With the Bristol Isambard-AI supercomputer now live, the UK government has unveiled a roadmap to delivering 420 Exaflops of compute power by 2030.
The £1bn set aside in the spending review is being used to increase the UK’s compute infrastructure and drive forward artificial intelligence (AI) development. The government’s goal is to reduce the UK’s reliance on foreign computing power, using sovereign compute capabilities to power transformation of public services and drive economic growth.
The Department for Science, Innovation and Technology (DSIT) said the roadmap also builds on the ambition of the 10-year infrastructure strategy and the modern industrial strategy to put the government’s vision into action – increasing investment and growing the industries of the future.
Chancellor Rachel Reeves said: “As technology advances, our plan for change is ensuring we are ahead of the curve, expanding our sovereign AI capabilities so we can make scientific breakthroughs, equip businesses with new tools for growth and create new jobs across the country.”
The strategy involves expanding the UK’s AI Research Resource (AIRR) 20-fold over the next five years. The system, delivered in partnership with UKRI, Nvidia, HPE, Dell Technologies and Intel, brings together supercomputers Isambard-AI in Bristol and Dawn in Cambridge.
In June, the Bristol Centre for Supercomputing unveiled its 5MW AI supercomputer facility for the 5,280 graphics processing unit (GPU)-powered Isambard-AI phase 2 system. The HPE Cray EX4000 supercomputer delivers over 21 ExaFLOP/s of 8-bit floating point performance for large language model training, and over 250 PetaFLOP/s of 64-bit performance. It integrates two all-flash storage systems: a 20 PiByte Cray ClusterStor and 3.5 PiByte Vast storage.
Unlike traditional supercomputers, Isambard-AI is designed to support users used to running GPUs in the cloud and offer access via Jupyter notebooks, MLOps and other web-based, interactive interfaces.
Technology secretary Peter Kyle formally launched the Bristol facility today. When the AIRR’s planned expansion is complete in the coming years, DSIT said it would be vastly more powerful than the world’s current leading supercomputers.
“Britain has top-of-the-class talent in AI, and our plan will put a rocket under our brilliant researchers, scientists and engineers – giving them the tools they need to make Britain the best place to do their work,” he said.
“This will mean we can harness the technology in Britain to transform our public services, drive growth and unlock new opportunities for every community in the country.”
University College London (UCL) researchers are already using Isambard-AI to line up pioneering AI tools that could revolutionise NHS cancer screening. Using prostate cancer as its initial test case, the UCL researchers are using the new supercomputer harnessing the system to develop one of the first scalable AI models dedicated to medical imaging – using AI to analyse magnetic resonance imaging scans and identify patients in need of treatment sooner.
The Dawn supercomputer in Cambridge is also being used for a research project funded by Cancer Research UK, assessing AI’s ability to analyse computerised tomography scans for kidney cancer compared with radiologists.
