Thirty Years Away, Forever

The magnet arrives on a low-loader at four in the morning, because that is when the roads are empty and the police escort is cheapest. It weighs ten tonnes. It is a flattened ring of silvery tape wound into a block the size of a small car, and it cost more than most houses.

Inside the test hall, engineers spend eleven days cooling it to twenty degrees above absolute zero, then run current through it and stand well back. The field it produces reaches twenty tesla — roughly four hundred thousand times the Earth's magnetic field, strong enough that a steel spanner brought within several metres would become a projectile. It holds. Somebody takes a photograph of the readout. It ends up on a wall.

This is what progress in fusion looks like now: not a reactor, but a component, tested alone, doing what it was supposed to do.

---

The joke is older than most of the people working on it. Fusion is thirty years away, and always will be. It has been repeated since roughly 1970, it is unfair, and it is not entirely unfair, which is the interesting part.

The underlying physics has never been in doubt. Force two light nuclei close enough and they merge, and because the product weighs slightly less than the ingredients, the difference is released as energy. It is the process powering every star, it produces no long-lived radioactive waste of the kind produced by fission, and it cannot run away — the reaction is so difficult to sustain that any disturbance stops it.

The difficulty is entirely in the word close. Atomic nuclei are positively charged and repel one another with enormous force at short range. Overcoming that repulsion requires the fuel to reach roughly a hundred and fifty million degrees, ten times the temperature at the centre of the sun, at which point it is a plasma — a gas so hot that electrons have been stripped from nuclei — and there is no material that can contain it. Every surface it touches, it cools itself against and contaminates.

The dominant solution has been a magnetic bottle: a doughnut-shaped chamber in which magnetic fields confine a ring of plasma away from the walls. Plasma, however, is a fluid carrying current in a magnetic field, and it is spectacularly unstable — it kinks, it wobbles, it develops edge eruptions that dump megajoules onto a small patch of wall in milliseconds. Fifty years of this discipline consist largely of discovering and suppressing successive families of instability, each of which had to be found before it could be fixed.

The target is a triple product: density, multiplied by temperature, multiplied by the time the energy stays confined. Push all three high enough at once and the reaction sustains itself. Laboratory triple products improved roughly a hundred-thousand-fold between 1970 and 2000 — a rate comparable to anything in semiconductors — and then, as funding flattened and the field consolidated into one enormous international project, the curve substantially slowed.

---

Two results have recently changed the conversation, and both are routinely misreported.

In December 2022, a laser facility compressed a fuel capsule and obtained more energy out of the reaction than the laser had delivered into it — about three megajoules from two. This was genuine, historic, and the first time humans had achieved it. It was described almost everywhere as fusion producing more energy than it consumed.

It did not. The lasers that delivered two megajoules to the target drew something over three hundred megajoules from the electricity supply, because the system was built for physics experiments rather than efficiency. Scientific gain — energy out versus energy into the fuel — had exceeded one. Engineering gain, which is energy out versus energy off the grid, remained below one by a factor of roughly a hundred. Both numbers are real. Only the second one is a power station.

The second development is quieter and, in the view of most people working in it, more consequential. Superconducting tape operating at higher temperatures and carrying current in far stronger fields became manufacturable in quantity. The relevance is a scaling relation: power density in a confined plasma rises with roughly the fourth power of the magnetic field. Doubling the field is worth sixteen times the fusion power in a given volume — so a device with better magnets can be dramatically smaller for the same output.

Smaller is not merely cheaper. It is faster. A machine that takes four years to build can be iterated; a machine that takes twenty-five cannot, and a field that cannot iterate learns very slowly. That, more than any single physics result, is why private capital arrived in this sector after decades of ignoring it.

---

Enthusiasm should nonetheless be rationed, because three problems remain substantially unsolved and none of them is about plasma.

The first is fuel. The favoured reaction uses deuterium, which is abundant in seawater, and tritium, which is not. Tritium has a half-life of twelve years, exists in trace quantities, and the entire global civil inventory is measured in tens of kilograms. A commercial plant would consume a significant fraction of that annually. The intended solution is to breed it inside the reactor, by surrounding the plasma with lithium that captures escaping neutrons and yields tritium. The chemistry works. Breeding more than you consume, continuously, while extracting it from a solid or liquid blanket in a working power plant, has never been demonstrated at any scale, and a machine that cannot achieve it is not a power source.

The second is materials. The reaction emits neutrons of very high energy, unaffected by magnetic fields, which pass into the surrounding structure and knock atoms out of their lattice positions. Over a plant's life, every atom in the first wall is displaced many times over. Metals subjected to this swell, embrittle and change properties. No facility currently exists that can test candidate materials under the correct neutron spectrum at the required intensity, which means the materials for a commercial reactor cannot yet be qualified — only estimated.

The third is that nobody has built the rest of the power station: heat extraction, tritium handling, remote maintenance of intensely activated components, availability targets of ninety per cent when the internal wall needs periodic replacement. These are engineering problems rather than scientific ones, which people say as though it were reassuring. It is not. Engineering problems of this class routinely take longer than the physics did.

---

Back in the test hall, the magnet is warmed slowly over several days, because thermal shock would destroy it and it is worth more than the building.

The engineer who ran the test is asked the inevitable question and gives the answer everyone in the field now gives, which is more precise than the old joke and less satisfying.

"Net electricity on a grid? Not in the 2030s. Possibly the 2040s, if the breeding works. What I can tell you is that ten years ago I'd have said the magnets were the hard part, and this week they aren't."