
Share
Amodei warns that allowing sales of powerful AI chips to China could have severe security implications, likening it to handing over nuclear weapons-a stark warning in the escalating tech cold war.
Dario Amodei, CEO of AI research firm Anthropic, has expressed strong opposition to the US government's decision to allow Nvidia to sell H200 accelerators to Chinese companies. In a recent interview at the World Economic Forum in Davos, Switzerland, Amodei likened this move to providing nuclear weapons to an adversary.
The debate over AI export controls is intensifying as the US and China vie for dominance in advanced technology. The Trump administration's decision to permit Nvidia to sell H200 GPUs to Chinese customers, with a 25% revenue cut for the US government, has drawn criticism from figures like Amodei who argue that such actions could undermine US technological leadership.
Amodei's concerns are rooted in the belief that access to these advanced chips would significantly enhance China's ability to develop and deploy competitive AI models. "The CEOs of these [Chinese] companies say it's the embargo on US chips that's holding them back," Amodei stated. "I think it's a big mistake to ship these chips."
Technological Decoupling: Stricter export controls could lead to a technological decoupling between the US and China, which is home to approximately half of the world's AI researchers. This could stifle innovation and collaboration in the global AI community.
Competitive Disadvantage: If Chinese companies gain access to advanced GPUs like the H200, they could develop more sophisticated models that compete with Western counterparts, particularly in enterprise adoption. Models from companies like DeepSeek are already open-source, offering enterprises greater control over their data and reducing concerns about data privacy.
Security Concerns: Amodei's analogy of providing nuclear weapons to an adversary underscores his concern that advanced AI capabilities could be misused. The US currently leads in advanced semiconductor technology, and export controls help maintain this advantage by slowing China's technological progress.

Maintaining Leadership: By restricting the export of advanced GPUs, the US can preserve its lead in semiconductor technology. As computing power continues to double every two years, maintaining a technological edge is crucial for long-term dominance.
Innovation and Collaboration: Despite the risks, there are opportunities for innovation and collaboration. The US could work with China on specific AI projects that align with mutual interests, fostering a more cooperative global AI ecosystem.
Regulatory Balance: Finding a balance between strict export controls and open collaboration is essential. Policies should aim to protect national security while encouraging international cooperation in areas where it is beneficial.
Not all industry leaders share Amodei's hawkish stance. Chipmakers like AMD and Nvidia have warned that cutting off China from advanced technology could lead to a technological decoupling, which would be detrimental to both sides. "Closing the door on China would result in a technological decoupling," said a spokesperson for Nvidia.
Anthropic has long advocated for stricter controls on AI exports. In a statement released last spring, the company argued that export controls capitalize on the trend of computing power doubling every two years. "While US chip technology continues advancing, China's progress is slowed," Anthropic stated.
The decision to allow Nvidia to sell H200 GPUs to Chinese companies is a contentious one with significant implications for global AI development. While Amodei's concerns about security and competitive disadvantage are valid, the potential benefits of maintaining technological leadership and fostering international collaboration must also be considered. As the debate continues, policymakers will need to strike a delicate balance between protecting national interests and promoting innovation.
Tags
Original Sources
About the author
Marcus began tracking AI's market implications in 2016, noticing AI-related patent filings accelerating ahead of earnings upgrades before most of the sell-side had caught on. A former fixed-income quantitative analyst, he spent two decades building models that priced risk across emerging markets before pivoting to cover the economic impact of AI full-time. His writing translates opaque technical developments into clear risk/reward terms — and he's rarely diplomatic about the gap between AI valuations and underlying fundamentals. He believes most market participants still underestimate AI's long-run deflationary effect on knowledge work.
More from The Analyst →This Week's Edition
21 January 2026
88 articles
Related Articles
Related Articles
More Stories