It’s a bit unusual for someone to rebuke an investor in their AI company, but we live in interesting times.
Anthropic Chief Executive Officer Dario Amodei has issued a renewed warning over the US decision to allow Nvidia to resume sales of advanced AI chips to China, saying the move could carry what he called “incredible national security implications.”
Speaking at the World Economic Forum in Davos during an interview with Bloomberg Editor-in-Chief John Micklethwait, Amodei argued that shipping these processors would be “a big mistake.”
“It’s a bit like selling nuclear weapons to North Korea and [bragging that] Boeing made the casings,” he added.
In November 2025, Microsoft and Nvidia said they plan a combined investment of up to $15 billion in Anthropic. We’ll find out later if that’s still happening.
Amodei’s comments come as US President Donald Trump moves to ease a ban originally intended to prevent Beijing and its military from building cutting-edge AI capabilities using American technology. The decision has reopened a long-running debate in Washington over how to balance economic competitiveness for US firms with efforts to slow China’s progress in strategic technologies.
The warning is particularly notable because it comes from the head of one of the most influential AI research companies in the world. While many technology executives have urged policymakers to avoid constraints that limit growth, Amodei has taken the opposite position, repeatedly advocating for more aggressive safeguards against advanced AI systems and the hardware required to power them.
Policy shift allows China access to Nvidia’s H200
The revised approach represents a meaningful win for Nvidia, which has argued that prolonged restrictions could harm US interests by encouraging China to speed up the development of homegrown alternatives. Under the updated rule, the H200 processor introduced more than two years ago would become the most advanced AI chip legally available to Chinese customers.
While the H200 is no longer Nvidia’s top-tier product, it remains a powerful processor for training and running large AI models. Analysts and industry watchers view it as capable enough to support advanced development in areas ranging from consumer chatbots to military-linked applications, depending on how it is deployed and who ultimately controls access.
Nvidia has maintained that limiting Chinese customers’ access to US hardware could backfire, creating long-term competitors and reducing American influence in the global AI ecosystem. The company’s position reflects a broader argument from the semiconductor industry: restrictions may delay China, but they also incentivize Beijing to prioritize self-sufficiency.
Chips remain restricted, but stakes remain high
Even as the H200 becomes available again, Nvidia continues to sell its more advanced Blackwell generation in the US, while preparing to transition to an even faster chip family named after astronomer Vera Rubin. Both of those platforms will remain restricted for national security reasons, preserving the government’s effort to keep the most cutting-edge American hardware out of the Chinese market.
Still, Amodei’s remarks suggest he believes the distinction may not be enough. From his perspective, even chips that are a step below the frontier could materially improve China’s ability to develop advanced AI systems at scale. That scalability is a major part of what makes modern AI strategically valuable, since more compute power typically translates into stronger models, faster training cycles, and more sophisticated applications.
The debate is also expanding beyond Nvidia. Rival Advanced Micro Devices is seeking approval to sell its MI325X chip in China, underscoring broader efforts by US chipmakers to re-enter a major market after a period of tightening restrictions. For investors and policymakers alike, AMD’s move signals that the issue is not isolated to a single company, but part of a larger push by the semiconductor sector to preserve revenue streams tied to overseas demand.
Why Amodei sees a national security risk
Amodei has consistently urged the Trump administration to keep tight controls in place. He has argued that China remains behind in AI development partly because of the chip embargo, and that loosening those controls could remove one of the most significant barriers slowing Beijing’s progress.
In this view, making compute power available to geopolitical rivals increases the likelihood that AI could be used for surveillance, military modernization, disinformation operations, or other state-linked programs.
Amodei’s warning is a reminder that the AI boom is not only a market story, but also a geopolitical one, where the next move from Washington could reshape both supply chains and the pace of technological competition worldwide.
Anthropic is shaking things up in the intense AI race as it expands its Labs.