The UK government is splashing out £45 million (c $60 million) on a new AI-driven supercomputer designed to help scientists model the chaotic physics of nuclear fusion, with the system expected to come online this summer at the UK Atomic Energy Authority’s (UKAEA) Culham campus.
The machine, called Sunrise, is being pitched as the world’s most powerful AI supercomputer that is dedicated specifically to fusion energy research. Funded by the Department for Energy Security and Net Zero (DESNZ), the 1.4MW system is slated to begin operating in June and will form the first major piece of infrastructure in what ministers describe as the UK’s planned “AI Growth Zone” at Culham in Oxfordshire.
Fusion research has long relied on large-scale simulations to understand the behavior of superheated plasma and the extreme materials in experimental reactors. The idea behind Sunrise is to combine high-performance computing with physics-informed AI models, allowing researchers to run more detailed simulations and develop digital twins of complex fusion systems before attempting costly physical experiments.
According to the government, the system will deliver up to 6.76 exaFLOPS of AI-accelerated modeling performance. That figure refers to AI workloads rather than the traditional supercomputing benchmarks used in global rankings, but it still represents a significant increase in modeling capability for the UK’s fusion research programs.
The machine will incoporate AMD EPYC processors and AMD Instinct GPU accelerators running on Dell PowerEdge infrastructure, with WEKA providing the storage platform. Intel is also supporting the project, alongside the University of Cambridge and the UK Atomic Energy Authority (UKAEA).
Officials say the system will help tackle several key challenges in fusion research, including modeling plasma turbulence, developing reactor materials, and advancing tritium fuel breeding technologies needed for future fusion systems.
Dr Rob Akers, director of computing programs at the UKAEA, said the system is intended to bring an “Apollo program” style approach to fusion development by allowing researchers to test and refine designs in a virtual environment before building them in the real world.
“Sunrise will bring that capability to fusion by combining high-fidelity simulation with physics-informed AI to develop predictive digital twins that reduce the cost, risk, and time of learning that would otherwise require expensive and time-consuming physical testing,” he said.
The supercomputer will support several UK fusion initiatives, including the LIBRTI program, which focuses on tritium fuel-cycle technologies, and the government’s flagship STEP project, a prototype spherical tokamak power plant that Britain hopes to build in Nottinghamshire in the 2040s.
Sunrise also fits into a broader push by the UK government to expand its domestic AI and supercomputing capacity. Earlier this year, ministers confirmed a separate £36 million (c $48 million) investment in the Cambridge supercomputing center, while Culham is expected to become a hub for AI-driven scientific computing tied to energy research.
Whether AI can meaningfully speed up the notoriously slow march toward commercial fusion power remains an open question. For now, the UK is betting that more computing power might help crack one of physics’ most stubborn problems a little faster. ®