AMD Leak Reveals Anthropic is Diversifying Chips

It is honestly kind of funny that a high-level executive's GitHub commit leaked this, but the signal is loud and clear: the "de-Nvidia-fication" of AI infrastructure is accelerating. For a long time, NVIDIA's H100s were the only game in town for top-tier LLM training, but the risks of vendor lock-in—both in terms of pricing power and supply chain fragility—have become too high for companies like Anthropic to ignore.

By integrating AMD's MI-series GPUs, Anthropic isn't just hedging its bets; it's gaining significant leverage. We've already seen them lean on Google's TPUs, and adding AMD to the mix creates a diversified compute matrix that makes them far more resilient. For AMD, this is a massive win. Landing a client of Anthropic's caliber proves their hardware can handle the most demanding frontier models, effectively breaking the aura of NVIDIA's exclusivity in the high-end training market. This shift suggests we are entering an era where the "compute war" is as much about supply chain diplomacy as it is about raw FLOPS.