A commercial nuclear fusion reactor begins supplying electricity under a power purchase agreement before 2032.
Private fusion companies have signed preliminary PPAs targeting grid delivery around 2028-2030. This tracks whether delivery under a commercial contract happens by 2032.
Across 6 frontier models answering independently, the panel has all but ruled it out: the median is 9%.
They broadly agree, 6 points between Gemini at 14% and Qwen at 8%.
Nothing is settled until 1 Oct 2036, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
GeminiGoogle
Unproven net electric gain must transition to licensed, utility-grade plant engineering in under six years.
- Helion and CFS target 2028-2030 grid power, but novel energy hardware historically slips by 3 to 5 years past pilot milestones.
- No commercial design has yet demonstrated net electricity (Q_eng > 1), leaving zero margin for regulatory and grid interconnect delays.
- A verified net-electric continuous burn by Helion or Commonwealth by late 2027 would push odds well above 35%.
14% -
GrokxAI
Any private firm must actually deliver net power under a signed PPA by end-2031.
- Helion-Microsoft PPA targets 2028 but still needs a net-electricity demo then a FOAK plant; none yet exist.
- Even small fission plants routinely take >5 years to build; fusion FOAK plus NRC licensing will exceed that.
- A 2027 net-power shot plus fast permitting would raise odds, but tritium/He-3 and materials remain blockers.
12% -
ClaudeAnthropic
Hinges on Helion, CFS or similar hitting net-engineering-gain and demo-plant delivery by 2031, which no plasma physics group has done.
- Helion's PPA with Microsoft targets 2028 but no fusion device has ever produced net electrical output, so a 2028-2031 commercial delivery requires skipping steps no one…
- Fusion timelines have slipped by years to decades historically (ITER, NIF, every prior 'commercial fusion by 20XX' claim), so base rate for hitting an aggressive…
- Regulatory licensing (NRC), engineering scale-up from lab shots to continuous grid output, and tritium fuel cycle issues all remain unresolved and typically add years…
9% -
GPTOpenAI
It hinges on one private plant proving net-electric, reliable operation and clearing nuclear licensing before 2032.
- No private fusion project has produced net electricity; the leading promised 2028–30 dates imply a historically unprecedented scale-up.
- Helion, CFS and others must finish a first-of-kind plant, secure permits, interconnect, and meet contracted availability by 2031.
- A sustained grid-power demonstration from Helion’s Polaris-class machine by 2028 would materially raise the odds.
9% -
DeepSeekDeepSeek
No private fusion device has ever produced net electricity, so a signed PPA delivering power by 2032 requires a physics first.
- Commonwealth Fusion's ARC in Chesterfield, Virginia targets ~2030s operation, but SPARC must first demonstrate Q>1, now slipping toward 2027.
- Helion's 2028 Microsoft PPA and TAE's 2030s deals are non-binding or contingent; no fusion device has reached scientific breakeven since NIF's 2022 single shot.
- Base rate for first-of-a-kind energy plants hitting a 6-year target is near zero; ITER is 20+ years late, so 9% reflects only a genuine surprise scenario.
9% -
QwenAlibaba
Even the most optimistic private fusion timelines (Commonwealth, Helion) face first-of-a-kind engineering delays that push past 2032.
- Commonwealth Fusion targets mid-2030s for SPARC→ARC; Helion's 2028 Microsoft PPA already slipped once and requires net-electricity demo first.
- No fusion device has yet achieved sustained net electricity output; the gap between plasma physics milestones and a grid-connected plant is typically 10+ years.
- A development that would raise this: Helion or CFS demonstrating sustained net-electricity generation before 2029, which would compress the engineering timeline.
8%
6 of 6 models answered · 6 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 4 Oct 2026. Forecast by google/gemini-3.8-flash, anthropic/claude-sonnet-5, openai/gpt-5.6-terra, x-ai/grok-4.6, deepseek/deepseek-v4.1-flash, qwen/qwen3.8-max-0902 via OpenRouter.