AMD is introducing an open rack design for agentic AI with up to 32,768 CPU cores

AMD $AMD introduced the AMD Helion agentic AI rack on Monday, an open blueprint for agentic AI infrastructure built on an Open Compute Project spec-design and powered by AMD compute.

The platform comes in two primary configurations. One is a CPU-only rack built on sixth-generation AMD EPYC silicon, capable of scaling to 32,768 CPU cores; it is designed for hyperscale customers seeking dedicated agent execution capacity separate from their model inference infrastructure. An all-in-one rack combines AMD EPYC CPU nodes with AMD Instinct MI350P GPU nodes and general-purpose compute on a single platform, targeting enterprise agentic AI deployments, the company said.

Both configurations are available in liquid- and air-cooled versions. AMD Helion also includes an integrated token router that directs AI agents' model calls across frontier and open-source models.

"Agentic AI is not a single workload. It spans personal agents, enterprise applications, code execution and complex workflows, each with different compute requirements," said Dan McNamara, senior vice president and general manager of Compute and Enterprise AI at AMD, in a statement. "With AMD Helion, we are bringing AMD EPYC performance, open software and flexible rack designs together so our partners can build systems around customer needs and make efficient use of available power."

AMD said it intends to submit the Helion specification to the Open Compute Project so the broader industry can build on it. Configurations from launch partners are currently in development and targeted for release in November 2026.

Eight companies have announced support for the platform: Cisco $CSCO Systems, Dell $DELL Technologies, Hon Hai Technology Group, Hewlett Packard Enterprise, Lenovo, QCT, Supermicro and Wiwynn.

Dell Technologies Senior Vice President Arun Narayanan said in a statement that AMD Helion, combined with Dell's PowerRack infrastructure, gives customers "a complete, validated path to deploy agentic compute at scale." HPE Senior Vice President Trish Damkroger said in a statement that the two companies are "ushering in the next era of AI with open large-scale systems," noting both air-cooled and direct liquid-cooled networking options.

AMD said the platform builds on the design approach of its existing AMD Helios rackscale solution and is intended to let partners and customers match CPU and GPU configurations to specific workload requirements, including performance, power and cost targets.