Helion insists it will deliver fusion power to Microsoft by 2028
Source · Energy desk
— Summary
Helion Energy, the US nuclear fusion start-up backed by Sam Altman and Peter Thiel, has told the FT it "remains on track" to supply Microsoft with commercial electricity through the grid in 2028 — a deadline that would make it the first fusion company ever to do so. CFO Pragav Jain said the Washington-state generator being built for Microsoft would be a "sub-scale commercial unit", less efficient than the full-scale systems Helion hopes to deploy later. Helion has also signed a 500-megawatt deal with steelmaker Nucor for 2030, and Axios reported last month that OpenAI was in talks to buy electricity from the company.
The claim divides the sector. Helion is one of the best-funded fusion players — $1bn raised, a $5.4bn valuation in its latest round — but smaller than Commonwealth Fusion Systems, which has raised about $3bn and plans its first commercial plant in the early 2030s with Google as offtake customer. US government scientists reached "net energy gain" (more power out than in to trigger fusion) but no start-up has matched that, let alone built a commercial plant. Rivals argue Helion's 2028 timeline "doesn't add up" and criticise its silence on how it will handle high-energy neutrons, which damage reactor structures. Helion says it is developing a material with "its own kind of healing mechanism" but has disclosed little to protect IP from "copycats". According to the Fusion Industry Association, 89% of private fusion companies believe the technology will be on the grid by the 2030s.
Helion's edge, if real, is its design. It does not plan to use turbines; instead it would generate electricity directly from the changing magnetic field as the plasma expands, inducing current in surrounding coils. One fusion investor described this as "the holy grail" — in theory, it would let Helion make commercially useful electricity with a smaller net energy gain than rivals need. The company is testing a pre-commercial machine meant to "demonstrate electricity from fusion", but Jain declined to say whether it was close to break-even. Source: Financial Times, 18 April 2026, Ryohtaroh Satoh.
The story in one line. Altman- and Thiel-backed Helion is publicly standing by its 2028 Microsoft commitment even as rivals call the timeline implausible and as no fusion start-up has yet hit net energy gain, let alone delivered commercial electricity.
Key numbers
Helion valuation:$5.4bn (latest round); $1bn raised to date.
Microsoft agreement: deliver grid power by 2028 — deadline no fusion firm has ever met.
Nucor agreement:500-MW fusion plant for 2030.
OpenAI: reportedly in talks to buy electricity from Helion (Axios, last month).
Commonwealth Fusion Systems (main rival):~$3bn raised, first commercial plant in early 2030s, Google as offtake.
Industry survey (Fusion Industry Association):89% of private fusion companies expect grid power in the 2030s.
Key backers:Sam Altman (OpenAI), Peter Thiel.
Why it matters
Helion is the first test case of the “AI scale-up buys fusion power directly” thesis: a hyperscaler (Microsoft) underwrites the commercial demand that nuclear fusion has historically lacked. If the 2028 delivery happens — even from a deliberately “sub-scale” unit — it is a step-change validation of private fusion and lends pricing power to every firm with a data-centre offtake story. If it slips, the lesson is the opposite: private capital can fund reactors but cannot compress physics timelines, and hyperscalers will shift back to SMRs, gas, and geothermal for 2030s compute loads.
Two technical doubts haunt the sector’s read on Helion. First, the company’s cryptic response on neutron damage (“a self-healing material”) has not convinced peers: neutron embrittlement is the main reason fusion plants keep slipping into the 2040s. Second, Helion’s turbine-free direct-conversion design is genuinely innovative — rivals call it “holy grail” — but unproven at scale, and Helion refuses to disclose how close its pre-commercial machine is to break-even. The IP-protection defence is plausible; it is also exactly what a company without results would say.
Takeaway
The 2028 Microsoft deadline is now a market signal. Watch for disclosure of a pre-commercial break-even run (the make-or-break technical milestone), regulatory updates on the Washington-state facility, and whether OpenAI firms up its own offtake deal. A second fusion-to-grid contract for a hyperscaler would confirm a genuine market; a Helion delay would reset the sector’s valuation baseline lower.
Source: Financial Times, 18 April 2026, Ryohtaroh Satoh.