AMD Unveils Helion Open Rack for Agentic AI
AMD's Helion rack lets customers mix EPYC CPUs and Instinct MI350P GPUs by workload, with systems planned for November.
AMD announced Helion, an open server rack design aimed at agentic AI infrastructure, according to inews24.com. The design lets customers combine CPU and GPU configurations based on workload, power and cost requirements.
AMD said it is developing Helion systems with server manufacturers and plans to launch them in November. The company will contribute the Helion design specification to the Open Compute Project.
Helion extends the open design approach of AMD's existing Helios server rack and supports both liquid and air cooling. A CPU-only Helion rack uses 6th Gen AMD EPYC processors and supports up to 32,768 CPU cores.
A combined rack pairs AMD EPYC CPUs with AMD Instinct MI350P GPUs for both AI agent execution and model inference. Dell Technologies, HPE, Lenovo, Supermicro and Hygon said they plan to support Helion.
Quick answers
What is AMD Helion?
Helion is an open server rack design from AMD for agentic AI infrastructure that lets customers combine CPU and GPU configurations according to workload, power and cost needs.
When will AMD Helion systems launch?
AMD plans to launch Helion systems in November, developed with server manufacturers.
Which companies plan to support Helion?
Dell Technologies, HPE, Lenovo, Supermicro and Hygon said they plan to support Helion.