UKRI opens 2 million GPU hours to power UK AI startups
This article covers UK Research and Innovation (UKRI), the UK research funding agency, opening a new Innovator route on 6 November 2025 that gives UK researchers access to large blocks of GPU time on the Isambard‑AI and Dawn supercomputers, with applicants able to request between 50,000 and 150,000 GPU hours for six‑month, resource‑only projects. The scheme aims to scale national AI compute capacity and support UK researchers, AI startups and cross‑sector teams undertaking multidisciplinary, compute‑intensive AI R&D.
The UK research agency UK Research and Innovation has opened a new “Innovator” route giving UK researchers access to large blocks of GPU time on the Isambard‑AI and Dawn supercomputers. Applicants can bid for between 50,000 and 150,000 GPU hours for six‑month projects, resource only, no cash, as part of a push to scale national AI compute capacity and enable more ambitious, multidisciplinary AI R&D across the UK.
Why it matters
Compute bottlenecks are one of the biggest constraints on advanced AI research. This Innovator call, run under the AIRR programme, aims to move beyond small pilot allocations and let teams test bigger ideas at scale. The scheme sits alongside the government’s wider plan to increase publicly available compute by at least 20 times by 2030, backed by a stated additional £1 billion commitment, and contributes to the UK Compute Roadmap’s objective of a diverse and accessible compute ecosystem.
Inside the offer
- Who runs it: UK Research and Innovation (UKRI) on behalf of the AIRR programme, with support from several UK research councils and government departments. Funders listed include AHRC, BBSRC, ESRC, EPSRC, Innovate UK, MRC, NERC and STFC.
- What’s on the table: an indicative pool of 2,000,000 GPU hours across the two AIRR services. Individual projects may request 50,000 to 150,000 GPU hours to be used over six months.
- Money: this route provides compute allocations only. There is no direct grant or cash funding attached.
- Timing: the application window opens 6 November 2025 at 9:00am UK time and closes 16 January 2026 at 4:00pm. Projects are expected to start from mid February 2026 and run for six months.
Who can apply
Eligible leads must be UK researchers employed by an eligible organisation: a UK registered company, a UKRI‑eligible research organisation, charity or not‑for‑profit. Project leads must already hold a contract of longer duration than the proposed project. Research technical professionals, such as research software engineers, are explicitly eligible to lead applications under the same terms as academic researchers.
Collaborative and cross‑sector proposals are encouraged, and there is no limit on the number of applications an organisation may submit. However, an individual can be project lead on only one Innovator application.
Technical prerequisites and staged access
UKRI expects applicants to have prior experience using AIRR through Gateway or Rapid Access awards. If a team has not previously used AIRR, UKRI will treat the first month of any Innovator award as an assessment period. The full block of GPU hours will not be released until the project has demonstrated feasible and scalable use of AIRR resources, for example by consuming at least 10,000 GPU hours during that initial phase.
How applications will be assessed
Applications go through an eligibility and remit check followed by peer review. Each Innovator submission will be reviewed by two experts from the AI and high performance computing community. Assessors judge projects on scope, potential to advance AI research and innovation, contribution to capacity building in the AI ecosystem, justification of requested resources, and consideration of ethical and responsible research issues.
Resource allocation uses a three‑tier system. Top tier applications are prioritised for allocation; middle tier projects may receive resources using partial randomisation and allocations continue until the GPU budget is exhausted. Only proposals that pass a competitiveness threshold will be considered for randomisation. UKRI and the Department for Science, Innovation and Technology will make final allocation decisions.
The machines and the money behind them
The Innovator route grants access to two purpose‑built UK facilities.
- Isambard‑AI at the University of Bristol is described as the UK’s most powerful public compute facility, comprising 5,448 Nvidia GH200 Grace‑Hopper superchips supplied by Hewlett Packard Enterprise and operated at Bristol’s National Composites Centre.
- Dawn at the University of Cambridge comprises 1,024 Intel Data Centre GPU Max 1550 devices and is a partnership between Cambridge, Intel and Dell based at the Cambridge Open Zettascale Lab.
The government has committed major capital investment in these clusters; UKRI states that more than £350 million will be invested in Isambard‑AI and Dawn by 2030 as part of the broader scaling effort.
What kinds of projects are sought
UKRI lists a wide set of AI R&D activities eligible for access: development of new algorithms and software, exploration of AI‑assisted workflows, AI‑driven data collection and synthesis, and early‑stage product development. Projects may range across fundamental research, feasibility studies, industrial research and experimental development. Proposals that align with the government’s five missions — growing the economy, modernising the NHS, safer streets, opportunity for all, and clean energy leadership — are of particular interest.
Rules, governance and ethical safeguards
Awards are made in line with the Subsidy Control Act 2022 where relevant, and projects must follow UKRI’s trusted research and innovation principles for due diligence on UK and international collaboration. Applications require supporting documentation on ethics and subsidy history. UKRI and DSIT reserve the right to pause access to AIRR for high priority national requirements, with prior notice where possible.
How to apply and practicalities
Applications are submitted through the AIRRPortal. The online form includes basic project details and uploads for supporting documents. Templates for the project details form and assessment questions are available on the portal. Applicants should ensure they meet eligibility criteria and include the required documents; incomplete or non‑compliant submissions may be rejected. For guidance, applicants can email airr@ukri.org.
What this means for the UK AI ecosystem
Providing larger, predictable blocks of public compute aims to lower the entry barrier for ambitious AI experiments and foster collaborations across academia, industry and the public sector. By shifting from small pilot allocations to multi‑month compute packages, the Innovator route seeks to help UK teams translate early ideas into demonstrable scaling behaviour, and to build skills and capability that are portable across institutions.
If uptake matches the resources available, the scheme could accelerate UK research that requires sustained GPU throughput while signalling a long‑term public commitment to national compute infrastructure. In the context of international competition for AI talent and infrastructure, access to sizeable public GPU clusters may be an important enabler for UK researchers and startups seeking to test compute‑intensive models without immediate commercial cloud spend.
Click here for a full list of 7,589+ startup investors in the UK