Thorium Reactors Are About to Get VERY Serious

ELI5/TLDR

A Danish startup called Copenhagen Atomics is trying to make nuclear power boring. They have built a small thorium reactor that uses hot liquid salt instead of solid fuel rods, fits inside a shipping container, and costs about 50 million dollars plus 2 million a year. If it works, nuclear energy goes from being a government mega-project to something closer to a wind turbine in price. They have not turned one on yet. They are close.

The Full Story

The pitch: nuclear at gas-turbine prices

Big nuclear has a pricing problem. A small light-water reactor starts at around 5 billion dollars. Hinkley Point C in the UK is heading for 60 billion. Only governments can write those checks. Copenhagen Atomics wants to sell a reactor unit for 50 million plus 2 million a year for fuel, maintenance, and replacement parts. They quote 9 dollars per megawatt-hour for the heat their unit delivers. Most power plants today sit north of 100 per megawatt-hour all-in.

“What we deliver only cost $9 per megawatt hour. So there’s lots of room for paying for your license and your financing cost and your buying the piece of land.”

Worth noting: they sell heat, not electricity. The customer still has to buy land, get a license, build the plant, and bolt on a steam turbine. But if the hardware and fuel really come in at those numbers, the economics shift entirely. Nuclear stops being a public works project and becomes a product.

Thorium, in the most boring way possible

Thorium does not, on its own, sustain a chain reaction. Think of it like wet wood. Put it next to a campfire (uranium-235 or plutonium-239 splitting and spitting out neutrons) and it slowly transforms. The neutrons stick to thorium-232, which becomes thorium-233, which decays through a middle stage called protactinium-233, and lands as uranium-233. That final product is fissile. It burns.

So thorium is a breeder fuel. Feed it in, it catches neutrons leaking from the main reaction, and quietly becomes new fuel. You can keep doing this for decades. The consequence: roughly 100 times less mining than a uranium reactor for the same energy. That was the number that grabbed the founder’s attention.

The onion core

The reactor itself looks like an onion sliced in half. Imagine a stack of concentric shells. The center is filled with heavy water. Wrapped around it, a thin channel of molten fuel salt flows through. Around that, more heavy water. Around that, a “blanket” of molten salt containing thorium, circulating to catch the neutrons that leak out of the center.

A few pieces to hold in mind:

Fuel salt is lithium fluoride mixed with uranium tetrafluoride. It is liquid at operating temperature, roughly 600 to 700 degrees C. Being liquid is the whole point — you can pump it, filter it, and swap it.
Heavy water is water where the hydrogen has been replaced with a heavier isotope. It slows neutrons down so they stick to things instead of flying off. It runs at 80 degrees, ambient pressure. No steel pressure vessel needed.
Blanket salt is lithium fluoride with thorium tetrafluoride. Sits on the outside and soaks up leaking neutrons.

Instead of sliding control rods in and out, they raise and lower the heavy water level. More moderator in the core means more neutrons slow down enough to split uranium, which means more heat. Less water, less reaction. Simple.

And if any of the pumps stop, the whole reactor turns itself off. The pumps sit below the core. Cut the power and gravity drains everything into separate tanks — salt here, water there, thorium somewhere else. Once the critical geometry is gone, the chain reaction stops within seconds. No human decision required.

The walk-away reactor

Light-water reactors need operators with years of training. Copenhagen Atomics wants to ship a reactor with one button. Stop.

“There’s only one button. There’s a stop button. That’s the only button that they can touch… and the reactor will go subcritical within seconds.”

The logic is brutal and consistent. If anything goes outside its limit, the action is the same: shut off power, drain the liquids. No branching procedures, no operator judgment. You scale this to “tens of terawatts” — their stated goal — and you cannot train a hundred thousand nuclear engineers. So the reactor does not need them.

Traditional nuclear builds triple-redundant critical systems. Copenhagen Atomics took a different approach: more than 40 computers across the reactor, sensors spread across them. Lose a third, it still runs. A dozen components break, operation continues. More breakage than that, it shuts itself down. The reactor is sealed at the factory, tamperproof. When something fails beyond the redundancy budget, you swap the whole unit and ship the old one back.

The worst-case envelope

For licensing, they have taken a shortcut that is secretly very clever. Instead of proving every possible failure mode is contained, they postulated the worst thing they could imagine — the entire core disintegrating at full power on the last day of a five-year run, all the fuel salt and heavy water mixing and boiling inside the containment, producing steam and pressure. Then they designed the shielding and the outer building to hold that scenario within radiation dose limits.

If you can contain the apocalypse, every lesser failure is automatically covered. It is a mathematical envelope. License the worst case, everything smaller is inside.

The iteration game

Big state-backed fast-reactor programs — French Phoenix, Russian BN series, various Chinese efforts — have spent decades and billions and mostly failed to hit commercial performance. Copenhagen Atomics decided they could not beat those budgets. So they chose different technology (thermal spectrum, molten salt, thorium, easier to simulate) and a different rhythm.

