Summarize with AI
AMD introduced Helion, an open rack design for agentic AI infrastructure, on October 12, 2026. It is based on an Open Compute Project (OCP) design, and AMD plans to give the specification to OCP. Racks are planned for November 2026.
The key part is an integrated token router. It directs AI agents' model calls across frontier models and open-source models.
Two configurations
Both come in liquid-cooled and air-cooled versions.
| CPU-only rack | All-in-one rack | |
|---|---|---|
| For | hyperscale cloud companies | enterprises |
| Hardware | 6th Gen AMD EPYC processors | EPYC CPU nodes, Instinct MI350P GPU nodes, general-purpose compute |
| Size | up to 32,768 CPU cores | not stated |
| Job | runs the agents | runs the agents and the models (GPU inference) |
| Models run | on separate inference infrastructure | in the same rack |
Partners and availability
The configurations are being developed with partners and are planned for November 2026. AMD names Cisco, Dell Technologies, Hon Hai, HPE, Lenovo, QCT, Supermicro and Wiwynn as partners. It does not say which of them will sell Helion systems in November.
Sources
Did this land? Leave a reaction.
Discussion
0 commentsMembers can join the discussion
Become a member to comment and reply to other members.
Become a MemberNo comments yet
Be the first to start the discussion.