023, demonstrated it in practice: in 2024, operators cut all power at full load and watched both modules cool themselves without intervention. That is passive safety functioning on a commercial machine, not a test rig.

Valar's reactor is far smaller than the HTR-PM — Ward 250 is designed to be deployable at a data center, industrial facility, or remote site, not a utility-scale power station. The modularity is the feature: if you need more output, you add another unit.

The funding and what it's for

On August 4, Valar announced a $1 billion Series B funding round led by Sequoia Capital, along with a $200 million credit facility. The stated purpose is not refinement of the existing machine — it is manufacturing scale. Valar says it intends to produce fleets of reactors, and that it will manufacture its own TRISO fuel in facilities co-located with the reactors that burn it. Every hour Ward 250 runs informs the next build.

The investor logic is legible. Data center construction is running into hard constraints: new builds are being refused or restricted over power draw and water consumption, and this problem is particularly acute in smaller communities where the resource demands of AI training infrastructure are wildly disproportionate to local supply. Some operators — SpaceX and Amazon among them — have floated orbital data centers as a workaround, which tells you something about how intractable the terrestrial problem is beginning to feel. An SMR that can be sited adjacent to the load it serves, requires no carbon fuel, and runs for decades without refueling is a genuinely attractive engineering solution to this problem. Whether it is also a practical, licensable, publicly acceptable one is a different question.

The gap between physics and machine

Here is where the honest accounting gets harder. Four reactors reached criticality under a Department of Energy program by July 4, and all four were zero-power tests. As IE Explains notes, this "proves the physics, but it proves almost nothing about how the machine behaves hot."

Gas-cooled reactors have a specific historical failure mode worth naming: they tend not to fail as reactors. They fail as machines. Peach Bottom and Fort St. Vrain, the American high-temperature gas reactor experiments of the 1960s and 70s, worked neutronically. They were wrecked by bearings, seals, moisture ingress, and corrosion — the mundane engineering problems that accumulate when you run hot gas through precision machinery for years. The neutron physics was fine. The helium wasn't staying clean.

The tests that actually matter are still ahead. Radiant Nuclear, whose Kaleidos microreactor fits inside a shipping container at 1.2 megawatts electrical, recently received its first TRISO fuel shipment at the Idaho National Laboratory's DOME facility, according to Radiant's own reporting, clearing the path to full-power, full-temperature testing — a program that includes 150 hours of unattended operation. That test will be worth watching closely, because it addresses exactly the question Ward 250's criticality test cannot: does this machine hold together hot?

A competitive field, not a single bet

Valar is not alone, and the approaches diverge meaningfully. X-energy's Xe-100 runs at 200 megawatts thermal using a pebble-bed design where spherical fuel elements circulate continuously and can be swapped out while the reactor operates — a fundamentally different manufacturing and refueling logic. Kairos Power's Hermes uses the same TRISO pebbles but replaces helium with molten fluoride salt as coolant; salt carries far more heat per unit volume and operates near atmospheric pressure, eliminating the pressurized vessel entirely. Westinghouse's eVinci and Antares' Mark-1 move heat through sealed sodium heat pipes — no pumps, no valves, no pressure — though that architecture trades simplicity for constraints on how large the machine can grow.

The subcritical reactor approaches being developed in parallel represent yet another branch of the safety engineering tree, one that sidesteps the decay heat problem through an entirely different mechanism. The field is genuinely plural right now, which is either a sign of healthy competition or a sign that nobody has yet found the dominant design.

The question public acceptance hasn't answered

There is one variable none of the engineering addresses, and it sits outside the reactor vessel entirely. IE Explains names it plainly: "the obvious objection is that local residents may not want to be in close proximity to a shoe factory or logistics hub that has the potential to cause a nuclear disaster."

The technical response to this objection is strong. A Gen 4 reactor that physically cannot melt down is not the same risk proposition as a 1970s light-water reactor under grid failure conditions. But risk communication around nuclear energy has a decades-long credibility deficit that better engineering alone does not repair. The communities most likely to host these reactors first — industrial zones, data center corridors, remote energy-constrained regions — are not uniform in their appetite for this trade-off.

Valar has a working reactor. It has a billion dollars. It has a market problem that its technology could plausibly address. What it does not yet have is proof that the machine works hot, that it can be manufactured at scale without accumulating the mechanical failures that killed the previous generation, or that the communities it needs to say yes will say yes.

The physics have been understood since the 1960s. What is being tested now — by Valar, by Radiant, by everyone else in this field — is whether anyone can build these machines over and over and keep them running. That test has barely begun.

Olivia Meng is a climate and environment correspondent for Buzzrag.