The Army Isn’t Buying Power. It is Buying an Industry.

The headlines declared the Army is going nuclear. On August 26 a new service called Janus commenced. It will cost up to $2.2 billion, include more than twenty small commercial reactors, is composed of five companies and five Army bases, with first power due by September 30, 2028. The project spans from Fort Bragg in North Carolina to Fort Drum in New York, with three posts in between. A quick glance at the headline reads like the Pentagon just bought a working product to keep the lights on for the 82nd Airborne. But not so fast, as these are small machines. Units from a company called Radiant, which holds the largest disclosed deal at up to $750 million, are just one megawatt each, and remain connected to the main power grid. In other words, the reactors are not the power source for an Army base of 50,000 soldiers. Instead, they are back up for critical loads if the commercial grid goes down. OK, but there is still much more going on here than the “Army goes nuclear.” In fact, officials running Janus told reporters, out loud, that most of the reactor companies will miss the deadline, that nobody yet knows what the reactors will actually cost, and that the American taxpayer is covering the brutal early costs that no private buyer would ever agree to pay.

The Army wasted no time sharing the project’s shortcomings, so what is really going on? Since none of these machines is a finished product, and not one of the five firms involved in Janus currently has a commercial microreactor available for purchase, is undoubtedly more to the story. Jeff Waksman, the Army official running the program, summarized the backstory, noting that the Pentagon intends to move these five companies “from experiments and prototypes to actual commercial products.” If it is not simply shopping for power, why is the Pentagon paying to construct an industry that cannot yet sell what it builds?

Waksman was forthcoming about the money, warning those involved not to divide the $2.2 billion by the number of reactors, because that is not what they cost. Until more than one has been built, “nobody knows what these microreactors are going to cost,” he said. Instead, he noted that the $2.2 billion is the ceiling on milestone payments (the money released as each company hits its technical targets), not the price tag, and it is understood that the first units produced will be more expensive. No individual Army base could ever buy a first-of-a-kind reactor on its own because “that is never going to pencil out,” he explained. So, in a clever workaround, the government is absorbing the early, brutal, money-losing part of the curve here, which would, in actuality, be the nail in the coffin for most commercial deals. The payments, Waksman shared, “can be seen as a form of subsidy to help these companies get there.” The Army’s hope is that most of the total investment will eventually be private. In the meantime, with no required split written into the contracts, public money is covering the early, unsellable part. But hoping is not the same as a done deal, and nothing in the contracts obligates these five companies to raise the private money that the Army is eventually counting on. Thus, if that aspect fails to materialize, the taxpayer’s share only grows.

Upon diving deeper into the deal, to be fair to the weary taxpayer, the headline here should have read something like “Industrial policy wearing a uniform.” Because, clearly, the Pentagon is the first customer, the first regulator, and the first insurer for a private industry, using military bases and military demand as the proving ground so that five companies can one day sell reactors to data centers and utilities. In this scenario, it is impossible not to notice that the Janus project is a taxpayer-funded launchpad for a commercial product line, cleverly dressed up as a defense procurement.

Who does this taxpayer funded launchpad actually lift? Two of the five companies involved in Janus are venture-backed startups with familiar investors. Radiant has raised more than half a billion dollars from Andreessen Horowitz, Peter Theil’s Founders Fund, and, in a twist, the venture arm of Chevron. Antares raised $470 million this summer from a similar crowd. And the wider microreactor rush, driven by the bottomless appetite for power of the AI data centers, runs straight back into the same defense-tech world we’ve been watching closely. Indeed, backers of one of the hottest startups in the field, Valar Atomics, include Palmer Luckey of Anduril and Shyam Sankar, the Palantir chief technology officer that the Army recently commissioned as a lieutenant colonel in its Detachment 201 tech-executive corps. Sure, Valar was not awarded a spot in the Janus project. Nonetheless, it is currently suing the Nuclear Regulatory Commission to loosen the very licensing rules that Janus is conveniently sidestepping. Again, the taxpayer absorbs the early losses, and when it is time for profits, they will flow to the funds that were already betting on it and making it happen.

The regulator aspect is another cause for concern because these reactors will not be licensed by the independent civilian body that has made up the Nuclear Regulatory Commission for half a century. Instead, they will be licensed by the Army itself, through an Army office called the Army Reactor Regulatory Office, currently run by a lieutenant colonel, which is legal under the Atomic Energy Act. Either way, this setup is one that we’ve seen before – the same department writing the checks is also writing the safety case and granting the approvals. The buyer, the funder, and the safety inspector are the same institution. Waksman has promised a “straight line” from Army approval to eventual NRC approval so the vendors can sell outside of the base down the road. Maybe. But until that happens, the safety sign-off is being handled in-house by the customer. What could possibly go wrong?

Well, let’s not forget about the waste, which is the quiet tell. Every Janus reactor runs for a limited time and then shuts down. Within two years of that time, all of its spent nuclear fuel, the used-up radioactive core, must be off the base. The Army has clearly stated that it has no intention of building its own waste facility. Instead, the Department of Energy will take that spent fuel and assume legal title to it, for a fee. Unfortunately, this is the oldest arrangement in the nuclear business: the private company touts the clean, modern front end, and the government quietly owns the radioactive back end forever. The massive profits are private, and the dangerous waste belongs to the taxpayer.

And the 2028 deadline framing the entire project exists thanks to a May 2025 executive order that requires the Army to have a reactor running at a military installation no later than September 30, 2028. That same order specified that the reactor be “regulated by the United States Army,” which indicates that the in-house licensing arrangement was not a bureaucratic accident. It was written into the directive from the top. Waksman openly conceded that four of the five vendors will not make the date. “We do not expect all five of these companies to turn on a reactor in 2028. In fact, they definitely will not,” he said. The portfolio is built so that one design can hit the political deadline and save it, while the rest slip. Makes one wonder which vendor will become the photo op.

None of this is a scandal, and it should not be read here as one. The technology is real and the fuel is not weapons-grade. Likewise, the reactor designs are built to shut themselves down without outside power, and hardened power on a base that an enemy could otherwise switch off in a war is a genuine strategic advantage. Either way, as the Army funds an industry into existence, questions remain. Is the readiness of American soldiers the reason for this program, or the cover story for it? And if it is the cover story, is that OK if the story ultimately benefits the private company more than the American taxpayer?