Every Industrial Revolution Has Run on a New Fuel

Synopsis

In February 2026, Valar Atomics airlifted the unfueled components of a nuclear microreactor across three military cargo planes — evidence, the company argues, that reactors can now be built like products rather than one-off megaprojects. This piece unpacks the technology behind Ward 250, a helium-cooled, TRISO-fueled reactor that reached criticality in roughly nine months, and the manufacturing logic (borrowed from 1930s aviation economics) behind betting that mass production can do to nuclear costs what it did to solar panels and batteries. It also traces the coalition — Nvidia, the Department of Energy, the US military, and state government, that compressed decades of regulatory and industrial timeline into months, and why AI's colossal electricity demand is the catalyst behind it all.

"An academic reactor or reactor plant almost always has the following basic characteristics: it is simple, it is small, it is cheap... it can be built very quickly... On the other hand, a practical reactor can be distinguished by the following characteristics: it is being built now, it is behind schedule, it requires an immense amount of development on apparently trivial items."
— Admiral Hyman Rickover, "Paper Reactors, Real Reactors" (1953)

Quite an understatement to say times have changed since then.

Take a second and imagine a nuclear reactor being airlifted by three separate C-17 Globemaster aircraft.

Three of the largest military cargo planes in the U.S. fleet, each hauling a piece of something that, until pretty recently, required a decade of permitting and a construction site the size of a small town.

This is exactly what happened with one of Valar Atomics' nuclear reactors.

In February 2026, the unfueled modular components of the reactor were trucked to March Air Reserve Base near Riverside, Southern California, loaded onto three US Air Force C-17s, and flown to Utah for final assembly and testing.

The whole point, as the company tells it, is that a reactor built like this can be rapidly transported and deployed, for a military base that needs consistent power in a remote location, or to power AI's ravenous need for energy.

But obviously, this isn't the nuclear reactor you're picturing.

Forget the immense cooling towers, the sprawling containment domes, the water reservoirs the size of lakes.

What flew to Utah was small enough that huge portions, at 100 kilowatts thermal (the nuclear reactions inside the reactor are currently generating heat at a rate of 100,000 joules every second), fits inside a package that can indeed go above the clouds.

So what actually is this thing?

Valar's reactor, called Ward 250, belongs to a category called a high-temperature gas-cooled reactor, or HTGR. Where a conventional nuclear plant circulates pressurized water through the core to move heat away from the fuel, Ward 250 circulates helium — an inert gas that can't become radioactive on contact and can't flash into steam in a loss-of-coolant accident.

The fuel itself is the more interesting part.

It's called TRISO — tri-structural isotropic; and each particle is roughly the size of a poppy seed, with a uranium kernel wrapped in ceramic layers and a silicon carbide shell that functions as its own individual containment vault.

Because each particle is essentially a self-contained unit, TRISO is considered one of the more robust nuclear fuel forms developed — which matters a lot for a reactor that's meant to be shipped around and dropped into varied locations rather than bolted permanently to one site.

At high burnup, the fuel retains over 99.99% of its fission products, which simplifies both accident tolerance and eventual waste handling.

The core itself is graphite-moderated and engineered to run at temperatures well above conventional reactors, which is what opens it up to industrial heat applications, not just electricity.

Right now, Ward 250 is a test article, rated at 100 kilowatts of thermal output for its criticality demonstration, with a design path toward 5 megawatts of electricity once fully scaled.

To put that in perspective, that 5 MW target would be enough to power somewhere around 5,000 homes — a fraction of what a conventional plant produces, and that's the point.

The company isn't trying to build one enormous reactor.

The company's actual plan is to build what it calls "gigasites" — clusters of thousands of these small HTGRs producing industrial power and carbon-based fuels more cheaply than oil.

Think of it less like building a cathedral and more like running a factory line: standardize the unit, then stamp out as many as demand requires.

The electricity generation itself is refreshingly unglamorous at this stage.

Hot helium exiting the core was converted to electricity during Valar's July demonstration using a thermoelectric generator — simple and reliable, if not especially efficient.

The commercial version is expected to swap that out for a Brayton-cycle turbine, which uses the hot pressurized helium to spin a generator directly, at efficiencies competitive with conventional power plants.

So no, it's not there yet.

But the architecture is designed to get there without redesigning the core.

Reactors as a learning curve

There's a piece of 1930s aviation economics buried inside Valar's pitch, and it's worth naming because it explains why Valar Atomics uses Wrights Law's scaling logic to argue that mass-producing modular microreactors like their Ward 250 can dramatically cheapen nuclear energy.

In 1936, an aeronautical engineer named Theodore Wright noticed something while studying airplane manufacturing: every time cumulative production of a given aircraft design doubled, the total cost to produce each additional unit dropped by a consistent percentage.

He measured it initially through labor hours, since that was the cleanest data available at the time, but the underlying principle turned out to be bigger than labor, it was about the full batch: tooling, materials, supply chains, process knowledge, all compounding downward together as more units got built.

Not a dramatic breakthrough, but instead the slow accumulation of a thousand small ones.

This became known as Wright's Law, and unlike Moore's Law, which ticks forward on a calendar — Wright's Law moves forward on production volume.

Cost doesn't fall because time passed; it falls because you built more units.

It's the same math that later explained why solar panels and lithium-ion batteries got dramatically cheaper as the world manufactured more of them, at a rate that was strikingly predictable once you knew the doubling curve.