Their Copenhagen facility has full-scale reactor mockups that have never gone critical. They run non-radioactive salts (“Fleck” — a cheap substitute with similar properties) through the same pumps, valves, heat exchangers, and sensors. No radioactivity means no nuclear license needed for iteration. They break things with angle grinders and rebuild them the next week.

“Where other people built one react unit every five or 10 years, we improve our stuff every day.”

The pumps are a good window into this. Generation one used lubrication bearings — solid, ran for two years. Generation two uses electromagnetic bearings. The rotor floats inside the pump, balanced by coils pulling on it from all sides thousands of times a second. No moving part touches another. No wear. Target lifetime: the full five-year reactor run with no maintenance.

Real numbers on what it costs to build one

The reactor core, pumps, heat exchangers, pipes, valves, electronics: about 6 million dollars in components. The fuel is where it gets spicy. You need roughly two and a half tons of 5%-enriched uranium, converted into uranium tetrafluoride, plus lithium enriched to pure lithium-7, plus thorium tetrafluoride for the blanket. Currently that fuel package runs around 30 million, including heavy water. Most of that is paying to set up supply chains that do not really exist yet — nobody in the Western hemisphere currently makes ton-scale enriched lithium-7. Part of their 100-million-dollar funding round is specifically to change that.

The ambition is a reactor a day out of a factory. The component count is lower than a modern car. The bottleneck is not engineering, it is licensing — you cannot run 3,000 parallel license applications. Eventually this only works with a “type license” model where a single approval covers a factory-produced design, not each individual unit.

Why not Europe

The founder is blunt. European steel is three times the Chinese price. Energy is expensive. Nuclear political will is thin. He expects to keep R&D in Copenhagen but relocate manufacturing somewhere the economics work — likely somewhere with surging energy demand. India and China have already embraced thorium. China is currently running a molten salt reactor with thorium in it. The US has thrown real money at a few molten salt startups but has been slower than hoped on regulatory support for this specific technology.

The first real criticality test is scheduled for the Paul Scherrer Institute in Switzerland — a 1-megawatt, one-month run, with PSI as the license applicant and Copenhagen Atomics supplying the hardware. That is the moment the engineering stops being theoretical.

Key Takeaways

Thorium does not burn by itself. It has to be placed next to fissile material where neutrons can transform it into uranium-233, which then sustains the chain reaction.
The “onion core” uses heavy water level — not control rods — to tune reactivity. Drop the water level and the reactor goes subcritical.
Every failure mode resolves the same way: cut the pumps, gravity drains the liquids into separate tanks, chain reaction stops in seconds.
The operator has one button: stop. Nuclear engineers are not required in the safety case.
Licensing uses a single extreme envelope — prove the whole core disintegrating at full power stays within dose limits, and every smaller failure is automatically covered.
Redundancy is achieved through 40+ computers and many sensors, not through triple-redundant critical paths. Lose a third and it still operates.
Iteration speed comes from running non-radioactive stand-in salts through full-scale reactor mockups. No license needed for hardware R&D.
Reactor unit cost target: 50 million upfront, 2 million per year, 9 dollars per megawatt-hour for delivered heat.
Thorium needs roughly 100 times less mining than uranium for equivalent energy.
First criticality test planned at Paul Scherrer Institute in Switzerland: 1 megawatt, one month, first real neutrons flying in a Copenhagen Atomics reactor.

Claude’s Take

This is a good interview because the host is a nuclear engineer and keeps asking the “yeah but what if” questions that journalist pieces skip. You hear the Copenhagen Atomics guy explain blockage scenarios, sensor credit, proliferation risk, and corrosion in actual detail.

Score 8. The technology is real — the hardware exists, is running, and the engineering approach is defensible. Molten salt reactors are not new. Oak Ridge ran one for five years in the 1960s, the MSRE, and its corrosion data is what gives Copenhagen Atomics confidence that fission products will not destroy their stainless steel. This is not a fusion-style “if we invent seven breakthroughs” story. The physics works. The materials work. The pumps pump.

The risks are regulatory and supply-chain, not scientific. Can they build ton-scale enriched lithium-7 production? Can Switzerland license the test reactor in reasonable time? Can a “one button” reactor get through a regulator whose entire institutional memory is built around hundreds of operators? Their business model assumes a type-license regime that does not yet exist in any Western country for this class of reactor.

The proliferation answer was honest and slightly uncomfortable. You could, in theory, chemically separate protactinium from the blanket salt and end up with pure uranium-233, which is weapons-usable. Their defense: the reactor is welded shut for five-year runs and subject to IAEA inspection during swaps. Anyone sophisticated enough to crack one open is already sophisticated enough to run a uranium enrichment program, which is cheaper. Probably true, still a little hand-wavy.

Marks against: “reactor a day” is aggressive marketing. Tens of terawatts is ambition, not forecast. And the Europe exit plan is a real signal — if your own continent cannot support your manufacturing economics, your scale path runs through other people’s regulators, which historically is where nuclear companies go to die. Still the most interesting small-modular-reactor story going. Worth watching for the PSI criticality test.