Anthropic's Claude Mythos Preview model has significantly reduced the time required to onboard AMD GPUs, cutting the process from up to 12 months to approximately three days with minimal human intervention. This advancement, which involves the AI autonomously writing nearly all necessary kernels, challenges Nvidia's long-standing CUDA dominance. Anthropic's co-founder, Tom Brown, highlighted that this task previously required a team of dozens of engineers, and the company's multi-chip strategy allows it to run models on various hardware including Nvidia, AMD, Google TPUs, and AWS Trainium. AI
IMPACT Accelerates hardware onboarding and potentially erodes Nvidia's CUDA moat by enabling AI to write low-level code.
RANK_REASON AI model autonomously writes GPU kernels, a significant research milestone in AI software development for hardware. [lever_c_demoted from research: ic=1 ai=1.0]
- AMD
- Anthropic
- Anush Elangovan
- AWS
- Claude Mythos Preview
- CUDA
- EE Times
- Lisa Su
- Meta
- MI355x
- Nvidia
- PyTorch
- ROCm
- Tom Brown
- Trainium
- Vamsi Boppana
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →