The Nuclear
Company

The most exciting thing here is repetition

Walk into the new engineering office in Columbia and you will not find a lab full of exotic prototypes. You will find schedules. Supply-chain maps. Sensors waiting to be bolted to a construction site. The Nuclear Company is in the business of building nuclear power plants, and its single most radical decision is to not invent anything new about the reactor itself.

That is the whole pitch. Take a reactor design that regulators have already studied and licensed. Build it. Then build it again, and again, the same way each time, until the construction crews could practically do it in their sleep. The company calls this "design once, build many." Most of the nuclear industry calls it heresy. The Nuclear Company calls it Tuesday.

Founded in 2023 and backed by roughly $70 million, the company has raised the money, signed the partnerships, and planted a flag in one of America's most nuclear-friendly states. What it has not yet done is pour concrete on its first plant. That is the part everyone is watching.

Nuclear's enemy was never the atom

Here is the uncomfortable truth about American nuclear power: the physics works fine. It is the construction that keeps going wrong. The last big U.S. reactor projects ran years late and billions over budget, and the reason is almost embarrassingly mundane. Every plant got built like it was the first one ever attempted - new designs, new crews, new mistakes, all paid for at full price.

Engineers have a name for this expensive habit: the "first-of-a-kind" penalty. Build something once and you pay for every lesson learned the hard way. The trouble is that the U.S. has spent decades treating nearly every reactor as a first-of-a-kind. Which is a bit like reinventing the wheel each morning and then acting surprised the commute is slow.

Meanwhile, the demand side stopped being patient. AI data centers are projected to push electricity demand up sharply by the end of the decade, and Google, Amazon, Meta, and Microsoft have all gone shopping for clean, around-the-clock power. The grid wants carbon-free baseload it can count on at 3 a.m. Wind and solar are cheap but moody. Nuclear is steady but slow to build. Somebody had to fix the "slow to build" half.

Three builders who had each scaled something hard

The Nuclear Company was started by three people who had already learned, in other industries, how brutal it is to scale physical things. Jonathan Webb ran AppHarvest, an indoor-farming company - he went, more or less literally, from greenhouses to gigawatts. Kiran Bhatraju built Arcadia into a large clean-energy software business. Patrick Maloney led the climate fund CIV. None of them is a reactor physicist, and that is rather the point. Their bet is that nuclear's bottleneck is not nuclear science. It is project management.

The wager attracted a specific kind of investor - ones comfortable with long timelines and heavy capital. Eclipse led the Series A, joined by Climate Investment, True Ventures, Wonder Ventures, Goldcrest Capital, and MCJ Collective. The money is not for a science experiment. It is for site evaluation, permitting, software, and the deeply unglamorous work of getting a coalition of utilities, regulators, and financiers to agree on something before the first shovel moves.

A reactor strategy, wrapped in software

The Nuclear Company is really two ideas stacked together. The first is the fleet-scale build: pick from roughly a dozen U.S. sites that already hold combined operating licenses or early site permits from the Nuclear Regulatory Commission, use large reactor designs that are already proven, and repeat the construction process across the fleet so crews and lessons travel from one site to the next.

The second idea is the layer that makes the first one credible. With Palantir, the company is building the Nuclear Operating System, or NOS - described as the first AI-driven, real-time software system built specifically for nuclear construction. Sensors across a site feed a digital twin, a live virtual model of the build. When reality drifts from the plan, the software notices before the budget does.

It is, if you squint, a slightly ironic arrangement: the most futuristic thing about The Nuclear Company is the software it uses to build something deliberately old-fashioned. The reactor is proven. The novelty lives in the spreadsheet.

The numbers behind the bet

A nuclear startup cannot show you a finished plant yet, so it shows you commitments. The funding closed. The partnerships are signed. The office is open and hiring. And South Carolina was not a random pin on the map - the state already generates more than half its electricity from nuclear and ranks third nationally in nuclear output, which means the workforce and the supply chain are already nearby.

Build the way America used to build

Strip away the AI and the digital twins and the mission is almost nostalgic. The Nuclear Company wants to make building nuclear plants feel routine again - the way the country once threw up dams, highways, and an interstate system without treating each mile as a miracle. The goal is not a single heroic reactor. It is a fleet boring enough that nobody writes a documentary about how hard it was.

Skeptics have fair questions. Reactors that come online in the early 2030s are betting on demand a decade out, and demand forecasts have a habit of aging badly. Big nuclear projects have humbled confident builders before. The Nuclear Company's answer is that repetition is precisely the hedge: a fleet absorbs cost shocks that would sink a one-off.

Five things that make it odd, in a good way