AMD introduces Helion rack blueprint for agentic AI

AMD introduced Helion at the Open Compute Project Global Summit 2026, according to TechRepublic.[2] AMD describes Helion as an open infrastructure blueprint based on an Open Compute Project design.[1] The company says the rack builds on the open-platform approach of its Helios rack-scale solution.[1] AMD plans to contribute the Helion specification to the Open Compute Project for further development and adoption.[1] That makes Helion a proposed shared design for partners to develop, rather than a single finished configuration described in the announcement.[1]

AMD’s CPU-only configuration uses 6th Gen EPYC processors and can support up to 32,768 CPU cores for hyperscale deployments.[1] AMD says this rack is designed to run agents and connect to separate infrastructure for model inference.[1] Its all-in-one enterprise configuration instead combines EPYC CPU nodes, Instinct MI350P GPU nodes and general-purpose compute on one platform.[1] AMD says Helion can let customers deploy agentic AI using IT infrastructure they already have.[1] For teams evaluating the design, the distinction is material: AMD describes one option around agent execution and another that combines CPU and GPU resources.[1]

AMD says Helion combines open hardware and software with a token router that directs agents’ model calls across frontier and open-source models.[1] The company says customers can match CPU performance, core density and power with GPU acceleration for the work their agents perform.[1] The rack blueprint supports both liquid-cooled and air-cooled configurations, according to AMD.[1] The Register reported that advertised configurations include up to 16 Instinct MI350P accelerators and up to 102 Tbps of networking bandwidth.[3] It also reported dense, balanced and high-performance CPU blades, with different processor specifications for balanced and performance options.[3]

AMD said Helion rack configurations were under development with ecosystem partners and planned to be available in November 2026.[1] Cisco Compute executive Jeremy Foster said combining Helion with Cisco’s infrastructure capabilities could give customers more flexibility to deploy and scale agentic workloads.[1] Dell Technologies executive Arun Narayanan said the company’s support for the reference design extends its commitment to agentic AI.[1] The summit also featured Intel Xeon-based rack designs, many using the ORv3 standard, TechRepublic reported.[2] The announcement therefore presents Helion as one design in a broader rack-infrastructure conversation, while its configurations remain in development.[1][2]