On March 23, 2026, Axios reported that OpenAI is in advanced talks to buy electricity from Helion Energy, the fusion startup backed by OpenAI’s own CEO. If the agreement lands, OpenAI would be guaranteed 12.5% of Helion’s future production — roughly 5 gigawatts by 2030, scaling to 50 gigawatts by 2035. The same day, Sam Altman stepped down as chair of Helion’s board.
This is not another data-center power contract. It is the first time the AI industry has written commercial fusion into a decade-long procurement script. Fifty gigawatts is close to half of the United States’ existing nuclear fleet and exceeds the total installed generation capacity of most countries. The talks are early-stage, but the numbers make the point on their own: the bottleneck for AI expansion has shifted from chips to electricity, and frontier companies are willing to lock in commitments for power that does not yet exist.
The Deal: 12.5% of Output, a Decade-Long Ramp
According to Axios, the structure is a guaranteed share: OpenAI would receive 12.5% of everything Helion produces. TechCrunch’s math puts that at 5 GW by 2030 and 50 GW by 2035. For scale, Helion’s two existing customers are Microsoft, which signed a power purchase agreement in 2023 with deliveries expected from 2028, and the steelmaker Nucor — both demonstration-scale, hundred-megawatt-class commitments. The OpenAI discussion is two orders of magnitude larger.
The guaranteed-share design deserves attention. OpenAI would not be buying a fixed quantity of electricity but a fixed fraction of every reactor Helion ever builds. That effectively converts AI demand into a long-term offtake contract for the fusion industry, binding capacity risk and demand risk together.
The Conflict of Interest and How It Was Handled
Altman is one of Helion’s earliest and most prominent investors; among the backers of its $425 million raise in early 2025 were Altman himself, Mithril, Lightspeed, and SoftBank. OpenAI buying 50 GW from a company its CEO invested in is a textbook conflict.
The handling came in two layers. First, Altman has resigned as Helion’s board chair — CEO David Kirtley told TechCrunch the move “enables Helion and OpenAI to partner on future opportunities.” Second, Altman reportedly recused himself from the negotiations throughout. There is precedent: last year he exited the chairmanship of Oklo, the fission startup, in similarly framed terms ahead of AI-company partnerships. Critics have a fair point — resigning a chairmanship does nothing about the equity stake — but at least the institutional overlap is gone.
Why Fusion, and Why Now
Helion’s technical path is unusual among fusion startups. Instead of heating a working fluid into steam to spin a turbine, it uses magnetic fields to convert the kinetic energy of charged fusion products directly into electricity. In principle that makes plants smaller and cheaper, and it explains the modular 50 MW rating per reactor.
On schedule, Helion’s Polaris prototype reached 150 million degrees Celsius in February 2026, approaching the roughly 200 million degrees the company believes commercial operation requires. In other words, it has not yet delivered a single commercial kilowatt-hour. But with the model race pushing compute and power demand to the limit — a trajectory we flagged in our opening outlook for 2026 — every lab is now locking future-decade electricity into contracts ahead of need.
The Math: One Thousand Reactors
At 50 MW per reactor, 5 GW implies roughly 100 machines built and grid-connected before 2030; 50 GW implies about 1,000. Against reality: Helion has zero commercial units today, and fusion as an industry has never sustained even a double-digit production cadence, let alone four digits.
The deal still makes sense for both sides. For Helion, an OpenAI offtake is proof of demand to show capital markets and supply chains. For OpenAI, it is a comparatively cheap option: if fusion commercializes on time, it has secured clean baseload power nobody else can bid away; if it slips, the loss is mostly priority, not capital.
What It Means for AI Infrastructure Teams
Three observations. First, power procurement is becoming a core AI-company function: from gas and renewables projects to geothermal PPAs to a fusion equity share of output, frontier companies now span every generation technology. Second, schedule risk is yours to manage — betting 2030 capacity plans on any single pre-commercial technology is a gamble, and a diversified portfolio remains the only robust answer. Third, conflict governance will face sharper scrutiny: when a CEO’s personal investments overlap with company procurement, a resignation is the starting point, not the answer — transparent pricing and independent audit complete it.
Sources
- OpenAI explores fusion energy in Helion power deal — Axios
- Sam Altman-backed fusion startup Helion in talks to sell power to OpenAI — TechCrunch
- Report: Helion is working on a massive fusion power deal with OpenAI — GeekWire
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.