AMD Unveils Helion Rack Platform for Agentic AI; Su Says CPU Market to Exceed $200B by 2030
On October 12, AMD announced at the OCP Global Summit in Santa Clara, California, the launch of Helion, a dedicated rack-scale computing platform for agentic AI, available in CPU-only and CPU+GPU configurations. The platform is designed around Open Compute Project (OCP) specifications and incorporates AMD's custom compute silicon.
The CPU-only rack configuration is powered by sixth-generation EPYC processors and supports up to 32,000 CPU cores, targeting hyperscale deployments and providing dedicated compute for agent execution. The second configuration combines EPYC CPU nodes, Instinct MI350P GPU nodes, and general-purpose compute for enterprise agentic AI applications. Helion also features an integrated token router that automatically directs model invocation requests from AI agents to frontier or open-source models.
An AMD senior vice president said agentic AI spans personal assistants, enterprise applications, code execution, and complex workflows, with different tasks placing significantly different demands on compute resources. Helion is designed to address these varied requirements through a flexible architecture. AMD plans to contribute the Helion specification to the OCP community. Rack configurations are currently being developed with ecosystem partners and are expected to launch in November 2026.
AMD CEO Lisa Su, in her keynote at the OCP Global Summit, characterized 2026 as the "year of inference." She said CPU market growth would accelerate sharply, driven by agentic AI workloads. Su disclosed that AMD has raised its CPU market forecast three times over the past nine months, and now expects the market to grow at a compound annual rate of roughly 50% through 2030, reaching a total size exceeding $200 billion. By comparison, the overall CPU market was about $20 billion a year earlier, when it was widely expected to grow at only single-digit annual rates.
