AMD unveils Helion agentic AI server rack with token router, plans to open-source design to Open Compute Project

AMD unveiled Helion, a rack-scale server built for AI agents, with an open design it plans to give to the Open Compute Project. It includes a token router that sends each request to the most cost-effective AI model, and comes in a CPU-only version with up to 32,768 EPYC cores for hyperscalers and an all-in-one version with EPYC CPUs and Instinct MI350P GPUs for businesses, backed by Dell, HPE, Cisco and Supermicro.

Detected & updated continuously · Source: Nebula