The processor complements the Helion agentic AI rack, an open-specification design arriving in November 2026. The Helion platform allows for configurations without GPUs to handle CPU-heavy agent tasks like code execution and tool calls, supporting up to 32,768 cores per rack via 128 EPYC "Venice" chips. Partners including Dell, HPE, Supermicro, and Cisco are supporting the rack, which includes a token router to direct requests to the most cost-effective AI models.
