A commercial nuclear fusion plant supplies electricity to a power grid before 2036.
Commercial fusion power remains a major clean-energy goal. Resolution requires a facility generating net electricity and transmitting it to a regional or national electrical grid.
Across 6 frontier models answering independently, the panel has all but ruled it out: the median is 13%.
They broadly agree, 29 points between GPT at 31% and Qwen at 2%.
Nothing is settled until 23 Sept 2036, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
GPTOpenAI
The key hinge is whether any leading pilot can convert promising physics into a licensed, grid-connected plant on schedule.
- CFS targets early-2030s delivery and Helion targets 2028, but neither has yet operated a commercial net-electric plant.
- A first grid export needs repeatable net power, turbine or direct-conversion integration, fuel handling, licensing, and transmission approval.
- Historically fusion schedules have slipped for decades; a sustained high-power demonstration by 2029 would raise this sharply.
31% -
GrokxAI
A private firm must convert an unproven net-electric demo into a licensed, grid-tied commercial plant inside ~9 years.
- Helion’s 2028 Microsoft PPA is the only near-term path, yet pulsed-FPP physics and tritium handling remain unvalidated at plant scale.
- CFS SPARC net-gain ~2027 still leaves ARC needing NRC-style licensing plus construction, historically 8–12 years even for fission.
- Fusion schedules have slipped 2–3× for 50 years; only a working net-electric prototype by 2029 would move this above ~25.
18% -
GeminiGoogle
Grid-connected fusion by 2036 hinges on private ventures compressing pilot construction and net-power demonstration into under ten years.
- Public mega-projects like ITER will not deliver power before the late 2030s, leaving success entirely to well-funded private startups.
- Front-runners like Commonwealth Fusion Systems and Helion target 2028-2032 grid demos, but first-of-a-kind engineering delays routinely add 4-6 years.
- A working demo producing net-electric output by 2029 with swift NRC licensing would raise these odds above 40%.
14% -
DeepSeekDeepSeek
Whether any private fusion developer can turn net-energy shots into a grid-connected commercial plant within roughly ten years.
- No fusion device has yet produced net electricity; NIF's 2022 ignition was laser energy gain, not grid power, and ITER won't run D-T until 2035.
- Commonwealth Fusion's SPARC targets net energy around 2027 and ARC ~2030s, but no developer has a licensed, financed commercial plant under construction today.
- Base rate for first-of-a-kind energy plants from concept to grid is 15-25 years; fusion's unresolved tritium breeding and materials issues push it below that.
12% -
ClaudeAnthropic
Success hinges on whether any fusion project can achieve net electric output and grid connection within ~10 years, a leap no current effort is close to.
- No fusion device has yet produced net electrical output; ITER (experimental, not grid-connected) delayed fusion power demonstration to the 2030s-2040s.
- Private firms like Commonwealth Fusion and Helion target pilot plants by early-2030s, but these are demonstration/prototype scale, not grid-supplying commercial plants.
- Historical base rate for fusion timeline promises is poor—every past 'decades away' estimate has slipped further, and engineering/materials challenges remain unresolved.
4% -
QwenAlibaba
No fusion project has even begun building a grid-connected plant; the engineering and licensing timeline alone exceeds 10 years.
- ITER won't achieve first plasma until ~2035 and doesn't generate electricity; DEMO-class successors target 2040s at earliest.
- Private firms (Commonwealth, TAE, Helion) claim 2030s demos, but none has a licensed reactor design or construction permit from any regulator.
- Even fission plants take 8-12 years from license to grid connection; no fusion entity has started that process, making a 2036 grid connection essentially impossible.
2%
6 of 6 models answered · 29 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 26 Sept 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.