AMD defines rack-scale infrastructure for agentic AI
AMD Helion is a rack-scale reference design for agentic AI infrastructure, combining open-source hardware and software. The design accommodates CPU-only configurations for hyperscale deployments and CPU/GPU configurations for enterprises. Based on Open Compute Project (OCP) specifications, the Helion 21-in. rack architecture lets customers balance CPU and GPU performance, core density, and power consumption to meet their application requirements.
A CPU-only rack can be powered by AMD EPYC processors supporting up to 32,768 Zen 6 x86 CPU cores. The all-in-one rack integrates AMD EPYC CPU nodes, up to 16 AMD Instinct MI350 GPU nodes, and general-purpose compute. A token router directs inference requests to frontier models in the cloud or open-source models running locally in the rack.
AMD plans to contribute the Helion specification to OCP for community development and adoption. Rack configurations are under development with ecosystem partners, with availability planned for November 2026.
The post AMD defines rack-scale infrastructure for agentic AI appeared first on EDN.
