The Race To Make Fusion Pay

AFusion has been described, unkindly but not unfairly, as the technology that is always thirty years away. The physics is not in doubt: forcing light nuclei together releases energy, as it does in every star, and a kilogram of fusion fuel holds some ten million times the energy of a kilogram of coal. What has proved stubborn is the engineering. A commercial plant must hold a gas at over one hundred million degrees, draw the heat out of it, survive a flux of high-energy neutrons for decades, and do all of this at a cost per unit of electricity that a market will accept. For fifty years the work was done almost entirely by state consortia on a single timetable, of which the ITER project in southern France is the surviving monument. Since 2018 the picture has changed. More than forty private companies now promise electricity on the grid in the 2030s, and the argument about whether they can now matters more than the argument about whether fusion works.

BCecile Kowalczyk, a plasma physicist at the Nordvik Institute for Energy Systems, believes those promises rest on a bookkeeping error. The favoured reaction consumes deuterium, which is abundant in seawater, and tritium, which is not: tritium decays with a half-life of twelve years and exists on the planet in quantities measured in kilograms. A single one-gigawatt station would burn about 56 kilograms a year, against a world civilian stock of some 25 kilograms held mostly in ageing Canadian reactors and falling by roughly five per cent annually. Every design therefore proposes to make its own tritium in a lithium blanket wrapped around the plasma, and every design assumes a breeding ratio above 1.05, meaning that each nucleus burned is replaced with a small surplus. Kowalczyk's group built three blanket mock-ups between 2019 and 2024 and recorded ratios of 1.02 at best, and then only when the lithium had been enriched to a level no supplier currently produces at scale.

CHer claim is not that fusion is impossible but that the first station has no supply of fuel which does not depend on a second station, and the second cannot be built until the first has run. Kowalczyk calls this the start-up problem, and she is unusually explicit about what would refute it: a blanket module holding a ratio of 1.08 in a neutron environment resembling a real reactor, sustained for a full year. She adds a caution that is often dropped when she is quoted. Her figures were taken in accelerator-driven test rigs rather than inside an operating tokamak, and the difference in neutron spectrum could work in either direction. What she rejects is the habit of citing a ratio computed in a simulation as though it had been measured; of the fourteen company roadmaps her group reviewed in 2024, eleven rested on modelled figures alone.

DMarisol Ayala, an engineer at the Cape Ferrol Laboratory, treats the fuel argument as a question of sequence rather than a barrier. Her own field is magnets, and here the change since 2015 has been abrupt. Coils wound from rare-earth barium copper oxide tape reach magnetic fields that copper and older superconductors cannot, and because the power density of a confined plasma rises with the fourth power of the field, moving from five to twelve tesla shrinks the machine that yields a given output by a factor of tens rather than of two. A device that would once have filled ITER's hall now fits inside a building of ordinary size, which matters less for elegance than for iteration: cheap machines can be built, broken and rebuilt often enough to learn from. Ayala accepts that the tape is still expensive, that a full-sized station would need several thousand kilometres of it, and that nobody has yet run a coil of this kind for a decade in a neutron field.

EThe dispute reaches well beyond the laboratory. Private investment in fusion passed twelve billion dollars in 2024, most of it raised on schedules that assume the magnet story and not the fuel story. Regulators in three countries have chosen to license fusion devices under rules written for particle accelerators rather than for fission plants, a decision that turns on the absence of a chain reaction and of long-lived waste, and one they would revisit if tritium at commercial inventories proved hard to contain. Kowalczyk notes, awkwardly for her own case, that the start-up problem is an argument about the first few stations only and says nothing about whether a mature industry could sustain itself. Ayala replies that first stations are precisely what investors are buying, and that an industry which cannot start will never mature. Both dislike the habit of announcing a record held for a second and calling it progress.

FA more useful way of framing the question is emerging. Field strength governs whether a plasma can be made to burn at a size and a price that somebody will pay for; the fuel cycle governs whether a machine that burns can be kept supplied. The two are not rivals, and progress on one does not purchase progress on the other, which is why the compact devices now under construction are best read as a test of the first question and not of the second. What remains genuinely open is whether a blanket can be qualified within the time the financing assumes, and no facility now operating produces neutrons at the intensity and volume such a qualification would need. Both researchers agree that the answer will come from an unglamorous programme of materials engineering, not from another record shot, and that the thirty-year joke will end, if it ends at all, in a laboratory rather than on a grid.