Anthropic's multi-platform computing strategy challenges nvidia's dominance

On October 26, Anthropic disclosed in its S-1 filing that it integrates computing power from multiple chip providers, including Amazon Trainium, Google TPU, Nvidia GPU, and AMD GPU, into a unified resource pool. This strategy, which took years and significant investment, allows the company to distribute AI workloads across different platforms for optimal efficiency. Analysts suggest this could reduce reliance on Nvidia and give large AI firms more leverage in pricing negotiations. Patrick Moorhead of Moor Insights & Strategy noted that Anthropic is leveraging this capability as a competitive advantage, potentially undermining Nvidia's pricing power. Ashish Nadkarni of IDC highlighted the benefit of diversification in mitigating supply fluctuations. However, experts like Gil Luria and Brad Gastworth argue that the strategy poses limited short-term threats to Nvidia due to the complexity of cross-platform task migration, particularly in model training. Ken Griffin of Castle Investment warned that excess supply could lead to a drop in GPU rental prices.