Thea Energy just got $20M from the feds to make fusion reactors cheaper to build

Thea Energy will spend $20 million in new federal money on its magnet factory, on the bet that fusion's commercial path runs through repeatable parts rather than plasma breakthroughs. The Department of Energy's ARPA-E awarded the funding to scale the company's high-temperature superconducting (HTS) magnet production, betting the same way: that the cost of fusion power can be attacked the way the data-center industry attacked server cost, by stripping a bespoke machine down to a few repeatable parts.

The grant, made under ARPA-E's SCALEUP program, targets the most expensive hardware in a magnetic-confinement fusion reactor: HTS magnets. HTS magnets are the coils that squeeze plasma hot enough for atoms to fuse. They are powerful, but the rare-earth and superconducting-tape supply chain is constrained, and each reactor historically needs thousands of unique coils, machined and assembled to the millimeter.

Thea is building a stellarator, a twisted, donut-shaped cousin of the more common tokamak that confines plasma with external magnetic fields. Stellarators avoid the tokamak's susceptibility to plasma disruptions, but they pay for that stability with geometry: a tokamak can be built from a small family of identical coils, while a stellarator's twisting field historically demands bespoke coil shapes across the whole machine. That is the manufacturing problem Thea is trying to dissolve.

The company's bet, previewed in its Helios power-plant design in December 2025, has two parts. First, the 12 largest magnets in the reactor will be built from just four templates, copies of a few master shapes, repeated and assembled. Second, more than 300 smaller magnets, arrayed around the reactor's periphery in what the company calls a "pixel" distribution, will all be identical, with each one individually steered by software rather than machined to a unique specification.

That software steering is the lever. By tuning each pixel-magnet's current independently, the company says it can compensate for small deviations in coil placement, which loosens the construction tolerances for the rest of the machine. The result, the company argues, is a stellarator that can be built with the repeatability of a server rack rather than the craftsmanship of a one-off scientific instrument.

Per TechCrunch's reporting on the award, the SCALEUP funding will let Thea add new magnet-manufacturing lines at its facility. Thea also announced the expansion on its own site, framing the program as a federal vote of confidence in its small-magnet architecture. The release was syndicated via GlobeNewswire.

The capital context is unusually heavy for an early-stage fusion startup. Thea closed a $100 million round in May 2026, on top of a $20 million Series A in 2024, putting it among the better-funded private fusion companies. The ARPA-E award is small relative to that private capital, but it functions differently: it is a federal signal that the manufacturing thesis is worth underwriting with public money, not just venture capital.

The thesis has obvious appeal, and obvious limits. Stellarators have a 70-year history of being harder to build than tokamaks, which is precisely why Thea's templating and software-tuning approach exists. HTS tape remains a constrained global supply, with most production concentrated in a handful of vendors. Thea says a commercial-scale fusion power plant is targeted for the mid-2040s, a timeline in line with peer fusion companies and roughly fifteen years out from any plant the grant would support manufacturing for.

The mid-2040s target is a company-stated goal, not an independently validated engineering schedule. ARPA-E's SCALEUP program is built to back technologies with credible paths to scale, but the program terms and milestones tied to this specific award have not been published, and the cost curve that Thea is targeting has not been independently benchmarked against tokamak magnet costs.

In the near term, the federal money buys a manufacturing line that can test the thesis. If the same four large-magnet templates and the same 300-plus identical pixel magnets can be built, assembled, and steered into a working plasma-confining field, the company's bet becomes a dataset rather than a deck slide. If they cannot, the stellarator's old reputation will be the more durable explanation.

The next read-out will be hardware, not physics. Thea says the new lines will come online as the magnets and software are validated against a working reactor, with milestones tied to magnet production rather than plasma performance. The 2040s commercial target is the prize; the SCALEUP award is a smaller bet that the factory, not the plasma, is the harder problem to solve.