Small Nuclear Is the Renaissance

For fifty years, nuclear power in America meant one thing, an enormous plant that took fifteen years and tens of billions of dollars to build. That era is ending. The market has decided the future of nuclear is small, reactors the size of a shipping container that a factory can build and a truck can deliver, and the government has started buying them. The Army went first. In August, under a program it calls Janus, it named five companies to build these reactors on its bases. Three are the names that have run nuclear for half a century: Westinghouse, General Atomics, and BWXT. The other two, Radiant and Antares, did not exist a decade ago. Both are venture-backed. Both are now building nuclear power for the United States military.

This is what people mean by a nuclear renaissance, but the word hides the real story. The old nuclear, the gigawatt plant, did not simply fall out of fashion. The partial meltdown at Three Mile Island hardened the politics, cheap natural gas broke the economics, and America all but stopped building large reactors for a generation. What brought nuclear back is not nostalgia. It is demand. AI has turned electricity into the scarce input for the most valuable companies on earth, and they cannot wait that long. A small reactor can be running years sooner. That is the renaissance, and the nuclear coming back is not the nuclear that left.

Here is the scale of that demand, because it is the whole story. The International Energy Agency expects the world's data centers to nearly double their electricity use by 2030, to around 945 terawatt-hours, close to what Japan uses in a year, and to triple in the United States alone. The grid cannot be expanded fast enough to catch that, so the buyers stopped waiting. Microsoft is bringing Three Mile Island, the very site that scared America off nuclear, back online. Amazon took a stake in the reactor startup X-energy. Google contracted Kairos for reactors that do not yet exist. Meta went looking for gigawatts of its own. The power they have lined up already dwarfs anything the military is building. Janus is not the demand in this story. The hyperscalers are. Janus matters because the most cautious buyer on earth reached the same conclusion, and picked the same small reactors to do it.

So look at what the Army actually committed to. Janus is a $2.2 billion program to put more than twenty microreactors, reactors small enough to truck to a base and run it on their own, across five installations, with the first units due online in 2028. The military reason is resilience: a reactor inside the fence line keeps the lights, the runways, and the command systems on even when the civilian grid goes dark. But the part that matters here is who it chose.

Two of the five are venture-backed startups: Radiant, founded in 2019, and Antares, younger still, neither of which had sold a commercial reactor before this. Three are incumbents who have built reactors since the Cold War. None of them brought a big plant; even Westinghouse, the name on thousand-megawatt reactors, came with a five-megawatt one. That the incumbents made the list is no surprise. That Radiant and Antares made it is the whole story.

So how does a company a fraction of the size win a seat next to firms with fifty years of reactor history? They treated the deadline as the product. An incumbent optimizes for the program: the cost-plus contract, the timeline that stretches to fill the budget. A startup optimizes for the ship date, because a startup that misses one dies. Antares had already taken a reactor critical before most competitors had metal in a room. Radiant built for transport from the first sketch. Neither won on lobbying or legacy. When the Army finally wanted speed, the companies built to move were already moving.

Here is how I read it, and it is not a stock tip. The two names that matter are private, and this is commentary, not a recommendation. Deep tech asks investors to underwrite two unknowns at once: will the science work, and will anyone buy it. This sector has answered the second one. The demand is not in doubt anymore, it is only a question of who captures it. A signed government contract is how you tell. It is the difference between a company the market believes will have customers and a company that already does, on a dated purchase order. That is why a Janus award is worth more than its dollar figure. It re-rates everything behind it. The signal is not the size of the contract. It is a venture-backed name appearing where only incumbents used to stand. Antares and Radiant did that this summer.

The Army was the first buyer. It will not be the last, and it is nowhere close to the biggest. The demand is certain, the first reactors are still years from switching on at scale, and the companies that will own this decade are competing fiercely now. Big nuclear had its century. This one goes to whoever builds small, fast, and now. AI will buy every megawatt they can make.

Bashar Aboudaoud
Managing Member, UpRound

Live Webinar: Thursday, October 15: You leave with three things. Where private valuations actually sit right now. Which sectors are repricing fastest. How I am reading prices into 2027. 20 minutes, then live Q&A. 20 slots only for UpRound readers.

Note: We have rescheduled the webinar from September 17th to October 15th to align with financial reporting cycles.

Share with 10 people to receive a quarterly call with Bashar.