Anthropic CEO Dario Amodei has dismissed concerns that Chinese AI company DeepSeek poses a significant threat to U.S. AI firms, calling such claims “greatly overstated.” In an essay posted Wednesday, Amodei criticized the reaction to DeepSeek’s recent AI models, particularly the stock drop of chipmaker Nvidia, which he found puzzling.
He argued that while DeepSeek’s V-3 model appears competitive in some tasks, it only matches the performance of U.S. models that were developed seven to ten months earlier. He also emphasized that DeepSeek’s progress follows an expected trend in AI cost reduction and that its overall spending is similar to that of U.S. AI labs.
Regarding DeepSeek’s R1 model, which recently gained public attention and impacted Nvidia’s stock, Amodei noted that multiple companies are capable of producing similar models.
However, he predicted that as AI technology scales further, only a few leading firms will maintain a competitive edge. A recent Bank of America report also questioned DeepSeek’s claim of training a foundation model for just $5.58 million, arguing that the figure excluded essential research and development costs.
Amodei used his essay to advocate for stronger export controls on AI chips to China, describing them as a crucial measure to maintain a technological lead over Chinese firms. He stressed that democratic nations must develop superior AI models to remain ahead but should avoid giving China unnecessary technological advantages.
