Industry change

AMD introduces Helion rack for agents

Published
Source
AMD press release

Summary

AMD announced Helion, an open rack blueprint for running AI agents. Based on an Open Compute Project spec, it proposes a CPU-only rack with up to 32,768 cores and an all-in-one CPU/GPU configuration, and includes an integrated token router that dispatches agents' model calls across models. AMD plans to contribute the spec to OCP, with partner systems expected in November.

Why it matters for our work

This signals that infrastructure dedicated to agent execution, not training GPUs, is becoming its own market. Power efficiency and open designs could shape the cost of operating agents at scale, which in turn affects enterprise adoption speed. Configurations and timing are AMD's stated plans.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.

Read original (opens in a new tab)