The AI Research Resource (AIRR) is a suite of advanced computers that provides world-leading, AI-specialised computing capacity to:<\/p>\n
public researchers
\n academia
\n small and medium size enterprises<\/p>\n
The\u00a0AIRR aims to address the acute under-provision of specialised public compute for AI users in the research and development (R&D) community. This is crucial for the acceleration of AI innovation and the government\u2019s growth agenda.\u00a0<\/p>\n
It is made up of \u202fAI-optimised, high-performance computing (HPC) platforms that support ambitious research projects, and provide access to advanced hardware, software and expert support.<\/p>\n
These projects:<\/p>\n
represent significant advancements in the UK\u2019s AI infrastructure
\n provide cutting-edge\u202fcomputational resources for researchers and start-ups
\n foster collaboration across various growth sectors such as healthcare and clean energy evidenced\u202f by the use cases for AIRR Phase 1<\/p>\n
Background\u00a0<\/p>\n
The AIRR programme intends to address the significant shortage of publicly available computing resources in the UK. In January 2025, the government announced expanding AIRR capacity, by at least 20x by 2030, as part of the AI Opportunities Action Plan<\/a>.\u00a0\u00a0<\/p>\n The government has committed to spending an extra \u00a31 billion to scale up our compute power by a factor of 20, giving Britain the power to become an AI leader.<\/p>\n AIRR compute clusters<\/p>\n The government is investing significantly in the Isambard-AI and Dawn AIRR clusters and will have invested over c.\u00a3350million by 2030.<\/p>\n Isambard-AI\u00a0(University of Bristol)<\/p>\n The Isambard-AI facility is the UK\u2019s most powerful public compute facility. It is made up of 5,448 Nvidia GH200 Grace-Hopper superchips (supplied by HPE) and operated by the University of Bristol at Bristol\u2019s National Composites Centre.<\/p>\n Further information on the system can be found on the Isambard website<\/a>.<\/p>\n Dawn (University of Cambridge)<\/p>\n The Cambridge Dawn facility is made up of 1,024 Intel Data Centre GPU Max 1550 GPUs. It is a close partnership between Cambridge, Intel and Dell at the Cambridge Open Zettascale Lab.<\/p>\n Further information on the system can be found on the Dawn website<\/a>.<\/p>\n Apply to use AIRR<\/p>\n 2 access routes are currently available:<\/p>\n Additional routes for larger-scale projects will be introduced as the AIRR service evolves.<\/p>\n All projects must:<\/p>\n be conducted in the UK Access to the compute capability will be provided in accordance with the Subsidy Control Act 2022.<\/p>\n Rapid Access<\/p>\n Open to UK-registered micro, small and medium organisations, this route supports early-stage development of AI products prior to market launch.<\/p>\n Eligible activities include:<\/p>\n feasibility studies Resources available:<\/p>\n Up to 20,000 GPU hours on AIRR services, to be used within 3 months of project start.<\/p>\n Read further guidance by UKRI on the Rapid Access route<\/a>.<\/p>\n Gateway<\/p>\n Designed for researchers from academia, industry, or other UK organisations, this route supports:<\/p>\n first-time users of AI supercomputing resources Eligible activities include:<\/p>\n fundamental research (Research organisations only) Resources available:<\/p>\n Up to 10,000 GPU hours on AIRR services, to be used within 3 months of project start.<\/p>\n Read further guidance by UKRI on the Gateway route<\/a>.<\/p>\n Apply via the AIRRPortal<\/p>\n You can use the AIRRPortal to apply to use AIRR.<\/p>\n Apply to use AIRR<\/a><\/p>\n Read further guidance by UKRI on how to use the AIRRPortal<\/a>.<\/p>\n Accessibility and inclusion<\/p>\n We welcome and encourage applications from people of all backgrounds and are committed to making our application process accessible to everyone.<\/p>\n Contact us at airr@ukri.org<\/a> to ask for guidance.<\/p>\n Privacy notice<\/p>\n
\n aim to deliver benefits within the UK.<\/p>\n
\n industrial research
\n experimental development<\/p>\n
\n testing of (novel) algorithms, code, and workflows
\n benchmarking of algorithms, code, and workflows before applying for larger AIRR opportunities<\/p>\n
\n feasibility studies
\n industrial research
\n experimental development<\/p>\n