Nvidia Pledges $250 Million of Computing Power to UK AI Safety Research Over Five Years

Nvidia will donate $250 million of computing power to British AI safety and security research over five years, the company announced this week. The pledge works out to about $50 million of compute access a year, and it is compute, not cash (Nvidia newsroom, The Daily Star).

How the commitment works

Nvidia will fund access to computing through UK-based Nvidia Cloud Partners, with the infrastructure hosted in Britain. Most of the capacity is for the wider UK AI safety and security research community, with a portion going to three named groups. Provider selection, technical architecture and commercial terms are still to be agreed, so the headline figure describes a commitment rather than a running grant.

Named recipient	Led by	Focus
BOLD (British Open-Ended Learning and Discovery Lab)	University of Oxford	AI systems that learn, adapt and work alongside people
SOFAIR (Science of Fundamental AI Research Lab)	University College London	Open-source AI architectures, cheaper ways to build and deploy AI
Imperial College London	Imperial	Safety engineering and robotics

No split of the compute among them has been published. One outlet reported a different breakdown, including grants and a Cambridge lab, that does not match Nvidia's release.

What the research will cover

The compute is for what Nvidia calls agentic safety and security: building secure AI agents, hardening open models and countering real-world threats. The work includes secure sandboxes, reproducible research and safety benchmarks, using Nvidia OpenShell and the Nvidia Sentry reference design. It builds on existing UK collaboration including the Isambard-AI supercomputer at the University of Bristol, which runs on Grace Hopper Superchips.

Agents are the focus because they act on their own: they browse, run code and handle files, so testing needs sandboxes that limit what an agent can reach, and evaluating that behaviour at scale takes a great deal of computing power.

Nvidia chief executive Jensen Huang said "AI capability and safety must advance together." UK AI minister Kanishka Narayan said the contribution would help ensure future AI is safe and serves people's best interests, and Jakob Foerster, founding director of BOLD, said treating safety as an afterthought "is no longer appropriate."

Why Britain, and why now

The UK has no major frontier AI developer of its own and has instead built a role in testing and evaluating risk, mainly through its AI Security Institute. A pledge of compute fits that strategy, because evaluating powerful models needs the same expensive hardware used to train them (EdTech Innovation Hub).

The timing follows a UK parliamentary hearing on October 13 where Anthropic's European representative argued voluntary commitments are not enough and OpenAI's Tom Duff Gordon backed mandatory incident reporting, while Google DeepMind and Meta favoured existing regulation and technical standards. Some analysts note a potential conflict of interest in Nvidia funding safety research into a technology its chips power.

What is not yet known

The cloud partners, hardware specifications and five-year milestones are unannounced, and the dollar value is Nvidia's own. The release carries standard forward-looking-statement language.

What to watch next

Selection of UK cloud partners, the first capacity allocations to Oxford, UCL and Imperial, and any AI Security Institute follow-up.

Related on Gadgets365

• Nvidia's Jensen Huang says chip sales will double next year
• OpenAI fires three employees over an alleged leak to an outside AI safety group
• Anthropic commits 5% of its shares to a public benefit fund