Anthropic claims its frontier model Mythos ported the Claude AI system to AMD GPUs in three days with nearly 100% kernel automation, collapsing a process that typically requires six to twelve months of engineering effort. Co-founder and Chief Compute Officer Tom Brown states the company now trains and runs Claude across all four major chip families, treating diverse compute as a single interchangeable pool. This multi-chip strategy, pursued for years despite industry skepticism, enables optimization for price-performance across training and inference workloads, challenging Nvidia's CUDA ecosystem lock-in and its reported 70%+ hardware margins. Anthropic attributes its efficiency to compounding "compute multipliers" — recursive research breakthroughs that amplify hardware utilization — and the "Anthropic Direct" program, which leases and operates chips in purpose-built data centers to lower total cost of ownership. The company asserts it extracts more value per compute dollar than any frontier lab. Financial ties reflect this shift: Nvidia committed up to $10 billion via a 2025 Microsoft deal and participates in a $26 billion conditional private placement per Anthropic's November 2026 S-1/A filing, while AMD pledged up to $5 billion in July 2026 for deployment of up to two gigawatts of Instinct MI455X GPUs in Helios rack-scale systems starting H1 2027. Figures derive from Anthropic's investor presentation and remain unverified.