Anthropic CEO Dario Amodei criticized the U.S. government's decision to allow Nvidia to sell H200 GPUs to Chinese companies, comparing it to giving nuclear weapons to an adversary. He believes this move undermines U.S. technological leadership and could strengthen Chinese AI capabilities. The Trump administration's policy permits these sales as long as the U.S. receives a 25 percent cut from revenues. Amodei argues that Chinese firms, claiming U.S. chip embargoes hinder their progress, would benefit significantly from access to advanced chips, which would level the playing field against Western competitors. Despite his concerns, Amodei's assessment of Chinese AI development suggests that they have not made substantial progress. He points out that while companies like DeepSeek produce models that perform well on benchmarks, they may not translate to real-world effectiveness. The availability of open weights models from Chinese developers presents a challenge, as these models can be easily downloaded and run on local hardware, giving enterprises more control over their data compared to American models, which are typically locked behind APIs. Anthropic remains focused on competing against major players like OpenAI and Google for enterprise contracts, while noting that they rarely lose contracts to Chinese models. However, Amodei's use of "almost" implies that Chinese models are starting to find some traction in the market. Anthropic has not released any of its flagship models publicly as open weights but has formed collaborations with select partners, such as Amazon Web Services. The dynamics of AI model deployment and the race for technological supremacy between the U.S. and China continue to evolve amidst these tensions.