Anthropic says AI is eroding Nvidia's CUDA lead, cutting an AMD chip switch from months to three days
The Decoder
Anthropic claims it used AI to bypass Nvidia's software lead by porting Claude to AMD GPUs in just three days.
Anthropic announced that it has successfully reduced its reliance on Nvidia through a multi-chip strategy, utilizing its own AI model, Mythos, to port Claude to AMD GPUs in only three days instead of the usual several months.
Nvidia has long maintained a competitive advantage through its CUDA software ecosystem, which locks developers into its hardware. However, according to Anthropic co-founder Tom Brown, the company now trains and operates Claude across four major chip families as a single interchangeable compute pool, focusing on the best price-performance ratio.
Anthropic's investor presentation also highlighted the use of compute multipliers to enhance performance and detailed plans to build its own data center capacity via the Anthropic Direct program. Both Nvidia and AMD maintain significant financial ties to Anthropic through multi-billion dollar investments and upcoming hardware deployments.
These claims have not been independently verified, but they suggest that AI models themselves could eventually dismantle hardware software monopolies and lower switching barriers in the compute market.
Why it matters: This development indicates that AI models could soon dismantle hardware monopolies like Nvidia's CUDA by automating the complex process of porting software across different chip architectures.



