The U.K. government has unveiled plans for its groundbreaking £225 million ($272 million) AI supercomputer, set to become the country's fastest computing system. Named Isambard-AI, this powerful supercomputer will reside at the University of Bristol and will be fueled by advanced hardware from Nvidia and HPE. It will utilize HPE’s Cray EX supercomputers combined with an impressive 5,448 Nvidia Grace Hopper 200 Superchips, offering unrivaled capabilities for researchers in areas such as robotics, climate research, and drug discovery.
With a staggering performance of 21 exaflops, Isambard-AI will have the capacity to execute up to 200 quadrillion calculations per second, placing it among the elite in the world of supercomputing. This innovative machine is expected to be operational by the summer of 2024 and is named in honor of Isambard Kingdom Brunel, an iconic figure in engineering whose work shaped transportation and construction during the Industrial Revolution.
Professor Simon McIntosh-Smith, director of the Isambard National Research Facility at the University of Bristol, noted that Isambard-AI will rank within the top 10 fastest supercomputers globally and will be a significant asset for open science.
### Performance Comparison
On a comparative scale, Isambard-AI is projected to be ten times more powerful than the U.K.’s existing fastest supercomputer, ARCHER2, located at the University of Edinburgh, which operates at around 20 petaflops. To put this into perspective, 1,000 petaflops equal one exaflop, making Isambard-AI's 21 exaflops a monumental leap in computing performance.
According to Alexander Harrowell, principal analyst for advanced computing for AI, the capabilities of Isambard-AI are set to "crush the whole Top500 list" of supercomputers. The latest Top500 ranking, published in June, lists Frontier (also known as OLCF-5) at the top. This HPE-based system became operational in 2022 and can perform a quintillion calculations per second.
### Future Developments and Collaborations
In addition to Isambard-AI, the anticipated Dojo supercomputer from Tesla is rumored to go live in 2024, potentially delivering a remarkable 100 exaflops. A portion of the Dojo cluster became operational in August, featuring $300 million worth of Nvidia's H100 GPUs.
Justin Hotard, executive vice president and general manager of HPE's HPC, AI, and Labs, emphasized that Isambard-AI would provide the necessary performance and scalability for demanding AI projects. Additionally, Ian Buck, vice president of hyperscale and HPC at Nvidia, highlighted that the supercomputer would grant researchers access to state-of-the-art AI and HPC resources, fostering new scientific breakthroughs in the U.K.
### The Dawn of a New Era in Supercomputing
Isambard-AI will also connect with a new supercomputer cluster at the University of Cambridge, called Dawn. Set to launch as early as January 2024 and powered by Dell and StackHPC, Dawn will expand the computational capacity available for research. The government’s Frontier AI Taskforce will receive priority access to these advanced computing tools to support its initiatives in AI safety research.
The establishment of Isambard-AI and its connections to other computing resources, including Isambard-3—another supercomputer located in Bristol designed for climate science and medical research—highlights a significant investment in the U.K.'s AI research landscape. Local academic institutions, including the universities of Bath, Cardiff, and Exeter, will also benefit from access to these powerful computing systems, reinforcing collaboration and innovation across the region.
Ranking
