AMD Unveils Open Helion Rack Design for Agentic AI Systems
AMD’s Helion blueprint targets flexible AI racks for hyperscalers and enterprises, with EPYC CPUs, Instinct MI350P GPUs and air or liquid cooling.
- published
- in the api
- on the website
on the website stories appear 15 minutes after publication, in the api at once. connect the api →
what happened
According to GuruFocus, AMD is moving further into full-stack AI infrastructure with Helion, an open rack blueprint for hyperscalers and enterprise customers. Built around the Open Compute Project framework, the design lets users combine CPUs and GPUs in different ways to suit agentic AI workloads, while providing more flexibility over computing capacity, power use and costs.
A CPU-only version uses sixth-generation AMD EPYC processors and can accommodate up to 32,768 cores. AMD is also working on an integrated rack that pairs EPYC chips with Instinct MI350P GPUs, targeting businesses that need agentic AI and conventional workloads on the same infrastructure. Both air-cooled and liquid-cooled designs are supported.
AMD plans to submit the Helion specification to the Open Compute Project community. Its partner ecosystem includes Cisco, Dell Technologies, HPE, Lenovo, Supermicro, Hon Hai, QCT and Wiwynn. The configurations remain under development and are expected to be available in November 2026.
why it matters
Helion would place AMD closer to the rack-level decisions that connect processors, accelerators, networking, power and cooling. Market attention is likely to focus on adoption, MI350P deployments, EPYC demand and partners’ progress in turning the blueprint into commercial systems.
instruments
sources
- AMD Raises Stakes in AI Race With Major New Move — gurufocus.com, 12 October 2026 at 18:24 UTC
prepared by the newsmetry newsroom with ai assistance from the sources listed below. not investment advice.
on the website stories appear 15 minutes after publication, in the api at once.
get stories in real time ↗