The fusion scoreboard

In January, Commonwealth Fusion Systems lowered the first of 18 superconducting magnets into SPARC, the doughnut-shaped reactor it is building outside Boston. That is the real news, and it is a physical one: the entire bet rests on these magnets. CFS proved in 2021 that a tape of high-temperature superconductor could hold a 20-tesla field, then designed a machine a fraction of the size of the giant international ITER reactor around that single result. Roughly three-quarters of SPARC is now assembled; the company expects all 18 magnets in by the end of summer.

"Aiming for first plasma in 2027 and then getting a Q greater than one as fast as humanly possible." — Brandon Sorbom, CFS co-founder, April 2026

The harder thing to pin down is the date. The pitch is that SPARC will be the first commercially relevant design to get more energy out of fusion than it puts in — Q greater than one. CFS's site says first plasma in 2026 and net energy in 2027. But the original 2018 plan said 2025, and by early 2026 the people running the project were saying first plasma is now 2027. The net-energy milestone is receding faster than the calendar advances.

Two things make this more than another fusion promise. CFS has raised about $3 billion, and the latest round brought in Nvidia and Google — the AI giants now scrambling for the electricity their data centers will need. And Google has already signed to buy power from ARC, the commercial plant CFS wants to build in Virginia in the early 2030s, a plant whose predecessor has not yet made a single joule of net energy. The useful way to hold this is as a scoreboard: a dated, falsifiable claim with a documented habit of slipping, and a magnet going into a machine you can watch get built.