Helion's 2028 Promise Runs on a Fuel It Has Not Yet Burned

Helion's Polaris prototype crossed 150 million degrees Celsius on February 13, setting a private-sector temperature record on a fuel its commercial machine will not use.

Helion's commercial target is 200 million degrees, what CEO David Kirtley calls the optimal sweet spot for a power plant. Polaris ran its February test on deuterium-tritium, not the deuterium-helium-3 its commercial plant will burn. At 150 million degrees, it is three-quarters of the way to the temperature ceiling on a fuel cycle the company is still developing.

Helion's February press release confirmed it will "continue testing to reach optimal temperatures for deuterium-helium-3 fusion."

Beyond fuel, Polaris has a third unfinished task: generating electricity directly from a plasma reaction. No fusion machine, public or private, has done that. Helion's Polaris page states it plainly: "If successful, Polaris will be the first fusion machine to demonstrate electricity from fusion."

The company harvests electrical current from the collapsing magnetic field that confines the plasma, bypassing the steam turbine. D-He3 produces more charged particles, making it better suited to that extraction.

Orion's Clock

Helion broke ground on Orion on July 30, 2025, on land leased from the Chelan County Public Utilities District. A conditional use permit for the reactor building followed in October.

Its binding contract with Microsoft requires 50 megawatts of grid power by late 2028, roughly 30 months from now. If Helion misses, it faces financial penalties payable to Microsoft and to Constellation, which manages the project's transmission; their amounts are not public. Mike Campbell, a fusion physicist at UC San Diego, told Fortune in February he would be "happily surprised" if the 2028 deadline was met.

On May 8, GeekWire reported that Helion is building Tiny Merge, a new testbed less than one-eighth the size of Polaris. Key issues remain in its larger prototypes that must be resolved before Orion's design can be locked in. With 30 months to the contracted delivery date, Helion is adding a prototype iteration, not removing one.

The Posting Mix

Helion had 491 employees as of March 2026 and posted roughly 110 open roles in mid-May. The positions cluster around plasma diagnostics, capacitor manufacturing, and materials science.

A firm 30 months from its first contracted megawatt-hour would typically be hiring grid interconnection engineers, utility interface specialists, and power plant commissioning managers. Those titles do not appear on Helion's careers page.

Microsoft, which backed a quantum chip before its foundational physics was established, is again the counterparty on a physics timeline it cannot independently verify.

In March 2026, Axios reported that OpenAI entered early talks to purchase 12.5% of Helion's total output, equaling five gigawatts by 2030. If that proportion holds, the implied total is 40 gigawatts, which at 50 megawatts per Orion reactor requires 800 machines. Sam Altman, who stepped back from Helion's board when negotiations began, holds a personal stake in Helion whose size has not been disclosed.

The mismatch between Helion's milestone sequence and its 30-month construction calendar changes the math on what "financial penalties" means in practice. Microsoft's penalty recovery covers Helion's contractual shortfall. The larger number (replacement power for 50 megawatts of contracted clean baseload at Pacific Northwest spot-market rates) falls outside the clause entirely.

The milestone that would derisk the 2028 contract is a first direct-electricity demonstration from Polaris, an achievement no fusion machine has produced. Helion's physics queue now runs through Tiny Merge before Polaris can pursue that result. If Tiny Merge does not resolve its outstanding plasma-formation questions before early 2027, the three-milestone gap closes the 2028 window from the inside.