Anthropic says AI is eroding Nvidia's CUDA lead, cutting an AMD chip switch from months to three days
Anthropic says its Mythos model helped port Claude to AMD GPUs in three days, with almost no human intervention. The company says it is pursuing a multi-chip strategy.
Key points
- Anthropic says Mythos ported or wrote nearly 100% of the kernels for the AMD conversion
- The company says bringing up a new chip platform previously kept dozens of engineers busy for months
- Anthropic says it trains and runs Claude on four major chip families
Sources
Details
Anthropic says it now trains and runs Claude on four major chip families. Co-founder and Chief Compute Officer Tom Brown said the company invested in a multi-chip strategy despite the cost and engineering work involved in supporting separate software and inference stacks.
The company says it first used AMD GPUs this year and used its Mythos model to port Claude to them in three days, with almost no human intervention. The Decoder reports that the figures are from an Anthropic investor presentation and have not been independently verified.
Anthropic says it treats compute from different vendors, chip architectures and locations as one pool, and claims it gets more out of each dollar of compute than any other frontier AI lab. It is also building data center capacity through its Anthropic Direct program, leasing chips and operating them in purpose-built facilities.
Anthropic has financial ties to both Nvidia and AMD. The Decoder reports that AMD announced an investment of up to $5 billion in July 2026, alongside Anthropic's plan to deploy up to two gigawatts of AMD GPUs, with the first gigawatt phase due to begin in the first half of 2027.
Written from coverage as of
Images
- The Decoder
- The Decoder
- The Decoder
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh