AMD unveils Helion rack reference design for AI agents at OCP Summit

Helion targets agentic AI workloads

AMD has unveiled a new rack-scale reference design called Helion, aimed at running AI agents. Revealed at the Open Compute Project (OCP) Summit 2026 in San Jose, California, the design offers customers a choice between a CPU-only hyperscale configuration and an enterprise configuration that mixes CPU and GPU blades within a single rack.

Helion follows the Helios system AMD launched earlier this year for frontier-scale AI training and large-scale inference. Where Helios targets those workloads, Helion's enterprise configuration combines agent execution and inference in one rack, while the CPU-only hyperscale version provides dedicated capacity for agent execution with connections to separate model inference infrastructure.

Configuration options and hardware

Depending on configuration, AMD advertises up to 32,768 of its "Zen 6" CPU cores, delivered via compute nodes fitted with 256-core versions of its Venice Epyc processor, up to 16 Instinct MI350P GPU accelerators, and up to 102 Tbps of networking bandwidth.

Both configurations offer dense, balanced, or high-performance CPU blades. Dense deployments use EPYC 9006 chips rated at 400 W, balanced blades use 256-core processors rated at 600 W, and performance blades sport a 96-core version of the chip clocked at 5 GHz. GPU inference blades pack eight Instinct MI350P GPUs into a 5U server, while general-purpose blades are intended for database and storage servers or technical compute workloads.

Open specifications and partner availability

AMD SVP for Compute and Enterprise AI Dan McNamara said agentic AI is not a single workload, spanning enterprise applications and complex workflows, each with different compute requirements. "To do that takes optimized compute and infrastructure that can adapt to different workloads. With AMD Helion, we are bringing Epyc performance, open software and flexible rack designs together so our partners can build systems around customer needs and make efficient use of available power," he added.

AMD says it plans to contribute the Helion specifications to OCP, making them available to the community for further development and adoption. The company cites Cisco, Dell, HPE, Foxconn, Lenovo, Supermicro and others as on board with the new kit. Partner-built Helion configurations are scheduled for November 2026.

Sources: Theregister