China built the world's largest fusion magnet, on time and at home. The burning plasma it's meant to hold is the part still running late.

The most impressive thing China built this summer is a magnet, and the most honest way to describe it is to say exactly what it is and exactly what it is not. It is a superconducting coil the shape of a capital D, twenty-one metres tall, twelve wide, and heavier than a fully loaded 747 — 582 tonnes of niobium-tin wire and specialty steel, cooled to a fraction of a degree above absolute zero, wound to carry a hundred thousand amperes and pin a magnetic field of 6.5 tesla in empty space. Its makers at the Institute of Plasma Physics in Hefei say every component, from the superconducting strand to the structural steel that keeps it from tearing itself apart, was produced in China. They finished it in about six years, tested it at full parameters, and announced that it works. All of that is true, and all of it is genuinely remarkable.

Here is what it is not. It is not a burning plasma. It is not net energy. It is not, by itself, a reactor, or electricity, or a solved problem. It is one of sixteen coils that will eventually ring a machine called BEST — the Burning Plasma Experimental Superconducting Tokamak — and the machine does not exist yet. The magnet is the cage. The hard part of fusion was never the cage.

I want to hold both of those thoughts at once, because the distance between them is the entire story, and it is the distance the coverage keeps collapsing. "China completes the world's largest fusion magnet" is a fair headline. "China nears its artificial sun" is a different claim, and the magnet does not support it.

What the coil actually does

A tokamak confines a fusion plasma — a gas of hydrogen isotopes heated past a hundred million degrees, hotter than the core of the sun — by wrapping it in a magnetic field so that it never touches a wall it would instantly destroy. There is no material that survives contact with that plasma, so the plasma is held in mid-air by force. The toroidal field coils are the part that does the holding: sixteen enormous D-shaped magnets stood in a circle, each one bending the field around the doughnut so the charged particles spiral along the field lines instead of escaping. The coil China just tested is one of those sixteen.

To make a field that strong and hold it steady, the coil has to be superconducting, which means it has to be cold — the reported operating temperature is around minus 269 degrees Celsius, roughly four degrees above absolute zero, colder than the space between stars. At that temperature the winding carries current with almost no resistance; the reported internal joint resistance is 0.04 nano-ohms, a number that means, in practice, that once you push the current in it stays there without the coil heating up and quenching. That is the engineering achievement, and it is not a small one. A magnet this size stores an immense amount of energy in its field, and everything about its construction — the steel case, the insulation, the way the joints are made — is a fight against the forces trying to release that energy all at once.

So when I say the magnet is real, I mean it in the most literal way. This is not a simulation or a rendering or a target on a roadmap. It is a physical object that passed its tests. In a field that runs on announcements, a piece of hardware that does what it was built to do is worth pausing on, and worth crediting plainly.

• Dimensions: a D-shaped coil roughly 21 m tall by 12 m wide, about 582 tonnes.
• Field and current: designed for around 6.5 tesla at the plasma centre, carrying about 100 kiloamperes.
• Stored energy: reported at roughly three times that of the equivalent toroidal-field coil built for ITER, in about 1.3 times the volume.
• Temperature: superconducting operation near minus 269 °C, about four degrees above absolute zero.
• Provenance: components reported as entirely domestically produced, over a six-year build that generated dozens of patents.
• Role: one of sixteen such coils intended for BEST, a tokamak still under construction in Hefei.

Demonstrated, and announced

My father, who taught physics, had a rule that arrived before I had the words for it: an experiment you cannot repeat is a story, not a finding. I would add a companion rule for fusion specifically. A component that passed a test is a demonstration. A machine that has not been built, doing a thing no machine has yet done, is an announcement. The two get printed in the same font, and the reader is left to sort them out.

What was demonstrated in Hefei is manufacturing. A very large, very cold, very precise magnet was built and met its specifications, and a companion central-solenoid coil was tested above its rated current. That is a supply-chain and engineering result, and it is the kind of result that is checkable, which is exactly why it deserves respect. What was announced, alongside it, is a chain of future events: BEST completed by the end of 2027; BEST achieving a burning plasma with an energy gain of around five, meaning five times more fusion energy out than heating power in; fusion power generation demonstrated around 2030; and behind BEST, a larger machine, CFETR, meant to put fusion electricity on the grid by roughly 2035. Those are goals. None of them has happened. The magnet does not make them more likely to happen on time; it makes one prerequisite ready earlier than the rest.

