July 17, 2025: Bristol, UK- In the same city where Isambard Kingdom Brunel once sketched out the Clifton Suspension Bridge, a new landmark in engineering has just powered on, not a marvel of steel, but one of silicon.
Named after the iconic British engineer, Isambard-AI, the UK’s most powerful supercomputer, has entered its second phase of operation, delivering unprecedented AI compute capacity to the nation.
The system, now live at the Bristol & Bath Science Park, is powered by 5,448 Nvidia GH200 Grace Hopper Superchips, making it the largest GH200 deployment globally. With 23 exaflops of AI performance, it is designed to train large language models, simulate complex systems, and accelerate cutting-edge research across science, healthcare, and climate modelling.
The system consumes around 5MW of power and was developed at a cost of USD 301.84 million (£ 225 million), underwritten by the UK Research and Innovation (UKRI) agency.
“This is a huge moment for the UK,” said Professor Simon McIntosh-Smith, director of the Isambard-AI project. “We’ve built the largest deployment of Nvidia GH200 superchips on the planet, bigger even than the US systems like those from Microsoft or Meta.”
The system runs on Nvidia’s GH200 Grace Hopper Superchips, which combine high-performance graphics processors with energy-efficient CPUs to accelerate AI workloads. These chips are named after Rear Admiral Grace Hopper, a pioneer of computer science best known for inventing the first compiler and helping create COBOL, a foundational programming language still used today in banking and government systems.
Her contributions to programming mirror the visionary engineering of Brunel, whose name Isambard-AI proudly carries.
Isambard-AI is not alone in its mission. It joins Dawn, a complementary AI supercomputer housed at the University of Cambridge, which became operational earlier this year.
While Dawn emphasizes AI inferencing and general-purpose research access, Isambard-AI is tuned for high-end model training and simulation workloads.
Together, these two machines, one in Cambridge, the other in Bristol, are now widely recognized as twin-sister systems forming the core of Britain’s emerging AI infrastructure. Both are funded through the UK’s Frontier AI Taskforce and aligned with its broader USD 1207.35 million (£ 900 million) compute initiative to build domestic capability in foundational and frontier AI research.
“We’re delivering on our promise to give British researchers the tools they need to lead in global AI,” said Michelle Donelan, Secretary of State for Science, Innovation and Technology, during the Dawn system launch.
Dawn, constructed by Dell and StackHPC, uses a hybrid CPU-GPU architecture optimized for versatility, while Isambard-AI’s full-stack Nvidia design makes it more suitable for handling extremely large datasets and training workloads for generative AI.
The systems are already being integrated into the UK’s National AI Research Resource, which aims to democratize access to advanced computing for academia, startups, and public sector institutions, a crucial step in ensuring that global breakthroughs in artificial intelligence remain accessible to UK-based innovation.
With Phase Two of Isambard-AI now operational, and Phase Two of Dawn expected to follow, Britain is staking a bold claim in the escalating global race for AI supremacy, one where compute access, not just algorithms, will define leadership.
And from the legacy of suspension bridges to the logic of neural networks, two historical titans, Brunel and Hopper, now leave their mark on a new kind of infrastructure, not just built to last, but built to learn.
