AMD introduces Helion, an open rack design for AI agents with up to 32,768 EPYC cores, due from partners in November

AMD on 12 October introduced Helion, which it calls an open blueprint for agentic AI infrastructure. It is built on an Open Compute Project specification, and AMD says it will contribute the Helion specification to the OCP.

The design comes in two versions. A CPU-only rack for hyperscale customers uses 6th Gen EPYC processors and “can support up to 32,768 CPU cores,” AMD says. An all-in-one rack combines EPYC CPU nodes with Instinct MI350P GPU nodes, for enterprises that want agent execution and inference in one rack. Both include a built-in router that sends agents' model calls to frontier and open-source models. AMD says there will be liquid-cooled and air-cooled versions.

“Agentic AI is not a single workload,” said Dan McNamara, AMD's senior vice president and general manager for compute and enterprise AI. AMD's pitch is that agents need CPU capacity to run tools, code and sandboxes alongside the models they call.

Cisco, Dell, Hon Hai Technology Group, HPE, Lenovo, QCT, Supermicro and Wiwynn gave supporting statements. AMD says partner configurations are still in development and are planned for November 2026.

• Confirmed The CPU-only Helion rack uses 6th Gen EPYC processors and “can support up to 32,768 CPU cores”. AMD Newsroom
• Confirmed Partner configurations are planned for November 2026, and AMD plans to contribute the specification to the Open Compute Project. AMD investor relations release
• Claimed A built-in router sends agents' model calls across frontier and open-source models. AMD Newsroom