World’s most powerful AI supercomputer dedicated to fusion energy begins work
EL.KZ Информационно-познавательный портал
The UK government announced that it will invest £45 million, about $60 million, to build a powerful new artificial intelligence supercomputer to accelerate nuclear fusion research,El.kz cites Interesting Engineering.
The system, named Sunrise, will be installed at the UK Atomic Energy Authority’s Culham campus in Oxfordshire and is expected to begin operating in June.
Officials say the machine will help scientists better understand the complex physics at work in fusion reactors. By combining advanced computing with artificial intelligence models, the supercomputer could allow researchers to test ideas virtually before building costly experimental systems.
AI computing system built for fusion science
Sunrise is being described as the most powerful AI-focused supercomputer dedicated specifically to fusion energyresearch. The system is funded by the UK Department for Energy Security and Net Zero and will operate at about 1.4 megawatts.
The computer will become the first major infrastructure project in what the government calls an “AI Growth Zone” planned for the Culham Science Campus. Officials hope the zone will bring together artificial intelligence development and advanced energy research in one location.
Fusion energy research requires enormous computing power. Scientists need to simulate the behavior of plasma that reaches temperatures hotter than those in the sun’s core. They must also study how extreme heat and radiation affect reactor materials.
Traditional simulations can take a long time and require significant resources. By using artificial intelligence models trained on physics data, Sunrise aims to make those simulations faster and more detailed.
Digital twins could reduce costly experiments
The system is designed to combine high-performance computing with physics-informed AI models. This approach will allow researchers to build digital twins of fusion systems. These virtual versions of reactors can be used to test designs and operating conditions before building real prototypes.
According to government officials, Sunrise will deliver up to 6.76 exaFLOPS of AI modeling performance. This measurement refers to AI workloads rather than traditional supercomputing benchmarks, but it still represents a major boost in computing capability for the country’s fusion programs.
The system will run on AMD EPYC processors and AMD Instinct GPU accelerators installed on Dell PowerEdge servers. WEKA will provide the system’s storage platform. The project also has support from Intel, the University of Cambridge, and the UK Atomic Energy Authority.
Scientists expect the computer to help address several difficult technical problems. These include modeling turbulent plasma behavior, studying materials that can withstand conditions in fusion reactors, and developing technologies to breed tritium fuel for future fusion plants.
El recommends 
