The United Kingdom’s first AI supercomputer dedicated to fusion research, SUNRISE, will soon connect with its American counterpart, PPPL’s STELLAR-AI, in a transatlantic effort to accelerate the development of fusion energy. Scientists from the Princeton Plasma Physics Laboratory and the UK Atomic Energy Authority convened to formalize the SUNRISE – STELLAR-AI Federation, a partnership designed to share data and use the combined power of two experimental facilities focused on spherical tokamaks.
“A fusion power plant is one of the most complex machines humanity has ever tried to build, and no single laboratory or nation will design it alone,” said Jonathan Menard, chief scientist at PPPL, explaining that the linked systems will allow researchers to train AI on data from both facilities and hasten progress toward a compact fusion power plant.
SUNRISE & STELLAR-AI Federation Accelerates Fusion Research
The SUNRISE – STELLAR-AI Federation will enable the shared training of artificial intelligence models using data from two distinct experimental fusion facilities, accelerating research into viable fusion energy. This transatlantic collaboration formally links the UK Atomic Energy Authority’s (UKAEA) SUNRISE with the Princeton Plasma Physics Laboratory’s (PPPL) STELLAR-AI, creating a unified platform for AI-driven fusion discovery. The initiative builds upon a memorandum of understanding signed in June between UKAEA and PPPL, establishing a framework for joint endeavors in advanced computing and artificial intelligence.
Both SUNRISE and STELLAR-AI focus on spherical tokamaks, devices shaped like cored apples that utilize magnetic fields to contain plasma, indicating a shared design prioritization within the collaboration. This focus allows for direct comparison of results and streamlined AI training across similar experimental setups.
The long-term vision extends beyond immediate gains, with plans to develop a shared foundation model specifically for spherical tokamaks and, eventually, a broader range of tokamak configurations. The partnership is bolstered by significant investment; SUNRISE received £45 million from the UK government, while STELLAR-AI represents a $13 million platform operated with support from Princeton University. This funding supports not only the supercomputers themselves, but also the infrastructure surrounding them, including the UKAEA’s Culham Campus designation as the UK’s first AI Growth Zone.
The campus’s direct connection to the UK’s 400 kilovolt supergrid provides essential power for the intensive computing demands of fusion research. “Our goal is to let models and experiments move freely between the two systems,” said Shantenu Jha, head of computational sciences at PPPL. “We will turn a collection of supercomputers into a single engine for fusion discovery.” The collaboration extends beyond simply sharing computing resources; it aims to foster a unified research approach.
Joe Milnes, executive director for engineering and computing at UKAEA, emphasized the importance of partnerships in tackling the immense scientific and engineering hurdles inherent in fusion research. “Fusion is one of the great scientific and engineering challenges of our time. To solve these challenges, fusion needs partnerships,” Milnes stated.
Both organizations operate leading compact fusion devices, the National Spherical Torus Experiment-Upgrade (NSTX-U) at PPPL and the Mega Amp Spherical Tokamak Upgrade (MAST-U) at UKAEA, providing complementary experimental data for the AI algorithms to analyze. This transatlantic effort also aligns with broader national initiatives; on the U.S. side, the work is part of the Department of Energy’s Genesis Mission, a program designed to unite national labs, industry, and academia to accelerate scientific discovery in energy, science, and security.
The federation is expected to facilitate a more efficient and comprehensive approach to fusion research, allowing scientists to use the collective expertise and resources of both nations. “The UK and U.S. are two global leaders in fusion research, and the federation of SUNRISE and STELLAR-AI can build on a long history of transatlantic cooperation to further advance fusion development,” Milnes added.
A fusion power plant is one of the most complex machines humanity has ever tried to build, and no single laboratory or nation will design it alone.
Jonathan Menard, chief scientist at PPPL
UKAEA-PPPL Partnership Advances Digital Twin Development
This coordinated effort goes beyond simply combining datasets; it focuses on building digital twins of both experimental devices, allowing researchers to test modifications and anticipate machine behavior in simulation before physical implementation. These virtual replicas will not only aid in the design of future power plants but also support real-time decision-making during ongoing experiments, which is a key step toward sustained fusion. The creation of these digital twins relies on a shared foundation model for spherical tokamaks, a design prioritized by both PPPL and UKAEA given their operation of leading compact fusion devices.
Models trained on a single machine often struggle to accurately predict conditions, highlighting the importance of data quantity and quality for building effective AI tools. This interoperability will allow researchers to use the strengths of both SUNRISE and STELLAR-AI, assigning computing jobs to the processor best suited for each specific problem.
Beyond the technical aspects of data sharing and model training, the collaboration aims to streamline the design cycle for future fusion systems. By coordinating priorities and expanding the federation over time, PPPL and UKAEA hope to create a more comprehensive and efficient approach to fusion research.
Rob Akers, UKAEA director of computing programmes and senior fellow, explained that the partnership will draw on experimental data from both leading machines, using simulation to fill gaps and extend datasets into unexplored regimes. The £45 million investment in SUNRISE, UKAEA’s new AI supercomputer, and the $13 million allocated to PPPL’s STELLAR-AI platform demonstrate a significant commitment to AI-driven fusion research.
These platforms are designed to host and mature AI tools, providing the computational power necessary to analyze complex plasma behavior and optimize fusion parameters. “The real power will come from working as one team. Together, we can learn faster and learn more, extracting maximum insight,” Akers said. It also aligns with broader national initiatives, including the U.S.
Fusion is one of the great scientific and engineering challenges of our time. To solve these challenges, fusion needs partnerships.
Joe Milnes, executive director for engineering and computing at UKAEA
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
