AMD unveiled Helion, an open rack-scale blueprint for agentic AI infrastructure, at the Open Compute Project Summit on October 12, 2026 in San Jose. The system comes in two versions: a CPU-only setup supporting up to 128 EPYC 9006 Series processors (32,768 Zen 6 cores) for hyperscale operators, and an enterprise version combining EPYC CPUs with up to 16 Instinct MI350P GPUs. Helion complies with OpenRack v3 standards and supports up to 102 Tbps of networking bandwidth. AMD plans to share specifications with the OCP community and work with hardware partners including Cisco, Dell Technologies, HPE, Lenovo, and Supermicro for specific configurations, with availability targeted for November 2026.
Source: Read the original article

