AMD Announces Helion Open Rack Blueprint for Agentic AI Infrastructure
DataNews Editorial Team
AMD has announced Helion, an open rack-infrastructure blueprint for agentic AI based on Open Compute Project specifications and AMD compute platforms. Partner-developed configurations are still in development and are expected to reach the market in November 2026.
- AMD has announced Helion as an open, OCP-oriented rack blueprint for agentic AI.
- The design includes air- and liquid-cooled options, plus CPU-only and combined CPU/GPU configurations.
- The CPU-only configuration supports up to 32,768 CPU cores using sixth-generation AMD EPYC processors.
- AMD intends to contribute the Helion specification to OCP.
- Partner configurations remain under development and are expected on the market in November 2026.
AMD has introduced Helion, an open blueprint for agentic AI rack infrastructure built around Open Compute Project (OCP) specifications and AMD compute platforms.
The design supports both air-cooled and liquid-cooled deployments, with a CPU-only option alongside a combined CPU/GPU rack configuration. The CPU-only design uses sixth-generation AMD EPYC processors and supports up to 32,768 CPU cores, an IT-scope specification. The combined configuration includes AMD EPYC CPU nodes, AMD Instinct MI350P GPU nodes and general-purpose compute resources.
AMD plans to contribute the Helion specification to OCP. The company is developing Helion configurations with ecosystem partners, with market availability expected in November 2026. Supermicro is identified as participating in the go-to-market effort, although the available information does not specify its commitments, delivery volumes or supply timetable.
For operators and system integrators, the announcement expands the potential set of standardised rack architectures for AI deployments with different CPU/GPU mixes, cooling approaches, density profiles and available power conditions. If partner systems arrive on the indicated timetable, Helion could offer OEMs and customers a modular alternative for tailoring infrastructure to particular agentic AI workloads.
The announcement does not provide pricing, rack power draw, energy consumption, shipment volumes or details of deployment sites. Helion should therefore be viewed as an announced infrastructure specification and partner-development programme rather than as commercially available rack capacity.
Helion could broaden the range of standardised rack architectures available for AI workloads with differing CPU, GPU, density and power requirements. For data-center operators and integrators, it may provide greater flexibility in configuring agentic AI infrastructure, including a choice between air and liquid cooling.
- Data Centre Magazine / October 16, 2026 · EnglishSource →
This article was prepared by the DataNews editorial team based on the sources listed above.