Nvidia CEO Says He Would Cut Off Customers Who Build Unsafe AI

AI research compute costs drop by up to $50 million annually for UK institutions and developers.

NVIDIA has announced a five-year commitment to donate $250 million in compute resources to advance agentic safety and security research in the United Kingdom. This initiative, totaling $50 million annually, will be distributed through U.K.-based NVIDIA Cloud Partners to support the broader AI research community. A portion of the funding will specifically aid research at BOLD (British Open-Ended Learning and Discovery Lab), SOFAIR (Science of Fundamental AI Research Lab), and Imperial College London. The research will focus on foundational models, robotics safety, transparent AI, and human-AI co-improvement, aiming to develop safeguards for increasingly powerful AI systems. The compute will be delivered via NVIDIA Cloud Partners, fostering the U.K.'s domestic compute ecosystem and research capabilities.

This significant investment by NVIDIA underscores the growing industry focus on AI safety and security as a critical component of advanced AI development, particularly for agentic systems operating in the real world. By providing substantial compute resources, NVIDIA is enabling U.K. researchers to tackle complex challenges in AI alignment, governance, and human-AI interaction, positioning the U.K. as a potential leader in responsible AI innovation. The strategy of routing compute through local cloud partners also strengthens the national AI infrastructure and ecosystem.

The initiative signals a long-term commitment from a major AI hardware provider to influence the direction of AI safety research, potentially setting standards and best practices for agentic AI development. This could lead to a more robust and trustworthy AI ecosystem globally, while also benefiting NVIDIA by fostering a secure environment for the adoption of its future technologies. Researchers and developers in the U.K. will have unprecedented access to cutting-edge resources, accelerating their work in building safer, more capable AI systems.