Jensen Huang: We will stop supplying chips to customers developing unsafe AI software; AI infrastructure shortage will persist for several years

NVIDIA CEO Jensen Huang stated that if a customer develops unsafe AI software, he will stop supplying chips to them. However, he also emphasized that the current market concern over AI safety risks is exaggerated, and the so-called AI bubble is similarly overstated.

Jensen Huang made these remarks during an exclusive interview with Bloomberg in London. He is touring Europe to promote the company’s upcoming event in Berlin. He noted that before cutting off supply, he would first communicate with the relevant customers and pointed out that NVIDIA’s technology is general-purpose, and there are limitations to the company’s judgment over how others use it.

At the same time, NVIDIA announced it will donate $250 million over five years to AI safety research in the UK, demonstrating a tangible commitment to this issue.

On Wednesday, NVIDIA and other chip stocks fell sharply. Jensen Huang reiterated his optimistic outlook on demand for AI infrastructure, stating that the imbalance between supply and demand will persist for years. “Supply and demand will ultimately balance, but definitely not today,” he said at the London event.

NVIDIA is the dominant supplier in the AI accelerator chip market, with its products serving as core hardware for training and running AI models—a position that puts it in the spotlight of the AI safety debate.

When pressed by outsiders, Jensen Huang made it clear that if a customer is determined to be developing irresponsible AI, he is willing to stop supplying—“This is hypothetical, but the answer is yes.”

However, he also set boundaries for NVIDIA’s involvement. “We don’t run other people’s companies, and there are limits to our judgment on how other companies use our technology; our technology is general-purpose.” He said there would be dialogue and communication before cutting off supply.

This statement reflects that, as a key hardware provider, NVIDIA is seeking a balance between industry responsibility and business neutrality when facing controversies over AI safety.

Regarding calls from some AI executives to slow down industry development as a whole—for example, the sector-wide slowdown suggestion by Anthropic CEO Dario Amodei—Jensen Huang takes a different view.

He characterizes current security concerns as “growing pains” of a rapidly expanding industry and believes AI software can be effectively managed and safely deployed.

“Many young companies are experiencing the pains of growth, and they are growing extremely rapidly,” he said. “As companies grow and mature, the corresponding sense of responsibility and accountability will also gradually mature.”

He also pointed out that some concerns are “self-imposed,” and that the whole industry has done “a terrible job” in explaining the benefits of new technology to the public.

As for concrete approaches to address safety risks, Jensen Huang proposed using technology to counter technology: AI itself can be a tool for monitoring and regulating AI.

“We should accelerate the integration of these control technologies into AI systems, and we should fast-track the development of monitoring technology—and AI is that monitor. So, we should accelerate AI technology,” he said.

To support this direction, NVIDIA announced a donation to UK AI safety research, with funding provided via access to advanced computing power.

Multiple academic institutions in the UK, including the University of Oxford and University College London, will participate. The research will cover AI safety protection measures and methods for regulating AI agents.