Nvidia CEO Jensen Huang says company will halt chip sales to unsafe AI developers

Nvidia chief executive Jensen Huang has said the silicon giant will stop supplying high-performance processors to any buyer found to be developing dangerous or unchecked artificial intelligence systems.

Speaking at an event in London, Huang stated that while Nvidia manufactures general-purpose technology and cannot directly govern how clients build software, it is prepared to cut off hardware deliveries to customers who deploy AI recklessly. He noted that any action would start with direct discussions, but confirmed that halting processor sales remains a necessary safeguard if safety boundaries are breached.

Funding British safety research

To reinforce guardrails around advanced software, Nvidia announced a $250 million grant spread over five years to support British research institutions. Rather than cash funding, the package provides high-performance computing time to universities including the University of Oxford and University College London.

Academic teams will use the processing capacity to study software containment mechanisms and safety controls for autonomous AI agents. Huang argued that the most practical way to manage technological risks is to use artificial intelligence itself as a monitoring tool. Accelerating the development of AI-driven containment technology, he argued, will help supervise software performance far more effectively than artificial delays.

Rejection of industry-wide pauses

The approach places Nvidia at odds with executives in the sector, such as Anthropic chief executive Dario Amodei, who have suggested a coordinated slowdown across major developers to evaluate safety risks. Huang rejected calls for a voluntary pause, describing public anxiety over AI capabilities as the typical growing pains of a rapidly scaling industry. He added that the sector has done a poor job explaining the technology's real-world benefits to the public.

Huang also dismissed financial concerns surrounding heavy industry spending, rejecting claims that investment in hardware represents a market bubble. Speaking during a market downturn for tech stocks, he stated that while supply and demand for AI processors will eventually harmonise, global demand currently continues to outstrip available hardware output.

Nvidia's control over the primary processors used to train and run complex generative models ensures its policies will heavily influence how safety standards are enforced across the industry.