Apollo Atomics found the one part of a reactor worth shrinking

For most of its history, nuclear power has had a scaling problem that runs backwards. In almost every other technology, the way to get cheaper is to get smaller and build more units. Nuclear went the other direction. Plants got larger, more custom, and more expensive with each generation, until a single reactor became a decade-long megaproject that only a utility or a government could finance.

That is the problem Apollo Atomics is trying to reverse, and its approach is unusual because of how little it tries to change.

Apollo builds a compact pressurized water reactor. The pressurized water reactor, or PWR, is not exotic. It is the design that already runs roughly 80% of the world's nuclear plants, with decades of operating history and a regulatory paper trail to match. Apollo's bet is that you do not need new physics or new fuel to make nuclear small. You need to fix the one component that forces the rest of the plant to be enormous. So that is the part they rebuilt, and they left almost everything else alone.

Why reactors got so big

To see what Apollo is doing, it helps to know where the size in a conventional reactor actually comes from.

A pressurized water reactor heats water under high pressure in the reactor core, then passes that hot water through a steam generator. The steam generator transfers the heat to a second water loop, which flashes to steam and spins a turbine. The core itself is compact. The steam generator is not. In a full-scale plant it is a massive, hand-built pressure vessel, one of the largest and most expensive single components in the entire facility, and everything around it has to be sized and built to match.

Because that component is so big and so bespoke, it drags the whole project into megaproject territory. It has to be forged by a handful of suppliers, shipped in pieces, and assembled on site over years. The construction timeline, the financing, the custom engineering, and the cost overruns that plague new nuclear all trace back in part to the fact that the core piece of equipment cannot be mass-produced. You are not building a product. You are building a one-off industrial structure around a giant vessel.

This is why the industry's answer to "make it cheaper" has usually been "make it bigger," to spread that fixed cost over more output. It is also why so many new reactor startups have gone the opposite way and thrown out the whole design, betting on molten salt, sodium coolant, or high-assay fuels that no regulator has licensed at scale. Those approaches are ambitious, and they carry the risk of ambition: new materials, new failure modes, and a first-of-a-kind licensing process that can take a decade on its own.

The one component they rebuilt

Apollo's insight is that the steam generator is the thing worth attacking, and that you can attack it without touching the rest.

The company's core technology is what it calls a Compact Steam Generator. It replaces the enormous conventional steam generator with a version that Apollo says is roughly 20 times smaller while delivering the same thermal power. Shrink that one component and the constraint it imposed on everything else falls away. The reactor no longer has to be a cathedral built around a giant vessel. Apollo describes the resulting plant as an order of magnitude smaller than existing ones, compact enough to be built in a factory and shipped on a truck.

That size change is not a spec-sheet detail. It is the difference between two entirely different businesses. A plant that has to be constructed on site is a decade-long project financed like infrastructure. A reactor that rolls off a production line is a product, with the cost curve, the lead times, and the quality control that come from building the same thing many times. Apollo says its target is a deployment time under 24 months, against the ten-plus years a conventional build can take, and it treats high uptime as a design goal rather than an afterthought.

The founders like to frame it plainly. They took the technology that already powers most of the world's reactors and flipped one part. Everything hard about nuclear that they could keep, they kept. The one thing that made it un-manufacturable, they redesigned.

Keeping everything the regulator already trusts

The most strategically interesting thing about Apollo is not the physics. It is what the physics lets them avoid.

Regulatory approval is the graveyard of nuclear startups. A genuinely novel reactor, with new fuel and new coolant, has to prove safety cases no regulator has seen before, which means years of review with no revenue on the other side. Apollo sidesteps most of that by staying inside the lines the Nuclear Regulatory Commission has already drawn. The reactor uses light water as coolant and low-enriched uranium as fuel, the same materials in the reactors operating today. It draws on components and a fuel supply chain that are already qualified. The novel part, the steam generator geometry, sits inside a design the regulator has licensed dozens of times.

That is a deliberate wager about where the real risk in nuclear lives. Apollo is betting that the binding constraint on new nuclear is not reactor physics but manufacturability and licensing, and that the fastest path to deployment is to innovate hard on the one component that gates cost while staying conservative about everything a regulator scrutinizes. It is a less glamorous story than a clean-sheet reactor. It is also a much shorter one.

Apollo has started putting that theory to the test in hardware rather than slideware. The company has built and run a demonstrator, and reports reaching criticality on a test assembly in mid-2026. The underlying work traces back to a research collaboration with MIT's Department of Nuclear Science and Engineering.

The customer that needs this now

Timing is doing a lot of work in Apollo's favor, and it has a name: the data center.

AI has turned electricity into a gating input for the largest technology companies in the world. Training and serving frontier models consumes enormous, around-the-clock power, and hyperscalers are finding that the grid cannot give them what they need where and when they need it. They want firm power, meaning power that runs 24 hours a day regardless of weather, sited close to the compute. That is a hard thing to buy right now, and it is close to a perfect description of what a small, factory-built reactor is meant to provide.

Apollo is pointing its roadmap at that demand directly. It describes reactors sized for data center operators and hyperscalers looking for firm power in roughly the 2028 to 2035 window, and has signaled that its early customers are likely to come from data center projects, with Pennsylvania mentioned as a probable first market. The pitch to that buyer is simple. You need reliable power on your own timeline, you cannot wait a decade, and you would rather buy a machine than commission a megaproject.

There is a useful alignment in this for a company that still has to prove itself. Data center operators are among the few customers with the balance sheets, the urgency, and the appetite to pay for firm power early, before a technology has decades of its own operating history. They give Apollo a way to reach commercial scale with a customer that genuinely wants the first units.

The founders

Nuclear is a field where credentials are not decoration. They decide whether anyone will let you build.

Apollo was co-founded by Assil Halimi and Drew Walker. Halimi, the CEO, finished a PhD in nuclear science and engineering at MIT in 2025, working on small modular reactor design and high-burnup fuel under Professor Koroush Shirvan. He has a stack of peer-reviewed papers and patents on reactor design and fuel to go with it, and before graduate school he worked as a reactor operator at Engie and as an advanced reactor designer at EDF, the largest nuclear operator in the world. That is an unusually direct line from operating and designing real reactors to trying to reinvent the economics of one.

Walker, the COO, comes at it from the other side. He previously co-founded Blue Innovations Group, an electric boat startup, and spent time working in the White House. The pairing is deliberate. One founder has spent his career inside the technical and regulatory world of nuclear; the other has built a hardware company and worked in government, the two arenas where a reactor company either succeeds or stalls.

Apollo is part of Y Combinator's 2026 batch and has raised early funding from Y Combinator, Orange Collective, and other investors backing the current wave of American nuclear.

What Apollo is really betting on

Step back, and Apollo Atomics is a wager about which kind of innovation the moment rewards.

The instinct in deep tech is to change as much as possible and build the most advanced version of the thing. Apollo is doing something more disciplined. It is changing the one variable that unlocks the business, the manufacturability of the plant, and holding constant everything that would slow it down with regulators and customers. If that instinct is right, the winner in the next era of nuclear will not be whoever designs the most novel reactor. It will be whoever ships a licensable one first, at a size a factory can produce and a customer can actually buy.

That is a real bet, and it is far from settled. Turning a demonstrator into a licensed commercial reactor is a long road, and the history of nuclear is full of promising designs that never got down it. But the shape of Apollo's approach is a genuinely different answer to why nuclear stayed expensive for fifty years. The reason was never that the physics was too big. It was that one part was, and nobody had made the whole thing small enough to build like a product. Apollo is trying to be the company that does.