Summary
At Advancing AI 2026 on July 23, AMD moved its Helios rack-scale system (72 Instinct MI455X GPUs, 18 6th-gen EPYC 'Venice' CPUs) into production, claiming up to 30% more tokens per dollar than the leading competitor, and detailed MI400-series GPUs including the MI455X at 34x the token throughput of MI355X. Anthropic committed to deploying up to 2 GW of MI455X GPUs starting 2027.
What changed
AMD launched 6th Gen EPYC 'Venice' CPUs, the Instinct MI400 series (MI455X, MI430X, MI350P), Helios rack-scale systems, and Ryzen AI Embedded X100/Kria robotics platforms. Helios (72 MI455X GPUs + 18 EPYC Venice CPUs) is in production with a claimed up to 30% more tokens per dollar than the leading competitive solution; MI455X claims 34x higher token throughput than MI355X. Deployments are planned with OpenAI, Anthropic, Meta, Microsoft, and Oracle, and Anthropic committed to up to 2 GW of MI455X GPUs beginning 2027.
Why it matters
Inference cost per token is the economic bottleneck for running agents at scale, and AMD is attacking it head-on with a rack-scale system and named commitments from the top model labs. Anthropic's 2 GW commitment and OpenAI/Meta/Microsoft/Oracle deployments signal a credible second source to Nvidia, which could ease supply and pricing pressure across the AI compute market.
Evidence excerpt
72 high-performance AMD Instinct MI455X GPUs and 18 powerful 6th Gen AMD EPYC 'Venice' CPUs... up to 30% more tokens per dollar than the leading competitive solution.