Huang: Nvidia May Cut Chips to Unsafe AI Developers

According to reports, Nvidia CEO Jensen Huang said the company would stop supplying chips to customers developing unsafe artificial intelligence (AI) software, although he believes current concerns about the technology may be overstated.

When asked whether Nvidia would cut off access to its chips for irresponsible AI developers, Huang replied, “It’s hypothetical, but the answer is yes.” However, he emphasized that communication should come before such extreme measures. Nvidia does not operate other companies’ businesses and has limited control over how its general-purpose technology is used.

Nvidia is a leading provider of AI accelerators, specialized chips used to train and run AI models. Its position has placed the company at the center of growing debates over whether increasingly powerful AI systems could pose risks or operate beyond human control. Some AI executives, including Anthropic CEO Dario Amodei, have called for the industry to slow its development pace to address these concerns.

During a European tour ahead of a company event in Berlin, Huang described AI safety concerns as growing pains for a rapidly expanding industry. He expressed confidence that AI software can be controlled and deployed safely, while acknowledging that companies must take greater responsibility as they grow.

As part of its AI safety efforts, Nvidia plans to donate $250 million over five years to support research in the United Kingdom. The contribution will take the form of access to advanced computing resources. Researchers at institutions including the University of Oxford and University College London are expected to study safeguards and methods for controlling AI agents.

“Many young companies are going through growing pains, and they are growing at an extraordinary pace,” Huang said. “As companies grow, responsibility and obligations come with that growth.”

Huang also argued that some concerns about AI are self-imposed and criticized the industry for doing a poor job of explaining the technology’s benefits. He added that AI itself could help improve the safety and control of AI systems.

Huang reiterated his view that fears of an AI bubble may be exaggerated, even as Nvidia and other semiconductor stocks led a broader decline in U.S. equities. He maintained that demand for AI infrastructure is likely to continue outstripping the supply of components over the next several years.

“Eventually, supply and demand will come into balance, but not yet,” Huang said at an event in London.