OpenAI talks fusion power with Helion, targeting 5 GW by 2030

The AI race is officially turning into a power grab, and Sam Altman is taking things into his own hands.

OpenAI is reportedly in advanced talks to purchase a massive amount of fusion electricity from Helion Energy, the nuclear fusion startup backed by Altman himself. According to Axios, the deal would give OpenAI an initial 12.5% cut of Helion’s total electricity output, aiming for 5 gigawatts by 2030 and scaling up to a staggering 50 gigawatts by 2035.

To clear ethical and regulatory hurdles, Altman has stepped down as Helion’s board chair after more than a decade and recused himself from the negotiations.

The astronomical math behind the deal

If these numbers materialize, Helion’s deployment schedule will have to border on science fiction.

Helion designs its modular reactors to produce 50 megawatts each. If OpenAI’s 5 gigawatts represents only 12.5% of total capacity, that means Helion plans to bring 40 gigawatts online by 2030. That equates to manufacturing and installing 800 reactors before the decade is out, followed by another 7,200 units by 2035.

While Helion told TechCrunch it has not formally announced any new customers beyond existing pacts with Microsoft and steelmaker Nucor, it confirmed Altman’s departure. Helion CEO David Kirtley stated that stepping down allows the two companies to partner on future zero-carbon power opportunities.

Helion is taking a radically different engineering route than most fusion contenders. Instead of using fusion heat to boil water and spin steam turbines, it uses magnetic fields to collide two plasma pulses inside an hourglass-shaped device. The reaction pushes back against the magnets, directly harvesting electricity.

Its Polaris prototype hit 150 million degrees Celsius in February, inching closer to the 200 million degrees needed for commercial operation. The company also secured $425 million last year from backers including SoftBank, Lightspeed, Mithril, and Altman.

Running the same playbook twice

This is not the first time Altman has performed this exact corporate shuffle.

Last year, Altman stepped down as chair of advanced fission startup Oklo to let it pursue deals with AI giants, including OpenAI. In both instances, Altman seeded the energy startups years ago, built out their governance, and then stepped aside at the boardroom door just as OpenAI arrived to sign the power checks.

Yet the technical risk remains sky-high. No private fusion outfit in history has officially achieved scientific breakeven—generating more energy than the reaction consumes. Helion says it is close, but commercial delivery remains unproven.

Why it matters

We think this move shows just how desperate frontier AI labs are becoming over grid constraints. While everyday users will not feel any immediate difference in ChatGPT today, the bottleneck for artificial general intelligence is no longer just Nvidia silicon, but base-load electricity.

For the tech industry, OpenAI is setting a precedent that traditional utility grids simply cannot support long-term scaling. Tech giants have already bought up vast amounts of solar, wind, and conventional nuclear. OpenAI skipping straight to fusion forces rival hyperscalers to evaluate untamed, speculative energy bets much sooner than planned.

Altman's strategy is undeniably clever, effectively turning his personal investment portfolio into OpenAI's private supply chain while removing direct conflict-of-interest red flags. But the risk is immense. If Helion fails to achieve commercially viable fusion by the late 2020s, OpenAI's ambitious mid-decade infrastructure roadmap could face a serious power outage.