TLDR
- Anthropic CEO Dario Amodei publicly criticized US policy allowing AI chip exports to China at Davos
- He compared selling Nvidia H200 chips to China as equivalent to selling nuclear weapons to North Korea
- This came despite Nvidia investing $10 billion in Anthropic just two months earlier
- Anthropic admitted losing some business contracts to Chinese open-weight AI models
- The conflict reveals deep divisions in the tech industry over AI hardware export controls
Anthropic CEO Dario Amodei delivered a scathing critique of US chip export policy at the World Economic Forum in Davos. His target was the recent decision allowing Nvidia to sell advanced AI processors to China.
The remarks came just two months after Nvidia announced a $10 billion investment in Anthropic. The timing made Amodei’s comments especially striking to industry observers.
“I think it’s a bit like selling nuclear weapons to North Korea and bragging,” Amodei told Bloomberg. He described the policy as “crazy” during his interview.
The new US policy permits Nvidia and AMD to sell certain AI chips to China. This includes Nvidia’s H200 processor line, which was previously banned from export.
Why Anthropic Opposes the China Deal
Amodei expressed concern about AI capabilities in the hands of authoritarian governments. He described AI models as “essentially cognition, that are essentially intelligence.”
He asked his audience to imagine these models as “100 million people smarter than any Nobel Prize winner.” This power concentrated in one country’s control worried him deeply.
Chinese companies are releasing open-weight AI models that anyone can use for free. These models reveal trained parameters but not the underlying source code or training data.
Amodei admitted that Anthropic has lost business to Chinese models. “I have almost never lost a deal, lost a contract to a Chinese model,” he said.
The word “almost” confirms some contracts went to Chinese competitors. He suggested the real impact might be larger since companies can test Chinese models without contacting Anthropic.
The Nvidia Investment Complicates Things
Nvidia invested $10 billion in Anthropic in November 2025. The company also pledged to optimize Anthropic models for better performance and efficiency.
Anthropic trains its AI models using hardware from multiple sources. This includes Nvidia GPUs, Amazon Trainium chips, and Google TPUs.
The company counts several tech giants as major investors. Amazon invested up to $8 billion, Google put in $3 billion for roughly 14% ownership, and Microsoft committed up to $5 billion.
Amodei’s criticism risks straining the relationship with Nvidia. The chipmaker remains essential to AI model development across the industry.
Industry Tensions Over Export Controls
The controversy highlights growing friction in the tech sector. Chipmakers want to expand their markets while AI developers worry about national security risks.
The US Commerce Department claims sales to China undergo rigorous vetting. Buyers are screened for connections to military operations.
Critics argue enforcement remains weak due to front companies and complex business relationships. These structures can hide the true end users of advanced chips.
Nvidia’s architecture has become foundational for training large AI models. Few alternatives exist though AMD and Intel are working to compete.
Amodei did not explicitly name China during his Davos appearance. However, the context made his target clear to everyone present.
The dispute shows how AI hardware exports have become a flashpoint in global tech competition. What happens with these chips affects how quickly new AI capabilities emerge worldwide.
The US policy change allowing H200 chip sales to China took effect after years of export restrictions.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
