AMD opens its agentic rack blueprint: CPU-only configs reach 32,768 cores, spec headed to OCP

AMD introduced the Helion agentic AI rack, an open blueprint for rack-scale agentic infrastructure in which customers match CPU performance, core density, power and GPU acceleration to the work their agents perform. The prior rack-scale offering was AMD Helios; Helion continues that open-platform approach and extends the adjustable range from the rack to combinations of CPU and GPU nodes. AMD gives two configuration examples: a CPU-only rack powered by 6th Gen AMD EPYC processors supporting up to 32,768 CPU cores for hyperscale customers, and an all-in-one rack combining EPYC CPU nodes, Instinct MI350P GPU nodes and general-purpose compute for enterprise agentic AI.

The rack blueprint is based on an Open Compute Project spec-design, supports open software, and comes in liquid- and air-cooled versions so customers can build the configuration they need. AMD plans to contribute the Helion specification to OCP for community development and adoption, rather than keeping it only as a proprietary product line. AMD says Helion agentic rack configurations are under development with ecosystem partners and are planned to be available in November 2026, with Cisco, Dell Technologies, Hon Hai Technology Group, HPE, Lenovo, QCT, Supermicro and Wiwynn listed as ecosystem supporters.

Helion puts an integrated token router into the rack blueprint to direct agents' model calls across frontier and open-source models. AMD describes agentic AI as not a single workload but one spanning personal agents, enterprise applications, code execution and complex workflows, each with different compute requirements. Dan McNamara, AMD senior vice president and general manager of Compute and Enterprise AI, said this takes optimized compute and infrastructure that adapts to different workloads, and that Helion brings EPYC performance, open software and flexible rack designs together so partners can build around customer needs and make efficient use of available power.

The blueprint targets partners and customers deploying agentic AI, covering hyperscale and enterprise settings, with configurations from a CPU-only rack to an all-in-one CPU and GPU rack, in liquid- and air-cooled versions. AMD plans to contribute the Helion specification to OCP, configurations are under development with ecosystem partners, and availability is planned for November 2026.

AMD describes Helion as an open blueprint, and specific configurations, performance and power behavior depend on partner implementations and customer workloads; development and adoption after the specification is contributed to OCP, along with the November 2026 availability plan, are forward-looking statements.