The fusion startup that raised $100M by promising 2034

The most truthful number in Thea Energy’s announcement last week was not the $100 million. It was 2034. That is the year the company says its first commercial fusion plant, Helios, could come online — eight years out, a decade and change from the founding, and notably not the round number everyone in this field reached for through the 2010s. For most of my adult life the answer to “when will fusion arrive” has been a horizon that recedes at exactly the speed you approach it: always about ten years away, always 2030-something, always just past the next raise. Thea putting a date eight years out, and a demonstration machine in front of it, is a small act of specificity in an industry that has historically preferred the comfortable vagueness of the verb “soon.”

On May 27, the Kearny, New Jersey company said it had closed an oversubscribed $100 million Series B led by Thomas Tull’s U.S. Innovative Technology Fund, bringing total private investment to roughly $130 million. That is real money, and it arrives at a moment when not every fusion company can count on the next cheque landing. The wonder here is genuine. So is the need to read the press release the way Thea reads its own plasma: carefully, and with the instruments calibrated to tell demonstrated from announced.

What a stellarator is, and why Thea’s is unusual

Start with the physics, because it is the part that is actually impressive. A stellarator, like its better-known cousin the tokamak, confines a superheated plasma — hydrogen isotopes stripped to ions and electrons, hotter than the core of the sun — inside a twisting magnetic cage so that nuclei can collide and fuse. The tokamak holds that cage partly by driving an enormous electrical current through the plasma itself, which works well right up until the current goes unstable and the whole thing disrupts. The stellarator does something harder and, in principle, calmer: it shapes the entire confining field with external magnets, so the plasma can in theory sit in steady state without that internal current and the disruptions that come with it. The catch, and it is the catch that has dogged stellarators since the 1950s, is that those external magnets have to be fabricated into baroque three-dimensional shapes, each one a custom sculpture, expensive and slow to build.

Thea’s bet is to refuse the sculpture. Instead of a few exquisitely warped coils, it uses an array of flat, mass-manufacturable “planar” magnets — twelve larger ones for the bulk confinement and more than three hundred smaller ones for fine adjustment — and then shapes the field in software rather than in metal. The company likes to compare the magnets to pixels on a screen: individually simple, collectively capable of drawing a complicated picture. It is a genuinely clever inversion. It moves the hardest part of the problem out of the machine shop and into code, where iteration is cheap and a mistake is a recompile rather than a recast.

What has been demonstrated

This is the line that matters, so let me draw it precisely. Thea has built and operated what it describes as the world’s first superconducting planar-magnet array capable of producing the complex shaped fields a commercial stellarator would need — a prototype it calls Canis. In testing, the team deliberately installed magnets out of alignment and then let the software correct for the error, which is exactly the kind of result that should reassure you: it shows the control scheme degrades gracefully, that real-world sloppiness can be compensated rather than fatal. That is a meaningful engineering demonstration. It is the part of the company’s story that has moved from announced to done.

Here is what has not happened, at Thea or anywhere else among the private startups. No one has shaped and held a fusion plasma in one of these machines and gotten more energy out than they put in. Thea’s array has been tested as a magnet system, not yet wrapped around a burning plasma. The company has not claimed scientific breakeven — the point where the fusion reactions release more energy than was needed to ignite them — because it has not run the experiment that could claim it. The Eos device that the $100 million is partly meant to build is the machine intended to integrate the magnets with an actual plasma and chase “power-plant-relevant, steady-state” operation. Its site has not been selected; that is slated for later this year. Eos is the proof. Right now it is a line item and a target date: completion around 2030.

A magnet array that corrects its own misalignment is a real result. A power plant that does it reliably for thirty years is a different problem by orders of magnitude — most of them engineering, not physics.

What the money is buying

Read the use of funds and the round tells you what stage Thea thinks it is at. The $100 million goes to three things: a second magnet-manufacturing facility in northern New Jersey, siting and starting construction on Eos, and roughly doubling the headcount. Notice what that list is and is not. It is a manufacturing-and-scale-up budget — factories, sites, people — not a basic-physics budget. Thea is spending as a company that believes the science of its magnets is largely de-risked and the remaining work is industrial: can you build these arrays by the hundred, cheaply and repeatably, and stand up the facilities to house a real machine. That is a confident posture, and the federal scaffolding underneath it is unusually solid: the company has taken six Department of Energy INFUSE awards and says it was the first to clear a DOE certification milestone for Helios’s preconceptual design.

So the honest version is this. The Series B buys time, factories, and a workforce to attempt the experiment that would actually validate the approach. It does not buy the result. The gap between “we built the magnets and they behave” and “we held a plasma and it gave back more than we fed it” is the entire remaining distance, and it is the distance every fusion company has found longer than its slide deck suggested.

The boom, and the cracks in it

Thea’s raise lands against a backdrop that is harder to read than a single happy headline. By the Fusion Industry Association’s most recent tally, private fusion companies had pulled in about $2.64 billion in the twelve months to July 2025 — the second-biggest year the sector has ever recorded — bringing cumulative investment across the roughly fifty-three surveyed companies to about $9.8 billion. The field has gone from a couple of dozen firms to more than fifty in four years. Money has not been the problem. Money has, if anything, been the story.

But in April, TechCrunch reported what a lot of people in the field had been saying privately: cracks are forming. Two of the older, better-known names, TAE Technologies and General Fusion, have moved to go public through mergers with listed shells — TAE via Trump Media & Technology Group, General Fusion via a reverse merger — and neither has hit scientific breakeven, the very milestone many in the industry think a company ought to clear before asking public markets to price it. General Fusion cut around a quarter of its staff in 2025 and needed a $22 million lifeline to keep going. Another startup, Zap Energy, recently hedged by folding nuclear fission into its plans. The disagreements are revealing: when is it legitimate to go public, and are the side businesses some companies are building — Commonwealth Fusion Systems and Tokamak Energy selling magnets, TAE and Shine in nuclear medicine — prudent revenue or a quiet admission that the reactor is further off than the timeline says?

None of this is a verdict against fusion. It is what a maturing field looks like when the easy capital meets the hard physics and the two discover they are on different schedules. The companies that diversify are buying themselves runway because they suspect the runway needs to be long. That suspicion is, scientifically, correct.

On what timescale

Which returns us to 2034, and to the question I always end up asking out loud. Is it real? It is real as an intention and unproven as a fact, and the difference is the whole of the matter. To make Helios light in 2034, Thea needs Eos to work around 2030 — a machine that does not yet have a site — and Eos needs the magnet array, which exists and behaves, to do something it has not yet done, which is govern a burning plasma at power-plant-relevant conditions. Each of those is a real step on a credible path. None of them has been taken. The schedule is a chain of contingent achievements presented as a calendar.

I keep a private list of every “fusion by [year]” promise and the year it actually turned out to be, and the list is long and the gap is always wider than advertised. So I will say what I think is true and let the company prove me happily wrong: Thea Energy has done the most respectable thing a fusion startup can do at this stage, which is build a real piece of hardware that works and then attach its biggest claims to a machine it has not built yet. That is not the same as having solved fusion. It is the honest shape of a field worth taking seriously — remarkable and unfinished, both at once. The $100 million does not collapse that distance. It just pays for the next attempt to cross it. Watch what Eos does. Not what 2034 promises.