The UK and US are connecting their fusion research programmes more closely through shared AI and supercomputing infrastructure, pairing scientific collaboration with regulatory cooperation as both countries try to move fusion from laboratory research towards an industry capable of attracting commercial investment.

The UK Atomic Energy Authority and the US Department of Energy’s Princeton Plasma Physics Laboratory have agreed to explore linking SUNRISE, the UK’s fusion-focused AI supercomputer, with the American laboratory’s STELLAR-AI platform. The proposed SUNRISE–STELLAR-AI Federation would allow researchers on either side of the Atlantic to combine computing, artificial intelligence, experimental data, and fusion expertise rather than treating national facilities as isolated programmes.

Alongside the research agreement, the UK and US governments are strengthening cooperation over fusion regulation. Britain has already created a dedicated domestic regulatory framework for the technology, while officials are trying to ensure that emerging rules do not create unnecessary divergence as private developers begin making longer-term investment decisions around prototype plants, supply chains, and specialist engineering facilities.

The agreements sit within a substantially larger British spending programme. The government has committed more than £2.5 billion to fusion between 2025–26 and 2029–30, including £1.3 billion for the next stage of the STEP prototype plant programme, £740 million for research infrastructure, and £125 million for the AI Growth Zone at Culham.

Compute is becoming part of the experiment

Fusion research has always depended heavily on computation because recreating and controlling the conditions inside a star requires scientists to model complicated physical systems. Plasma behaviour, magnetic confinement, materials performance, heat loads, component geometry, and fuel cycles all interact, while building and testing full-scale hardware remains costly and slow.

AI can change the economics of some of that work by creating faster surrogate models from expensive simulations and experimental data. Instead of repeatedly running the most computationally intensive physics models from scratch, researchers can train machine-learning systems to approximate certain behaviours quickly enough to explore larger design spaces or support real-time control.

SUNRISE sits at the centre of the UK’s attempt to develop that capability. The government has allocated £45 million to the 1.4MW system, which is intended to support fusion-specific AI, simulation, design, data analysis, and automation while connecting with other British computing resources. A further £80 million within the Culham AI Growth Zone supports research computing and preparations for larger-scale AI infrastructure.

The proposed link with STELLAR-AI takes that model across national boundaries. Rather than merely exchanging papers or research staff, the two laboratories are examining how their computing platforms, data, and AI methods can work together. For fusion, where experiments are relatively scarce and expensive compared with the quantity of data available to mainstream commercial AI, pooling research infrastructure could increase the amount and variety of evidence available to models.

Regulation is becoming part of the investment case

The second part of the UK-US relationship is less technically dramatic but commercially important. Fusion sits awkwardly between established nuclear regulation and an emerging industry that argues its risk profile differs materially from conventional fission, particularly because a fusion reaction does not operate through a self-sustaining chain reaction.

Britain has opted to regulate fusion under a framework distinct from conventional nuclear fission, and the government has presented that approach as an advantage for attracting developers. It has also published a draft Fusion National Policy Statement and committed to developing a dedicated market framework intended to give investors greater certainty around eventual commercial deployment.

Closer regulatory cooperation with the US could reduce the likelihood that companies developing equipment, software, materials, or entire plant designs have to navigate substantially different requirements in two of the largest fusion research markets. Complete harmonisation is unlikely, but greater alignment over safety evidence, licensing approaches, and technical standards can lower the cost of selling the same technology internationally.

That is relevant because the emerging fusion economy reaches beyond reactor developers. Commercial plants would require specialist magnets, robotics, advanced materials, power electronics, tritium-handling technology, data systems, remote maintenance, control software, computing infrastructure, and engineering services, giving suppliers a direct interest in whether national programmes develop compatible technical and regulatory expectations.

Public infrastructure carries much of the early risk

Despite renewed private investment in fusion, much of the enabling infrastructure remains publicly financed. The UK’s programme includes £180 million for the Lithium Breeding Tritium Innovation facility, £110 million for industry support and commercialisation, £80 million for international collaborations, and £50 million for skills alongside the STEP prototype plant and Culham computing investments.

That spending reflects the maturity of the technology. Fusion companies may be raising private capital, but no commercial fusion power station yet supplies electricity to a national grid, while major scientific, materials, fuel, and engineering challenges remain. Governments are consequently financing facilities that individual startups would struggle to justify independently but which could support a wider industrial base if the technology advances.

Britain’s fusion strategy targets a prototype STEP plant at West Burton by 2040, leaving a long development path between today’s research systems and commercial electricity generation. Linking SUNRISE and STELLAR-AI will not shorten that path by itself, but it gives researchers another way to test designs, learn from experiments, and coordinate work across two large public research programmes.

The pairing of computing and regulation is more revealing than either announcement in isolation because governments increasingly treat fusion as an emerging industrial system that requires compute, standards, capital, skills, supply chains, and regulation to develop together.