AMD launches Helion agentic AI rack solution, supporting up to 32,768 CPU cores per rack.

Beating AI News Flash: AMD today announced the launch of the AMD Helion agentic AI rack solution, featuring an open infrastructure design based on Open Compute Project (OCP) specifications, aimed at meeting the compute demands of large-scale AI agent task execution and supporting flexible combinations of CPUs and GPUs.

AMD Helion offers two main configurations: first, a CPU-only rack powered by 6th Gen AMD EPYC processors, supporting up to 32,768 CPU cores, designed for hyperscale deployments, providing dedicated compute for agent task execution and connecting to separate model inference infrastructure; second, an integrated rack combining AMD EPYC CPUs, AMD Instinct MI350P GPUs, and general-purpose compute nodes, designed for enterprise-grade agentic AI applications.

AMD stated that Helion supports liquid-cooled and air-cooled configurations, allowing customers to adjust CPU and GPU resources based on workload, performance, power, and cost requirements. The solution also combines open hardware and software with an integrated Token Router, which can distribute agent model calls to frontier models and open-source models.

AMD plans to submit the Helion specification to OCP for further community development and adoption. Currently, the solution is being co-developed with ecosystem partners and is expected to launch in November 2026.