The U.S. administration recently reversed an earlier ban, officially approving the sale of Nvidia’s H200 chips and a chip line by AMD to approved Chinese customers. These processors, though perhaps not the most advanced, are high-performance components crucial for artificial intelligence, making their export a contentious issue. The decision drew significant criticism from Anthropic CEO Dario Amodei at the World Economic Forum in Davos, a critique notably pointed given that Nvidia, one of the chipmakers involved, is a major partner and investor in Anthropic.

Amodei expressed incredulity regarding the new rules, stating, “The CEOs of these companies say, ‘It’s the embargo on chips that’s holding us back.’” He warned that this decision would ultimately harm the U.S., emphasizing, “We are many years ahead of China in terms of our ability to make chips. So I think it would be a big mistake to ship these chips.”

Painting an alarming scenario, Amodei highlighted the “incredible national security implications” of advanced AI models that he described as “essentially cognition, that are essentially intelligence.” He conceptualized future AI as a “country of geniuses in a data center,” asking his audience to imagine “100 million people smarter than any Nobel Prize winner,” all potentially under the control of a single nation. This vivid imagery underscored his deep concern about the implications of chip exports.

The most striking aspect of Amodei’s criticism was his stark analogy: “I think this is crazy,” he declared about the administration’s move, adding, “It’s a bit like selling nuclear weapons to North Korea and [bragging that] Boeing made the casings.” This comparison, delivered publicly, sent shockwaves through the industry, particularly impacting Nvidia, which supplies the crucial GPUs that power Anthropic’s AI models across various cloud providers.

Nvidia’s financial relationship with Anthropic is substantial, involving an investment of up to $10 billion, announced just two months prior alongside a “deep technology partnership” focused on mutual optimization. Amodei’s public comparison of his partner to an arms dealer raises questions about his motivations. It could have been an unguarded moment, perhaps swept up in his own rhetoric. However, given Anthropic’s strong market position, having raised billions and valued in the hundreds of billions, and with its Claude coding assistant being a top-tier tool, it is more probable that he felt confident in expressing such strong opinions.

It’s also plausible that Anthropic harbors genuine fears about Chinese AI advancements and sought to use provocative language to gain Washington’s attention. Nuclear proliferation comparisons are indeed an effective way to command notice. Perhaps the most remarkable takeaway is Amodei’s apparent fearlessness in delivering such a impactful statement on a global stage without evident concern for potential adverse effects on his business. While news cycles move swiftly and Anthropic is currently on solid footing, this incident suggests that for leaders in the AI sector, the stakes of the AI race have become so existential that conventional constraints like investor relations, strategic partnerships, or diplomatic niceties no longer apply. Amodei’s uninhibited candor, more than any specific statement, highlights a significant shift in the discourse surrounding AI development and its geopolitical implications.