Anthropic CEO Dario Amodei used a stage at the World Economic Forum in Davos to make an unusually sharp argument against allowing high-performance AI chips to be sold to approved Chinese customers.
His target was not only the U.S. administration, which had reversed an earlier ban, but also the chip companies involved. That made the moment especially notable: Nvidia is not just central to the AI industry, it is also a major partner and investor in Anthropic.
The chip decision behind the dispute
Last week, after reversing an earlier ban, the U.S. administration officially approved the sale of Nvidia's H200 chips, along with a chip line by AMD, to approved Chinese customers.
The source article notes that these are not necessarily the chipmakers' newest or most advanced products. But they are high-performance processors used for AI, which is why the export decision became controversial.
Amodei responded at Davos on Tuesday during an interview with Bloomberg's editor-in-chief. Asked about the new rules, he challenged the idea that export restrictions were the main obstacle for chip companies.
"The CEOs of these companies say, 'It's the embargo on chips that's holding us back,'"
Amodei said the move could harm the U.S. over time. His argument rested on a simple claim: the U.S. has a large lead in chipmaking capability, and shipping these processors would weaken that advantage.
"We are many years ahead of China in terms of our ability to make chips,"
He followed that by saying:
"So I think it would be a big mistake to ship these chips."
Why Amodei sees AI chips as a security issue
Amodei's objection was not framed as a narrow business disagreement. He connected AI processors to national security and to the future power of AI systems themselves.
He described the "incredible national security implications" of AI models that represent "essentially cognition, that are essentially intelligence." In his view, the hardware used to train and run advanced AI systems matters because it can help determine who controls powerful models.
To explain the scale of what he believes is at stake, Amodei compared future AI capacity to a "country of geniuses in a data center." He asked the audience to imagine "100 million people smarter than any Nobel Prize winner," under the control of one country or another.
That image was meant to show why chip exports are not just another commercial policy question. If AI systems become a strategic asset, then the processors behind them become part of the strategic contest too.
Amodei then made his strongest comparison, saying of the administration's latest move:
"I think this is crazy,"
He added:
"It's a bit like selling nuclear weapons to North Korea and [bragging that] Boeing made the casings."
The analogy was deliberately severe. It put the sale of AI processors in the same rhetorical category as proliferation, and it placed chipmakers in an uncomfortable role inside that comparison.
The Nvidia relationship makes the criticism more striking
The comments landed with extra force because of Anthropic's ties to Nvidia. Anthropic runs on the servers of Microsoft and Amazon and Google, according to the source article. But Nvidia alone supplies the GPUs that power Anthropic's AI models, and every cloud provider needs Nvidia's GPUs.
Nvidia also recently announced it was investing in Anthropic to the tune of up to $10 billion. Just two months ago, the companies announced that financial relationship along with a "deep technology partnership." The partnership was described with promises that the companies would optimize each other's technology.
That context makes Amodei's Davos remarks unusual. A CEO criticizing policy is common enough. A CEO sharply criticizing a decision that benefits a major partner and investor is different.
The source article suggests several possible explanations. Amodei may have been speaking in an unguarded moment. He may also have felt confident because of Anthropic's strong position in the AI market.
Anthropic has raised billions, is valued in the hundreds of billions, and its Claude coding assistant has developed a reputation as a highly beloved and top-tier AI coding tool, particularly among developers working on complex, real-world projects.
What the moment says about the AI race
Amodei's remarks also point to a broader shift in how some AI leaders appear to view the stakes around advanced models, GPUs, export controls, and China.
According to the source article, it is entirely possible that Anthropic genuinely fears Chinese AI labs and wants Washington to act. In that reading, the intense analogy was not accidental noise but an attempt to make policymakers pay attention.
The more remarkable part may be that Amodei seemed willing to say it publicly despite the business relationships involved. The usual concerns around investor relations, strategic partnerships, and diplomatic niceties appeared secondary to the argument he wanted to make.
That is why the Davos exchange matters beyond one quote. It showed a leading AI company executive treating chip exports as an existential policy issue, even when doing so risked tension with a company that sits at the center of Anthropic's own infrastructure.
The debate over Nvidia H200 chips, AMD processors, approved Chinese customers, and U.S. export policy is therefore also a debate over who gets access to the hardware behind the next generation of AI. Amodei's message was blunt: if AI is becoming a form of national power, then shipping the processors that enable it is not just a commercial decision.