• UKAEA and PPPL plan to federate two fusion-focused supercomputers, allowing researchers to share data and train AI across UK and US facilities.
• The UK-backed SUNRISE system has received $60.6 million, while PPPL’s STELLAR-AI platform cost $13 million.
• The partnership could support future fusion power plants, including the UK’s STEP programme and PPPL’s proposed STAR reactor.

The UK and US are deepening their fusion energy partnership by linking two specialist supercomputers to accelerate research using artificial intelligence.

The UK Atomic Energy Authority (UKAEA) and Princeton Plasma Physics Laboratory (PPPL), a US Department of Energy laboratory, signed a joint declaration of intent at the Global Fusion Policy Summit in London.

The agreement will connect SUNRISE, the UK’s AI supercomputer dedicated to fusion, with PPPL’s STELLAR-AI high-performance computing platform. The partnership will operate as the SUNRISE–STELLAR-AI Federation.

The initiative brings together substantial public research infrastructure. The UK government has committed $60.6 million to SUNRISE, its first AI supercomputer dedicated specifically to fusion energy. STELLAR-AI cost $13 million and is operated by PPPL with support from Princeton University.

Linking fusion data across the Atlantic

The collaboration builds on a memorandum of understanding signed by UKAEA and PPPL in June.

Researchers plan to combine experimental data from the MAST Upgrade facility in Oxfordshire with data from NSTX-U in New Jersey. Both are compact spherical tokamaks with similar designs.

Connecting the computing systems will allow researchers to share experimental datasets and train AI using results from both facilities. They will also be able to run models on whichever machine is best suited to a particular task.

That access could help researchers build more capable machine-learning models. By drawing on multiple facilities, the models could better capture the underlying physics of fusion and help researchers explore conditions that have not yet been tested experimentally.

“To solve these challenges, fusion needs partnerships. The US and UK are two global leaders in fusion research,” said Joe Milnes, Executive Director for Engineering and Computing at UKAEA.

Rob Akers, Director of Computing Programmes and Senior Fellow at UKAEA, said the partners will combine experimental data with simulation. The approach could fill gaps in existing datasets and extend research into unexplored operating regimes.

The work is also expected to support digital twins of both fusion machines. These virtual models could help researchers test designs and operating conditions before applying them to physical systems.

RELATED ARTICLE: Thea Energy Raises $100 Million To Scale Fusion Power Plants, Expand U.S. Energy Infrastructure

AI targets future commercial fusion plants

The research has a longer-term commercial objective. UKAEA expects the collaboration to support the UK’s Spherical Tokamak for Energy Production, or STEP Fusion. In the US, the work could contribute to PPPL’s proposed Spherical Tokamak Advanced Reactor, known as STAR.

Both concepts are part of efforts to move fusion research beyond experimental facilities and towards power-producing plants.

Jonathan Menard, Chief Scientist at PPPL, said the complexity of designing a fusion power plant makes international cooperation necessary. Federating SUNRISE and STELLAR-AI would allow AI models to learn from two major fusion facilities rather than relying on a single experimental dataset.

Shantenu Jha, Head of Computational Sciences at PPPL, said the partners want models and experiments to move between the two computing systems.

Over time, the federation could expand further. Its long-term goal is to develop a shared foundation model for spherical tokamaks. That model could eventually incorporate a broader range of tokamak configurations.

Fusion partnerships broaden beyond computing

The supercomputing agreement comes as the UK and US expand cooperation across other technical barriers facing fusion energy.

The University of Birmingham and US nonprofit EPRI have separately launched a collaboration under the Fusion Reactor Shielding Materials programme, known as FURESHMA.

The $3.54 million programme, which includes support from UK Research and Innovation, will test advanced boride and carbide shielding materials. These materials are designed to protect critical reactor components from extreme conditions and intense neutron irradiation.

For governments and investors, these collaborations address two major challenges facing commercial fusion: computational complexity and material durability.

Fusion has attracted growing public and private capital, but commercial deployment still depends on solving significant engineering problems. Linking national research infrastructure gives scientists access to larger datasets and greater computing capacity without requiring each country to replicate every capability.

The UK-US federation also shows how AI infrastructure is becoming part of national energy strategy. If the model expands as planned, shared computing and research platforms could play a larger role in turning decades of fusion science into deployable power technologies.