Challenges and opportunities for AI to help deliver fusion energy
Organizations: STFC Hartree Centre, SciTech Daresbury, Warrington, WA4 4AD, UK · UK Atomic Energy Authority, Culham Campus, Abingdon, Oxfordshire, OX14 3DB, UK · digiLab Solutions, Generator Hub The Gallery, Kings Wharf, Exeter, EX2 4AN, UK · Institute of Physics, 37 Caledonian Road, London, N1 9BU, UK · Zenithon AI, 1111B S Governors Ave, Dover DE 19904, USA · Eindhoven University of Technology, 5600 MB Eindhoven, Netherlands · Oliver Lodge Laboratory, University of Liverpool, Liverpool, L69 7ZE, UK · Fusion Energy Insights, Culham Innovation Centre, Culham Campus, Abingdon, Oxfordshire, OX14 3DB, UK
Abstract
There is great potential for the application of AI tools in fusion research, and substantial worldwide benefit if fusion power is realised. However, using AI comes with its own challenges, many of which can be mitigated if responsible and robust methodologies are built into existing approaches. To do that requires close, long-term collaborations between fusion domain experts and AI developers and awareness of the fact that not all problems in fusion research are best tackled with AI tools. In April 2025, experts from academia, industry, UKAEA and STFC discussed how AI can be used to advance R&D in fusion energy at the first edition of The Economist FusionFest event. This Perspective is an expanded and updated summary of the round table discussion, providing more context and examples.
Figures & tables
| Application | Opportunities | Challenges |
|---|---|---|
| Plasma physics & control | Disruption and instability prediction; sub-millisecond real-time plasma control; control of magnetic topology; improved operational regimes | Trust and acceptance by operators; limited interpretability; risk of out-of-distribution predictions and hallucinations; latency constraints for foundation models |
| Materials & components | Accelerated alloy discovery; prioritisation of irradiation experiments; uncertainty-quantified prediction of material properties; faster materials qualification and TRL progression | Severe scarcity of fusion-relevant data; reliance on fission surrogate datasets; evolving performance targets; uncertainties in ground truth; relatively low TRLs |
| Engineering & design | Rapid parameter sweeps; component and system optimisation; coupling of multi-scale models; faster design iteration | Generalisation uncertainty outside training domain; high training cost; explainability challenges; fidelity of underlying simulations |
| Diagnostics & data | Improved inference of hidden plasma parameters; faster analysis of heterogeneous data with diagnostic drift; uncertainty-aware data interpretation; digital twins | Limited ground truth; diagnostic noise; extrapolation risks; uncertainty sources that cannot be learned by ML |
| Operations & maintenance | Automated remote handling and robotics; predictive maintenance; reduced downtime; safer operation in high-radiation environments; digital twins | Deployment trust and safety assurance; validation of AI-driven actions |