The magnets are in. Now comes the plasma.
SPARC is about 80 per cent assembled and aiming at first plasma in 2027. The bet it represents is narrower, and more testable, than the word fusion suggests.
As of late July 2026, Commonwealth Fusion Systems reported SPARC roughly 80 per cent assembled at its site in Devens, Massachusetts, with all eighteen toroidal-field magnets expected in place by the end of the summer. First plasma is targeted for 2027, and with it the claim the company has organised itself around: that SPARC will be the first machine to get more fusion energy out of a plasma than is put into it.
Why the machine is small
The reason SPARC is the size of a tennis court rather than a stadium comes down to one material choice. Its magnets are wound from REBCO — rare-earth barium copper oxide — a high-temperature superconducting tape that carries useful current in fields where conventional niobium-tin magnets quench. SPARC's coils are designed for about 20 tesla, roughly double what the previous generation could hold.
That doubling matters more than it sounds, because tokamak fusion performance scales approximately as the fourth power of the magnetic field. Two times the field is about sixteen times the performance at fixed size — which can be traded back for a machine of about one fortieth ITER's plasma volume chasing comparable physics. SPARC is not a different theory of fusion. It is the same theory, bought with better wire.
The money followed
In July 2026 Commonwealth raised a billion dollars in a round that included pension funds — an industry first, and a signal about which side of the risk curve fusion is now thought to sit on. Total capital raised reached roughly four billion dollars, on the order of thirty per cent of all private investment in fusion worldwide. Concentration on that scale is its own kind of experiment.
What 2027 will and will not settle
Q greater than 1 is a scientific gain threshold: fusion power out versus heating power into the plasma. It excludes the electricity that runs the cryoplant, the magnets, the pumps and the control systems. A machine can pass it and still be a very large net consumer of power, and SPARC — which is not designed to generate electricity at all — will be exactly that. The plant that follows, ARC, is where the question of a grid connection is supposed to be answered.
What first plasma and the campaign after it will settle is narrower and more valuable: whether high-field, compact tokamaks behave the way the scaling laws say they should when you actually build one. Confinement in a small, very high field device has never been measured. Neither has the behaviour of REBCO magnets under years of neutron flux, or tritium breeding at any practical rate, or the survival of first-wall materials facing a burning plasma. Those are the load-bearing unknowns.
Meanwhile the reference machine has moved further away: ITER now expects deuterium–tritium operation to begin in 2039, four years later than its previous plan. The honest state of play in 2026 is that no fusion plant anywhere has delivered electricity to a grid, no company is within a year of doing so, and the early aggressive milestones have all slipped. What has genuinely changed is that the magnets exist, the capital exists, and a falsifiable test is now about eighteen months out.