The magnet is ahead of schedule. The sun it is meant to cage is not. — On the gap between a finished component and an unlit fire

This matters because the coil is the part of a tokamak we are actually good at. Big superconducting magnets are hard, but they are a known kind of hard — an engineering problem with a supply chain, a tolerance budget, and a test you can pass. The burning plasma is a different kind of hard. No tokamak anywhere has yet held a plasma that heats itself, at the density and confinement time and duration you need, long enough and cleanly enough to call it a power source. That is the problem BEST is being built to attack, and it is unsolved everywhere, by everyone, at every funding level. A finished magnet gets you a better-equipped attempt. It does not tell you the attempt will succeed, or when.

The number just past the money

BEST's stated goal is an energy gain — the physicists write it as Q — of about five. It is worth being precise about what that number is and is not. Q is the ratio of fusion power produced to the heating power delivered to the plasma. A Q of five would be a landmark: only a handful of experiments have ever exceeded Q of one for an instant, and none has done it in the kind of self-sustaining, repeatable way a power plant would need. But BEST's five is a design target for a machine that is still being assembled. It is the number the program is aiming at, not a number it has hit. In fusion, the space between the aimed-at Q and the achieved Q is where two decades of the field's history live.

And there is a further number the excitement tends to skip. Even a Q of five in a pulse is not electricity. To make power you need a burning plasma that runs continuously, not in shots; you need to breed your own tritium fuel inside the machine, because the world does not have enough of it to run reactors any other way, which means a tritium breeding ratio above one — a thing no fusion device has ever demonstrated; and you need materials that survive years of neutron bombardment without becoming brittle or radioactive waste. BEST is a step toward those problems. CFETR, the machine after it, is where China proposes to actually confront them, at a scale and on a timeline — grid electricity by 2035 — that would be the most aggressive in the history of the field if it holds.

On what timescale

I keep a private archive of fusion's promised dates and the years they were actually made. "Within ten years" has been roughly ten years away for most of my life. "By 2030" is a phrase I have watched migrate quietly toward "by the mid-2030s" and then toward "within a generation," each slide small enough to pass without comment. I mention the archive not to be cynical — the science is real and worth the patience — but because the moving goalpost is the specific thing honest coverage is supposed to catch, and it is easiest to catch on a good-news day.

So here is the timescale, stated plainly. The magnet is, if anything, ahead of schedule — an on-time or early piece of a machine due in 2027. The machine's physics goals sit in 2030. The grid-electricity goal sits in 2035, and it belongs to a reactor that has not begun full construction. Every one of those later dates depends not on building a better magnet but on solving the plasma, the breeding, and the materials — the parts that have humbled every program that has reached them. It would be a mistake to read an early magnet as evidence the later dates will hold. Components are where fusion tends to be on time. Burning plasmas are where it is late.

What the 'made entirely at home' part means, and what it doesn't

The line the program is proudest of is that the coil is fully domestic — the wire, the steel, the insulation, all Chinese-made, the director saying the work has "completely broken foreign technological monopolies." That claim is industrial, not physical, and it is worth taking seriously on its own terms. Being able to build these magnets at home, without waiting on an export licence or a foreign supplier, is a real strategic advantage: it lets a program iterate faster and scale without permission. It is the same logic that has driven so much of China's push across the chip and battery supply chains, applied to fusion hardware.

But it accelerates the engineering, not the physics. A domestic supply chain can get you sixteen coils faster and cheaper. It cannot tell a plasma to stay confined, or make tritium breed at a ratio above one, or make a first wall survive a neutron flux. Sovereignty over the cage is a genuine achievement and a genuine advantage. It is not sovereignty over the fire, because no one has that yet.

It is also worth setting the international frame honestly, because the comparison is being made for us. ITER, the thirty-five-nation reactor in France, is the machine whose toroidal-field coils this one is measured against, and ITER is famously slow — first plasma not expected until around 2034, decades into the project. That China built a larger, higher-energy coil domestically in about six years is a real contrast, and it says something true about the difference between a consortium and a national program in a hurry. What it does not say is that China has leapt ahead in the science, because ITER, too, is a research machine that has not yet produced a burning plasma. Both are still upstream of the hard part. One of them just got its magnets faster.

The honest version

There is a version of this story that is pure wonder, and it is not wrong to feel it. A 582-tonne object, colder than the space between stars, built without a single imported part, designed to hold a fragment of star in mid-air — that is one of the more extraordinary things human beings make, and the people who made it deserve to be told so. Fusion is worth the wonder, and the wonder is not the problem. Credulity is.

So the honest version keeps both halves. China has built, on time and at home, the best cage yet made for a fire the world has not yet kept lit long enough to matter. The cage is finished and real. The fire is a goal with a date on it, and the date is the part I would watch, not because I doubt the science but because I have watched this field spend the public's patience by conflating the coil with the plasma before. When BEST holds a burning plasma — actually holds one, at gain, repeatably — that will be the story, and it will be a bigger one than this. Until then, the correct sentence is the modest one: they built the magnet. That is a lot. It is also not the sun.