AMD to Launch Helion Agentic AI Rack in November 2026
The new hardware rack supports up to 32,768 Zen 6 cores to meet the scaling demands of emerging agentic AI workloads.
Updated on Oct. 16, 2026 in Data Centers
Live Poll
Is now a good time for local data centers to invest in upgrading their power infrastructure?
AMD is preparing to launch the Helion Agentic AI Rack, a modular hardware system designed to support agentic AI workloads, in November 2026. This announced rack architecture scales up to 32,768 Zen 6 x86 CPU cores and 16 Instinct MI350P accelerators per unit.
Why it matters
The system represents a move toward infrastructure specialized for agentic AI, which requires distinct handling of execution, inference, and general-purpose compute. Its density, supporting up to 102 Tbps of bandwidth, highlights the rising demand for higher power and cooling capacity in next-generation data centers.
The rack provides 74 petaflops of FP4 computing performance and 102 Tbps of networking bandwidth, utilizing ORv3 44OU hardware. Processors range from 96-core high-frequency models for technical computing to 256-core variants for personal and enterprise agents.
The players
AMD
A semiconductor company focused on x86 processors, high-performance GPUs, and adaptive computing solutions for enterprise and data center markets.
Open Compute Project
A collaborative community that develops and publishes open-source standards for efficient data center hardware, including the ORv3 power architecture.
The details
The Helion Agentic AI Rack uses modular server nodes, allowing operators to prioritize either high core density for personal and enterprise agents or high-frequency processing for technical computing. The enterprise configuration physically isolates agent execution, inference, and general-purpose computing tasks into distinct server nodes to manage resources. The architecture adheres to the Open Compute Project's ORv3 44OU hardware standard, which optimizes power delivery and thermal management for high-density deployments.
Timeline
The Helion Agentic AI Rack is scheduled for availability in November 2026.
The Tech Race
AMD is positioning the Helion rack to compete in the specialized agentic AI market, where throughput requirements are pushing beyond standard cluster designs. By adopting the Open Compute Project's ORv3 hardware standard, the firm is racing to establish a compatible template for future hyperscale deployments.
Data center operators will need to assess if their existing power and cooling facilities can support the 400W to 600W-per-processor requirements of the new racks. Once available in November 2026, the system will offer enterprise IT managers a new method to partition inference and execution tasks.
The takeaway
The development of hardware specifically for agentic AI workloads marks a shift toward dedicated infrastructure for autonomous computing tasks. Interested parties should monitor subsequent announcements from AMD regarding the rack-level power consumption benchmarks as the November 2026 launch approaches.
Further reading
For broader trends in hardware requirements, explore the Data Centers section.
Source note: This article includes information reported by TechRadar.
Live Poll
Is now a good time for local data centers to invest in upgrading their power infrastructure?
