A fusion startup is testing magnetic mirrors with salvaged MRI magnets
Victoria Shinder2 min readIEEE Spectrum reported on 23 September 2026 that Terra Fusion, a startup in Baltimore, is developing a reactor around magnetic mirrors — a confinement approach set aside for roughly half a century — and testing it on an experimental device at the University of Maryland.
The construction detail is the striking part. The CMFX experiment uses superconducting magnets upcycled from secondhand MRI scanners, alongside repurposed brewing equipment. The stated aim is a route to commercial fusion at a cost the mainstream approaches are not reaching.
Jason Cassibry, professor of mechanical and aerospace engineering at the University of Alabama in Huntsville and an adviser to several fusion startups, gives the trade-off in one line:
"Magnetic mirrors have the advantage of being simple. If a nation or company wants to be first and have the highest power density, they might go with a tokamak. But if they want something that's cheaper to build and easier to maintain, they might go with a mirror concept."
What it means
The interesting claim here is about capital, not physics. Magnetic mirrors were not abandoned because nobody understood them; they were abandoned because they leak plasma out of the ends, which is why tokamaks — closed loops with no ends — took the field. Nothing in this announcement repeals that. What changes is the cost of being wrong: a device assembled from salvaged MRI magnets can be built, measured and rebuilt on a budget where a tokamak programme can only be planned.
That matters for the grid in a narrow and unglamorous way. Fusion's timeline problem is not only technical — it is that each iteration of a mainstream design costs enough that very few happen. An approach that tolerates more iterations per unit of money is buying attempts, and attempts are what a field this far from delivery is short of.
⚠️ We are reporting an experiment, not a result. There is no performance figure here, no claim of net energy, and no connection to any grid. The interest is in the method and the price of trying it, and the article itself frames the company as racing to demonstrate something before better funded rivals.