Nvidia commits $250M of compute to UK AI safety research

Nvidia has committed $250M over five years to UK agentic safety and security research, in the form of compute access rather than cash. Provider selection and commercial terms have still to be agreed.

Nvidia has committed $250M to British AI safety research over the next five years. The commitment is compute access, not cash, and none of it has been contracted.

It works out at $50M of compute a year, the company said, delivered through UK-based Nvidia Cloud Partners with the infrastructure hosted in Britain.

Nvidia says provider selection, technical architecture and commercial terms are all still to be agreed. The donation is valued at $250M rather than priced at it.

Most of the compute goes to the wider UK safety community. A proportion goes to BOLD at Oxford, SOFAIR at UCL, and Imperial College London, which will work on safety engineering and robotics.

American science got four times as much. Nvidia committed $1B over five years to US universities, quantum computing and government cloud providers, five days earlier.

Capability and safety must advance together, Jensen Huang said, adding that the UK brings world-class scientific rigour to what he called an essential engineering challenge.

The timing is a tour. Huang was in Paris on Tuesday saying Nvidia would announce $2B of European investments within a week, with London and Berlin to follow.

Whether this $250M counts towards that $2B is not stated, and nor is it clear how either figure is constructed. Nvidia has not published a breakdown of the $8B it says it has put into Europe over the past year.

AMD got there first. At London Tech Week in June it pledged up to £2B over five years, including direct stakes in British startups, while Nvidia pledged nothing.

The structure is worth noting. Nvidia is funding access to compute sold by its own cloud partners and running on its own chips.

Ministers welcomed it anyway. AI minister Kanishka Narayan called it a vote of confidence in British research and tied it to Britain’s G20 presidency, and business secretary Jonathan Reynolds said it would help British researchers lead.

The work itself is specific. It covers secure agents, open-model security and threat mitigation, using Nvidia’s own OpenShell runtime and its Sentry reference design.

It builds on existing ground. Bristol’s Isambard-AI supercomputer already runs on Nvidia Grace Hopper chips, and the findings are meant to inform safeguards across the wider computing ecosystem.