Anthropic says Claude completed GPU kernel adaptation of a model to AMD accelerators in 3 days
Anthropic says Claude handled almost all of the GPU kernel code porting and writing involved in adapting the model to AMD AI accelerators, with very little human intervention, and the entire process took only 3 days. According to the company's past experience, onboarding a new chip platform using traditional engineering methods typically requires dozens of engineers and takes 6 to 12 months. The chips actually adapted so far include NVIDIA, AMD, Google TPU, and Amazon Trainium. AMD and Anthropic previously announced they will deploy up to 2GW of GPU compute based on AMD's MI450 series, with the first 1GW planned to begin in the first half of 2027; the two sides will also use Claude to optimize AMD GPU workloads and accelerate ROCm software development.
Why read this
AI score 70
The original text provides a time comparison between traditional adaptation and the AI-assisted process, allowing readers to gauge the room for change in the cost of migrating to a chip platform.
Topics
Body · AI translation
Claude ported or wrote almost all of the GPU kernel code in this work, with very little human intervention during the process. The case has sparked outside discussion about Nvidia's CUDA software ecosystem and competition in AI chips; CUDA is Nvidia's moat, but how long this moat can last in the AI era remains to be seen.
English text is being prepared.
Nvidia's CUDA brings together a large number of development tools, compute libraries, and GPU-optimized software, and has long been an important foundation for AI development, because adapting code originally optimized for specific hardware to other chips requires substantial engineering resources.
With AI involved in low-level code porting, it can help engineers write and adjust GPU kernels faster, and the cost and time for enterprises to onboard different chip platforms have the opportunity to decrease. At this stage, the chips Anthropic has actually adapted include but are not limited to Nvidia, AMD, Google TPU, and Amazon Trainium.
AMD and Anthropic previously announced that they will deploy up to 2GW of GPU compute based on AMD's MI450 series, with the first 1GW deployment planned to begin in the first half of 2027; the two sides will also use Claude to optimize AMD GPU workloads and accelerate ROCm software development.