NVIDIA commits compute access valued at $250 million to UK AI safety research

NVIDIA has committed five years of compute access to support agentic safety and security research in the UK. The donation is compute access rather than a cash grant. NVIDIA values it at $250 million in total and describes the annual access as worth $50 million.

NVIDIA says most of the commitment is intended for the wider UK AI safety and security research community. A portion will support the British Open-ended Learning and Discovery Lab, or BOLD; the Science of Fundamental AI Research Lab, known as SOFAIR; and Imperial College London. The company has not disclosed how it will divide the access among those recipients or provided application rules, eligibility criteria or an allocation schedule.

The company said it will pay for researchers to use infrastructure hosted in the UK through UK-based NVIDIA Cloud Partners. It has not named the cloud providers or specified the hardware, technical architecture, commercial terms or start date. NVIDIA also has not explained how it calculated the $250 million valuation, including whether it reflects retail, partner, internal or another form of pricing.

The planned work includes secure AI agents, security for open models, mitigations for real-world threats, secure testing environments, reproducible research and benchmarks for safety assessments, according to NVIDIA. Those are intended uses and expected benefits, not demonstrated outcomes. The opened sources do not establish that compute access has begun or that it has improved AI safety or security.

BOLD is an Oxford-led national research programme involving the University of Oxford, University College London and Imperial. Oxford said in September that NVIDIA was already a collaborator and had provided academic grants, compute resources and engineering support. SOFAIR is hosted by UCL, works with the universities of Cambridge, Edinburgh and Oxford, and lists open-source model development, new architectures and distributed AI compute among its areas of work.

The commitment follows an October 12 UK government announcement about preparations for serious incidents involving autonomous AI agents and the AI Security Institute’s evaluations of advanced models. That announcement does not establish a government role in NVIDIA’s donation. NVIDIA has not disclosed governance terms, research-independence protections, publication requirements or other conditions attached to the access.