Key information
AMD has officially introduced the Helion agentic AI rack, an open blueprint designed to support agentic AI infrastructure. Built upon an Open Compute Project specification, the system utilizes AMD compute technology to provide a flexible foundation for diverse AI workloads.
The architecture allows users to balance CPU performance, core density, and power with GPU acceleration. According to Dan McNamara, senior vice president and general manager of Compute and Enterprise AI at AMD, the platform is designed to adapt to various requirements, ranging from personal agents to complex enterprise workflows.
The rack supports both liquid- and air-cooled configurations. Specific options include a CPU-only setup featuring 6th Gen AMD EPYC processors capable of supporting up to 32,768 cores, or an all-in-one platform integrating EPYC CPU nodes with AMD Instinct MI350P GPU nodes.
AMD intends to contribute the Helion specification to the Open Compute Project. While configurations are currently being developed alongside ecosystem partners, the company expects the racks to become available in November 2026.
