A nuclear fusion facility achieves sustained engineering net-energy gain before 2032.
Net energy gain has occurred only in brief laboratory pulses. Resolves if a fusion reactor produces net positive electrical or thermal output across continuous operation.
Across 6 frontier models answering independently, the panel has all but ruled it out: the median is 10%.
They broadly agree, 27 points between GPT at 31% and Qwen at 4%.
Nothing is settled until 12 Nov 2036, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
GPTOpenAI
It turns on whether a grid-relevant pilot can sustain plant-level net output, not merely repeat high-gain shots.
- ITER is not scheduled for deuterium-tritium operation until 2039, removing the clearest public path before the deadline.
- Private developers target the early 2030s, but no project has yet demonstrated a burning plasma or full power-conversion balance.
- I would raise this sharply if a funded pilot begins construction by 2028 with credible tritium, heat-extraction, and turbine designs.
31% -
GeminiGoogle
Turns on whether CFS's SPARC/ARC or a well-funded stellarator overcomes materials degradation and Q_eng>1 before 2032.
- ITER's DT campaign is delayed past 2035, leaving the timeline entirely reliant on aggressive private ventures like Commonwealth Fusion.
- True engineering gain requires wall-plug Q_eng>1 over hours, facing brutal neutron damage and tritium breeding challenges by 2031.
- CFS completing SPARC ahead of schedule with net plasma gain by 2027 would be the primary signal to substantially upgrade these odds.
12% -
GrokxAI
Private firms must leap from first plasma to continuous net-electric output in <6 years.
- ITER DT ops slip past 2035; no public machine can meet the 2032 cutoff.
- CFS SPARC and Helion target scientific Q, not plant-level net electricity or continuous thermal/electric surplus.
- 70-year delay pattern plus unresolved first-wall, tritium and divertor issues keep this far below even aggressive private base rates.
11% -
DeepSeekDeepSeek
No fusion project has a credible path to sustained net-energy gain before 2032; the physics and engineering gaps are too large.
- NIF's 2022 ignition was a laser shot yielding ~3.15 MJ from 2.05 MJ, but no electricity and no continuous operation.
- ITER's first plasma is scheduled for 2035, and its DT campaign with Q=10 is planned for 2039, after this deadline.
- Private firms like Commonwealth Fusion and Helion target net gain in the early 2030s, but none has demonstrated sustained operation or a working pilot plant.
8% -
ClaudeAnthropic
Sustained (not pulsed) engineering net-energy gain by 2032 requires facility operation decades ahead of current roadmaps like ITER and private startups.
- NIF and similar facilities only achieve brief pulsed ignition, not continuous sustained net output needed here.
- ITER itself targets first plasma around 2035 and full deuterium-tritium fusion power experiments even later, missing 2032 entirely.
- Private fusion firms (Commonwealth, Helion, TAE) target demo reactors in early-to-mid 2030s, but sustained net engineering gain is a much higher bar than any announced…
5% -
QwenAlibaba
No existing or under-construction facility can plausibly reach sustained Q_eng>1 in continuous operation within six years.
- ITER won't produce electricity and targets first plasma ~2030-2035; private firms like CFS and Helion target brief Q>1 demos, not sustained continuous output.
- The gap between a single-shot scientific gain (NIF, Dec 2022) and sustained engineering net gain spans tritium breeding, heat extraction, and materials durability—none…
- A surprise would require a private company (likely Helion or CFS) to leapfrog multiple engineering milestones simultaneously; no precedent exists for compressing a…
4%
6 of 6 models answered · 27 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 15 Nov 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.