Fusion in 2026: Helion's D-T plasma, ITER's delays and a $100M stellarator bet

Three fusion stories ran through 2026, and they do not point the same way. A private company claims an industry first. The world's largest publicly funded reactor is still absorbing cost overruns. A New Jersey startup argues the right machine has not been built yet.

Polaris runs on deuterium-tritium

Helion, based in Everett, Washington, announced on 13 February 2026 that its seventh-generation Polaris prototype had demonstrated measurable deuterium-tritium (D-T) fusion and reached plasma temperatures of 150 million degrees Celsius. The company calls both results firsts for the private fusion industry. Polaris began operating at the end of 2024, and Helion says that in January it became the first and currently only private fusion machine to run on D-T fuel.

The number matters because of where it sits. Helion says the industry treats 100 million degrees Celsius as the threshold plasma temperature for a commercially relevant fusion machine. Helion's sixth-generation Trenta prototype set that record. Polaris broke it. The company also says it was the first to win regulatory approval to possess and use tritium for the purpose of demonstrating fusion energy production.

D-T is not, however, the fuel Helion intends to burn commercially. The company says it will keep testing Polaris to reach optimal temperatures for deuterium-helium-3 fusion, which is what its commercial operations will use. Helion describes the D-T campaign as one step in the Polaris testing programme, meant to show the company can operate and scale across multiple fuels.

"We believe the surest path to commercializing fusion is building, learning and iterating as quickly as possible," said David Kirtley, co-founder and CEO of Helion. "We've built and operated seven prototypes, setting and exceeding more ambitious technical and engineering goals each time."

Two outside researchers are quoted in the same release. Jean Paul Allain, associate director for Fusion Energy Sciences in the Department of Energy's Office of Science, said the Polaris data, "including record-setting temperatures and gains from the fuel mix in their system, indicates strong progress." Ryan McBride, who has worked on inertial confinement fusion and pulsed power at Sandia National Laboratories and taught at the University of Michigan, said he had reviewed diagnostic data and that it was "exciting to see evidence of D-T fusion and temperatures exceeding 13 keV or 150 million degrees Celsius."

Alan Hoffman is quoted too. He is an FRC plasma specialist with more than 40 years in the field, and he traces the research line back to Los Alamos, the private company Math Sciences North West, and the Large S device at the University of Washington built with Department of Energy funding. The release does not describe any independent verification of the Polaris results. The claims rest on Helion's own diagnostics and on the researchers who say they reviewed them.

On the commercial side, Helion says it began building in July 2025 on the site of Orion, its first commercial machine, in Malaga, Washington. That plant, the company says, will deliver electricity from fusion to the grid for Microsoft.

ITER: $22 billion, and a heat shield that cracked

In southern France, the International Thermonuclear Experimental Reactor is a different kind of project. CNET, which visited the site, puts the estimated cost at $22 billion and describes a doughnut-shaped vacuum chamber designed to hold plasma 10 times hotter than the core of the Sun. Containing 150-million-degree plasma requires superconducting magnets kept just a few degrees above absolute zero, with a thin heat shield separating one of the hottest environments ever created from one of the coldest.

Cracks in the piping of that heat shield turned up in 2020, according to CNET, along with distortions caused by welding and disruption from the COVID-19 pandemic. The result was a years-long delay and an additional $5 billion for repairs. Private fusion startups have multiplied against that backdrop, several of them hoping to reach major milestones before ITER does.

ITER's defenders frame the overruns differently. Javier Artola, a scientist modelling ITER's plasma behaviour, told CNET: "This is a publicly funded project. It is the knowledge of the world." The argument is that publicly funded work de-risks the research and development commercial fusion needs, so private companies can place bigger bets. Every problem ITER solves is one fewer problem for them. More than 30 countries are party to the ITER agreement, and CNET reports that every member state will have access to the science that comes out of the project, with the construction itself developing a global fusion supply chain.

"That China and Russia were going to collaborate with the US and Europe, and add in Korea, India, and Japan, that's either genius or insane," said Laban Coblentz, ITER's chief strategic advisor.

CNET's reporter came away more hopeful than the project's reputation would suggest. Seeing the build first hand, he wrote, made him think fusion may be in its last decade as a distant dream. That is a judgement, not a data point.

Thea Energy bets on the stellarator

Thea Energy, based in Kearny, New Jersey, announced on 27 May 2026 that it had raised $100 million in Series B funding. Thomas Tull's US Innovative Technology Fund led the round, with participation from General Innovation Capital Partners, Linse Capital, Calm Ventures, Climate Capital, Divergent Capital, Emerald Technology Ventures, Gaingels, Idemitsu Kosan, Overlay Capital, Timescale Ventures and Whatif Ventures. Existing investors also took part, among them Alumni Ventures, Hitachi Ventures, Lowercarbon Capital, Mercator Partners, Orion Industrial Ventures, Prelude Ventures and Starlight Ventures. The company calls the round oversubscribed.

The money goes to magnet manufacturing, including a second facility in northern New Jersey, and to the siting and construction of Eos, a large-scale integrated stellarator. Thea says it will pick a site for Eos later this year and double its team. It also says it is in talks with more than a dozen power offtakers, hyperscalers and utility partners, and aims to start construction of its first Helios power plant before the end of the decade.

The technical case rests on a specific claim. Stellarators, the company argues, are inherently more stable than other magnetic confinement designs, but their complex three-dimensional magnets historically made them impractical to build. Thea says it has moved that complexity out of precision mechanical fabrication and into software-defined controls, using arrays of mass-manufacturable planar magnets. It also says it has built and operated the world's first superconducting magnet array capable of producing the fields commercial stellarator systems need, and that it has de-risked the full-scale planar shaping magnets for Eos.

"Our architecture is simpler to manufacture, faster to construct, and more tolerant of real-world conditions compared to all other approaches," said Brian Berzin, co-founder and CEO of Thea Energy.

There is a regulatory footnote. Thea says the Department of Energy certified its Helios preconceptual design milestone, making it the first awardee to receive that distinction. The company spun out of Princeton University and Princeton Plasma Physics Laboratory in 2022, was an inaugural awardee of the DOE's Milestone-Based Fusion Development Program, and says it holds six DOE INFUSE awards.

What the three add up to

Helion reports a machine that ran on the fuel most fusion experiments use and hit a temperature record for the private sector. ITER is the large, slow, publicly funded attempt to solve problems once, for everyone. Thea is arguing about architecture, not just speed, and has raised money to build a machine that will test the argument.

None of the three has put fusion power on a grid at commercial scale. Helion's Orion is under construction and contracted to Microsoft. ITER is a research project. Thea's Helios is targeted for before the end of the decade. The milestones are real; the timelines remain claims.