The number that actually matters isn't the valuation

In early August 2026, Sequoia led a $1 billion round for Valar Atomics, at a transaction valuing the company at $6 billion including the new capital.

Big number, sure.

But the headline figure isn't the interesting part.

What's interesting is the shape of the curve underneath it: Valar raised a $19 million seed round in February 2025, grew that to $130 million by November 2025, and then closed a $450 million round in March 2026 at a $2 billion valuation.

Four months after that $2 billion mark, the company was worth $6 billion.

That's roughly a 3x jump in a little over 100 days, for a company whose actual product — molecularly speaking — is still a single test reactor running at 10 kilowatts of thermal output in the Utah desert.

Capital won't move like that on vibes alone.

It moves like that when a specific piece of physical evidence removes a specific category of doubt.

And the evidence, in this case, was Ward 250 actually working, achieving criticality, ramping power, and powering real hardware.

Which brings us to the deeper question: why did investors, the Department of Energy, and Nvidia all converge on this exact moment to compress a multi-decade industry into an 18-month sprint?

Every industrial revolution has run on a new fuel

Coal didn't cause the Industrial Revolution so much as permit it — it was the energy substrate that let steam engines, then factories, then railroads exist at scale.

Oil did the same thing a century later: it restructured global trade, war, and geopolitics around wherever the oil was.

Valar's own framing puts it plainly: coal powered the Industrial Revolution, oil fueled global commerce, and together they built the modern world.

Each leap in what civilization could do was gated by a leap in what civilization could burn.

AI is now running into the same wall electricity has always run into, except faster, and with a training run's worth of GPUs instead of a locomotive's worth of coal.

The numbers are genuinely startling.

According to the International Energy Agency, global data center electricity demand hit roughly 485 terawatt-hours in 2025 (an amount roughly equal to the entire annual power consumption of Germany), up 17% year-over-year, with AI-specific facilities growing at 50%.

Gartner expects worldwide data center power demand to rise 27% in 2026 alone, reaching 132 gigawatts, on its way to roughly 290 gigawatts by 2030.

And the U.S. specifically is already short: Goldman Sachs Research puts the country's structural data center power shortfall at 9.3 gigawatts in 2026, a gap expected to widen to 45 gigawatts by 2028 — equivalent to the annual electricity needs of about 34 million American households for example.

That's an entire new fuel-and-infrastructure sector being demanded on a timeline the existing grid was never built to meet.

Solar and wind can't be situated fast enough or run reliably enough for the always-on, firm power eons of GPU clusters need.

Gas turbines have years-long order backlogs.

What's left, if you want dense, reliable, transportable energy on a timeline measured in months rather than decades, starts looking a lot like nuclear — provided someone can figure out how to build reactors the way you build products instead of the way you build cathedrals.

That's the opening companies like Valar Atomics, Kairos Power, Antares Nuclear, and Oklo amongst many others are accelerating through.

Sam Altman, OpenAI CEO, and chairman of nuclear startup Oklo:

"I don't see a way for us to get there without nuclear."
— CNBC, 2021

Nvidia and the state

These are the ghost notes that gets sidelined in most coverage: it's not just that AI companies want nuclear power.

It's that Nvidia and the federal government are actively compressing the timeline that makes it possible, in ways that look nothing like how nuclear power normally gets built.

On the Nvidia side: Valar didn't just sign a power purchase agreement, weeks before its Series B closed, Ward 250 became the first reactor in U.S. history to power an Nvidia AI chip directly, converting reactor heat to electricity in a live demonstration.

That's a chipmaker treating a nuclear startup's reactor as a proof point worth staging a demo around, a level of integration that doesn't happen unless Nvidia sees firm, on-site nuclear power as a real bottleneck-breaker for its own customers' data center buildouts.

On the government side, the timeline compression is even starker.

Ward 250's criticality happened under a Department of Energy Reactor Pilot Program established by a May 2025 executive order that set a target of at least three advanced reactors reaching criticality by July 4, 2026 — the country's 250th birthday.

That February airlift wasn't a commercial shipping decision — it was a joint operation between the DOE and the Department of War, with U.S. Energy Secretary Chris Wright personally accompanying the flight and framing it as a milestone for military energy resilience.

And at the state level, the project sits inside Utah Governor Spencer Cox's "Operation Gigawatt," an initiative explicitly aimed at doubling the state's power production within a decade.

Stack that up: a chipmaker demoing reactor-powered silicon, a president's executive order setting a national deadline, the military airlifting reactor components, and a state government building industrial policy around the same site.

That's four different institutional actors, none of whom build things on the same clock as venture capital, all synchronized around one 27-year-old's startup reaching a specific milestone by a specific date.

For context on just how unusual that clock speed is: traditional nuclear projects like Georgia's Vogtle Units 3 and 4 took roughly 14 years from initial license application to commercial operation.

Valar went from breaking ground to a critical reactor in about nine months.

That compression is the real story sitting underneath the funding round.

Sequoia isn't betting on only a reactor.

It's betting that the entire apparatus around nuclear deployment; regulation, logistics, political commitment, just got rebuilt in real time, specifically to make companies like Valar Atomics' timeline possible.

"Energy is the only universal currency; it is necessary for getting anything done."
— Vaclav Smil, Energy and Civilization: A History (2018)