On October 28, Beijing time, Business Insider reported that Anthropic's prospectus filed ahead of its public listing revealed that it integrates computing power from multiple strategic partners into a unified resource pool, allocating tasks through software based on price and performance. Analysts noted that this multi-chip strategy could weaken NVIDIA's pricing power in the long term. Anthropic uses four types of chips—Amazon Trainium, Google TPU, NVIDIA GPU, and AMD GPU—to train Claude and provide services, with its software scheduling layer optimizing hardware selection based on task requirements. Analysts believe this flexibility could enable large AI companies to reduce their reliance on NVIDIA, enhance their bargaining power, and become less sensitive to fluctuations in NVIDIA's chip supply. However, implementing a multi-chip strategy is challenging, especially when training AI models, and only large AI companies possess the necessary resources. Therefore, the strategy poses limited threats to NVIDIA in the short term.