The scary part of nuclear power has always come down to the same picture. A solid uranium core, water pumped around it to carry off the heat, and a thick containment dome wrapped around the lot in case the cooling ever stops. Take the water away and the worry is that the fuel overheats, melts, and ruins your whole decade. Nearly every power reactor running on Earth today is built on some version of that arrangement.<\/p>\n
One reactor is not. Out near the city of Wuwei, on the edge of the Gobi Desert in China’s Gansu Province, a small experimental machine called TMSR-LF1 uses molten salt instead of water, runs at ordinary atmospheric pressure, and last November became the first reactor anywhere to convert thorium into usable nuclear fuel while it was still running.<\/p>\n
The lab behind it, the Shanghai Institute of Applied Physics under the Chinese Academy of Sciences, says it is currently the only operating molten salt reactor on the planet loaded with thorium at all. It is also rated at just 2 megawatts, which is the kind of figure that tends to get lost in the excitement, and thorium turns out to be a far more stubborn fuel than a clean headline suggests.<\/p>\n
That makes it a real first. It also leaves it a long way from putting power on anyone’s grid. The distance between those two facts is most of the actual story.<\/p>\n
TMSR-LF1 sits at SINAP’s campus in Minqin County, just outside Wuwei, in a stretch of Gansu desert that China picked partly because the design does not need a river or a coastline to cool itself. Construction started in September 2018. The reactor reached first criticality, meaning a self-sustaining chain reaction, on October 11, 2023, and was running at full operation by June 2024. In October 2024 it managed something conventional plants cannot: it was refueled without being shut down first.<\/p>\n
That last part sounds minor and is not. A standard pressurized water reactor has to power down and have the top of its pressure vessel unbolted to swap solid fuel assemblies. TMSR-LF1 carries its fuel dissolved straight into the molten salt that also cools it, so fresh fuel can be stirred in and fission products skimmed off while the reactor keeps humming.<\/p>\n
According to World Nuclear News<\/a>, the salt is a lithium-beryllium fluoride mix, the fissile starter is low-enriched uranium kept under 20 percent, and the reactor holds roughly 50 kilograms of thorium. None of those numbers are large. This was never built to sell electricity. It was built to prove the chemistry works in a real machine instead of on paper.<\/p>\n Here is the catch that has kept thorium on the shelf for decades. Thorium is not actually a nuclear fuel. It is fertile, not fissile, which means a lump of it will sit there and refuse to sustain a chain reaction no matter how you arrange it. To get energy out, thorium-232 has to absorb a neutron, turn into thorium-233, then decay through protactinium-233 into uranium-233, which is the thing that actually fissions. That conversion only happens if you already have a fissile driver like uranium or plutonium in the mix supplying the neutrons.<\/p>\n So the real test was never whether you can put thorium in a reactor. It was whether you can run that breeding loop inside a molten salt machine and measure it. That is what SINAP says it pulled off. The institute reported the first thorium addition to the reactor in October 2024 and the first confirmed thorium-to-uranium conversion<\/a> on November 1, 2025, calling it the first international experimental data of its kind from inside an operating molten salt reactor.<\/p>\n The Chinese Academy of Sciences framed the result as confirming the technical feasibility of using thorium in this kind of system. The institute has not released detailed technical data on the conversion run, though, and the reactor’s conversion ratio sits around 0.1. This is a measured proof of concept, not a reactor breeding its own fuel at scale.<\/p>\n The reason anyone bothers with molten salt comes down to how these reactors are supposed to fail. In a conventional reactor the fuel is solid and the danger is that it gets hot enough to melt, which is the literal meaning of meltdown. In TMSR-LF1 the fuel is already a hot liquid by design, circulating as part of the salt, so there is no solid core left to melt in the first place.<\/p>\n Two other things follow from that. The salt runs at atmospheric pressure with no water in the core, which removes the high-pressure steam that forces conventional plants to build those enormous containment domes. And the whole class of reactor is built around a drain tank. A plug at the bottom of the vessel is kept frozen solid by active cooling, and the idea is that if the reactor overheats or simply loses power, that plug melts, the fuel salt drains down by gravity into a holding tank shaped so the chain reaction cannot continue, and the reactor shuts itself off with nobody touching it.