TL;DR
- OpenAI is negotiating to buy electricity from Helion Energy — a fusion startup Sam Altman personally backs — to power its AI data centers.
- The deal would scale from 5 gigawatts by 2030 to 50 gigawatts by 2035, enough to run millions of GPUs.
- The move mirrors Microsoft’s fusion bets and positions OpenAI ahead of Google and Meta’s nuclear power plays.
- It’s also a remarkable conflict-of-interest tightrope: the CEO of the buyer is an investor in the seller.
OpenAI Chases Gigawatts from Altman’s Fusion Bet
OpenAI is in advanced talks to purchase electricity from Helion Energy, a nuclear fusion startup that CEO Sam Altman has personally invested in, according to Axios. The deal would secure power for OpenAI’s sprawling AI data centers, starting at 5 gigawatts by 2030 and ramping to 50 gigawatts by 2035.
That’s not a typo. Fifty gigawatts.
To put it in perspective, 50 gigawatts is roughly the entire generating capacity of South Korea. OpenAI reportedly plans to consume that much electricity within a decade to train and run the next generation of AI models.
Why OpenAI Can’t Scale Without Solving Power First
The energy bottleneck is the single biggest constraint on AI progress right now — and OpenAI knows it. Training runs for frontier models now consume as much electricity as small cities. Inference at scale isn’t much better.
Since GPT-4 launched, OpenAI’s power needs have reportedly doubled every year. Data centers are already straining regional grids. Utilities can’t build coal plants fast enough, and renewables alone won’t cut it when you need 24/7 baseload power with zero downtime.
Fusion — if Helion can actually deliver it — solves that problem. It’s carbon-free, generates massive amounts of power from tiny amounts of fuel, and doesn’t depend on the sun shining or the wind blowing. It’s the holy grail for hyperscalers drowning in compute demand.
But here’s the uncomfortable part: Altman isn’t just OpenAI’s CEO. He’s also one of Helion’s early backers. That means the company negotiating this deal is led by someone with a direct financial interest in the outcome.
I’m not saying it’s corrupt. But it’s certainly… cozy.
The optics are tricky. If Helion lands this contract, Altman’s personal investment portfolio benefits directly. If Helion stumbles, OpenAI’s entire scaling roadmap craters. Either way, the CEO is on both sides of the table.
It’s like hiring your brother-in-law’s construction company to build your house. Maybe he does great work! But you’d better document every decision.
Microsoft Already Bet on Fusion — OpenAI Is Playing Catch-Up
OpenAI isn’t the first hyperscaler to chase fusion power. Microsoft signed a deal with Helion back in 2023 to buy electricity starting in 2028. That agreement reportedly covers a smaller initial capacity, but it validated the broader thesis: fusion is real enough to bet on.
Google and Meta, meanwhile, are chasing nuclear fission deals. Google reportedly signed agreements with small modular reactor startups. Meta is exploring partnerships with existing nuclear operators to secure dedicated grid capacity.
The stakes are existential. Whoever locks in reliable, scalable, carbon-free power first can train bigger models, run more inference, and undercut competitors on cost. Power is the new moat.
And OpenAI is gambling that Helion — a company that has never delivered a single watt of fusion power to a commercial customer — will beat decades of physics challenges and deliver 5 gigawatts within four years.
That’s optimistic. Aggressively optimistic.
Fusion Promises Everything — If It Works
Helion’s approach uses a pulsed non-ignition fusion process. It heats plasma to over 100 million degrees, slams deuterium and helium-3 together, and captures the energy directly as electricity. No steam turbines. No radiation-soaked reactor walls.
It’s elegant. It’s also unproven at scale.
Fusion has been “20 years away” for the last 50 years. Helion claims it’ll break that curse by simplifying the engineering and focusing on a fuel cycle that produces less neutron radiation. But building a prototype that works in a lab is very different from delivering gigawatts of reliable power to a data center that can’t tolerate a single millisecond of downtime.
If Helion delivers, it changes everything. OpenAI gets unlimited clean power, Altman looks like a visionary, and the entire AI industry pivots to fusion overnight.
If Helion doesn’t deliver, OpenAI is stuck scrambling for grid capacity in 2030 while competitors who bet on boring old nuclear fission are already online.
Three Things to Monitor as This Deal Unfolds
First, watch whether OpenAI discloses the financial terms and governance safeguards around Altman’s dual role. If the board signed off on this without independent review, that’s a red flag. If they brought in outside advisors and structured the deal to firewall Altman’s personal interests, that’s a green light.
Second, track Helion’s technical milestones. The company needs to demonstrate net energy gain at commercial scale — not in a lab, but in a facility that can actually feed power to the grid. If they hit that benchmark by 2028, the 2030 target becomes plausible. If they don’t, the whole timeline collapses.
Third, monitor how Google, Microsoft, and Meta respond. If they start signing their own fusion deals, it validates the strategy. If they double down on fission or renewables instead, it suggests the smart money thinks fusion is still too risky. The next 18 months will reveal whether this is a calculated bet or a desperate Hail Mary.
FAQ
Why does OpenAI need 50 gigawatts of power by 2035?
Training and running frontier AI models consumes staggering amounts of electricity — reportedly as much as small countries. OpenAI’s power needs have doubled annually since GPT-4, and the company expects that trajectory to continue as models grow larger and inference demand explodes. Fifty gigawatts would provide the baseload capacity to train next-generation models and serve hundreds of millions of users simultaneously without relying on unstable grid power.
What is Helion Energy and how does fusion work?
Helion Energy is a fusion startup backed by Sam Altman that uses a pulsed non-ignition process to generate electricity. The company heats plasma to over 100 million degrees, fuses deuterium and helium-3 atoms, and captures the energy directly as electricity without steam turbines. Fusion produces no carbon emissions, generates massive power from tiny amounts of fuel, and creates far less radioactive waste than fission. The challenge is making it work reliably at commercial scale — something no company has achieved yet.
Is there a conflict of interest with Sam Altman backing Helion?
Yes, and it’s significant. Altman is both OpenAI’s CEO and a personal investor in Helion Energy. That means he has a financial stake in the company OpenAI is negotiating to buy power from. Whether this deal was structured with proper governance safeguards — like independent board review and arms-length negotiations — will determine whether it’s a responsible business decision or a problematic entanglement. Transparency around the terms and approval process will be critical.
How does this compare to Microsoft and Google’s energy strategies?
Microsoft signed a fusion power deal with Helion back in 2023, making it the first hyperscaler to bet on fusion commercially. Google and Meta are reportedly pursuing nuclear fission deals instead, partnering with small modular reactor startups and existing nuclear operators. The difference comes down to risk tolerance: fission is proven but slow to deploy, while fusion promises unlimited clean power but remains unproven at scale. OpenAI’s 50-gigawatt target by 2035 is the most aggressive timeline any hyperscaler has announced.
