AMD Announces Helion Rack Solution for Agentic AI Workloads
AMD announced the Helion rack solution designed for agentic AI workloads, featuring flexible CPU and GPU configurations with a pure CPU version that scales up to 32,768 cores. The system is scheduled for an official launch in November 2026. This architecture addresses the heavy computing and multi-turn execution demands of agentic AI by providing massive scalability and open industry standards. By contributing the specifications to the Open Compute Project (OCP), AMD aims to drive ecosystem-wide collaboration and efficient data center deployments. The Helion rack relies on OCP design specifications and leverages sixth-generation EPYC processors for its pure CPU configuration. AMD is currently developing the platform with ecosystem partners ahead of its late 2026 release.
## BACKGROUND
Agentic AI shifts large language model serving from isolated prompt requests to stateful, multi-turn executions that repeatedly invoke models and call tools. Meanwhile, the Open Compute Project (OCP) is a collaborative organization that shares efficient data center hardware designs and industry best practices among major technology companies.