<\/p>\n That is the pitch, and it is a good one. In places it also runs well ahead of the evidence. SINAP credits molten salt reactors with what it calls “inherent safety features,” waterless cooling, and atmospheric-pressure operation, and leaves it about there. The louder “this reactor cannot melt down” line comes mostly from enthusiasts and headline writers, not from the institute itself. TMSR-LF1 is a 2-megawatt research unit with no detailed operating data made public, so the meltdown-proof framing is best read as how the design is supposed to behave, not something independently stress-tested for anyone to check.<\/p>\n None of this chemistry is new, which is the part people tend to find surprising. The molten salt reactor was largely an American invention. Oak Ridge National Laboratory ran a Molten Salt Reactor Experiment from 1965 to 1969, proving you could operate a reactor on liquid fluoride salt at high temperature and ordinary pressure. Then the United States put its money behind solid-fuel uranium reactors instead, the program wound down, and the idea sat mostly dormant for half a century until China picked it back up.<\/p>\n Why did such an apparently attractive fuel get shelved? A handful of stubborn reasons, and they have not gone away. Thorium is cheap as a raw material but awkward to source on purpose, because it mostly comes up as a low-value byproduct of mining monazite for rare earth elements, the same supply chain that feeds the magnets inside EV motors and missile guidance systems<\/a>.<\/p>\n Building thorium reactors is capital-intensive precisely because almost nobody has commercial operating experience with them. The fuel needs a fissile driver to work at all. And as POWER magazine<\/a> has laid out, there is still no mature regulatory framework or fuel-fabrication supply chain for thorium anywhere in the world, which is a polite way of saying a developer has to invent the paperwork and the parts at the same time.<\/p>\n SINAP has been open about the roadmap, and it is a long one. The institute’s stated next step is a 100-megawatt-thermal demonstration reactor, which it is aiming to have running by 2035. Commercial thorium molten salt reactors, the kind meant to actually sell heat and produce hydrogen, are penciled in for somewhere around 2040. Director Dai Zhimin has put the institute’s name behind the 2035 demonstration target, and SINAP says it plans to work with major energy companies to build out the industrial supply chain for it.<\/p>\n The one number China is genuinely proud of is domestic content. SINAP’s deputy director said more than 90 percent of the reactor’s components are made in China, with the key parts fully localized and the supply chain entirely under Chinese control, which is the part Beijing cares about as much as the physics.<\/p>\n This is the same national push that produced a truck-mounted reactor China is pitching at AI data centers<\/a>, and it lands the same way: an impressive engineering claim where most of the numbers still come from the developer rather than an outside lab.<\/p>\n China is also not doing this alone, even if it is currently ahead on the molten salt route specifically. Denmark’s Copenhagen Atomics, the Dutch-French startup Thorizon, and Indonesia’s ThorCon, which got a 500-megawatt plant licensed in August 2025, are all chasing versions of molten salt or thorium reactors. In the United States, where the technology started, the same milestone has pushed thorium back into the conversation, with one industry proposal now circulating in Washington arguing the country has two or three years to act before it falls permanently behind China on this particular branch of nuclear tech.<\/p>\n That is a race better measured over the next decade than next quarter, and right now China is the one with a reactor actually running. Set it next to the government-lab fusion machines now being switched on<\/a>, which are even further from a working grid, and the desert reactor starts to look like the nearer-term bet of the two.<\/p>\n Strip away the breathless framing and what is left is still real. China took a reactor concept the United States abandoned in 1969, rebuilt it in a desert, and last year became the first to make thorium do useful work inside a running molten salt machine, something no other country can currently claim.<\/p>\n That is the achievement, and it is a solid one. What it is not is a finished answer. A 2-megawatt research reactor proving the chemistry is a different animal from a 100-megawatt plant feeding a grid, thorium has been the fuel of the future for longer than most of us have been alive, and the hard parts left are the unglamorous ones: cost, regulation, supply chains, and time. China has cleared the science. The engineering bill comes next.<\/p>\n Image credit: AFP\/Xinhua\/Lin Shanchuan<\/em><\/strong><\/p>\n","protected":false},"excerpt":{"rendered":" The scary part of nuclear power has always come down to the same picture. A solid uranium core, water pumped … <\/p>\nThorium Is the Genuinely Hard Part Here<\/h2>\n
Salt Instead of Water Changes the Safety Math<\/h2>\n
Thorium Has Been Sitting on the Shelf Since the 1960s<\/h2>\n
What China Says Comes Next, and Who Else Is in the Race<\/h2>\n