AMD is developing the Helion Agentic AI Rack, a rack-scale system designed for artificial intelligence agents. According to TechRadar, its top configuration supports up to 32,768 Zen 6 x86 CPU cores, 16 Instinct MI350P accelerators and 74 petaflops of FP4 computing performance. The rack also supports up to 102 Tbps of networking bandwidth.
AMD describes configurations for personal assistants, enterprise applications and technical computing. The source says personal-agent workloads use 256-core Venice processors rated at 400W, while enterprise workloads use 256-core Venice processors rated at 600W. Technical computing uses 96-core, high-frequency Venice processors. Operators can configure core density, performance and power use for their workloads.
The enterprise setup separates agent execution, inference and general-purpose computing across server nodes. CPU sandbox nodes handle agent execution, Instinct MI350P accelerators support inference, and general-purpose servers handle databases, storage and technical computing. The system also includes a token router for directing model requests across frontier and open-source models, programmable networking and ORv3 44OU hardware compliant with Open Compute Project specifications.
TechRadar reports that existing data centers may need changes to power delivery, cooling and networking to support the rack, depending on the configuration. AMD has not disclosed full facility upgrade costs or cooling requirements. The rack is scheduled to become available in November 2026.
Comments
0No comments yet. Be the first to comment.