AMD and HPE Unveil Rack-Scale AI Platform Built on New Helios Architecture

AMD (AMD) has expanded its strategic collaboration with Hewlett Packard Enterprise (HPE) to deliver open, rack-scale AI infrastructure aimed at accelerating high-performance computing (HPC) and advancing Sovereign AI research. The partnership marks one of AMD’s most significant moves in large-scale AI systems.

Helios: AMD’s New Rack-Scale AI Architecture

HPE Among First to Adopt Helios

HPE will be one of the earliest system providers to integrate AMD’s Helios architecture, designed to simplify deployment of massive AI clusters. The companies will incorporate:

HPE Juniper Networking scale-up switch, co-developed with Broadcom (AVGO)
Full-stack AMD compute technologies
Optimized rack-scale integration for hyperscale AI workloads

Leadership Commentary

AMD CEO Lisa Su emphasized Helios as a breakthrough platform enabling deeper collaboration:

“We’re bringing together the full stack of AMD compute technologies and HPE’s innovation to deliver efficiency, scalability, and breakthrough performance for customers in the AI era.”

HPE CEO Antonio Neri added that the platform provides:

Faster deployments
Greater flexibility
Reduced risk for cloud service providers scaling AI compute

Helios Hardware Stack: Built for Massive AI & HPC Workloads

The Helios platform combines AMD's most advanced components:

AMD Instinct MI455X GPUs
AMD EPYC Venice CPUs
AMD Pensando Vulcano NICs for high-bandwidth scale-out networking

This full-stack integration is designed to deliver superior performance across AI training, inference, and traditional HPC simulations.

Powering the “Herder” Supercomputer in Germany

Helios will also power Herder, a new supercomputer at the High-Performance Computing Center Stuttgart (HLRS) — one of Europe’s leading HPC research hubs.

HLRS Director Michael Resch highlighted Herder’s hybrid capabilities:

Supports classical HPC numerical simulations
Enables emerging AI and machine-learning workloads
Integrates both approaches for new hybrid HPC/AI workflows

Herder is expected to go online by the end of 2027.

Strategic Significance

The expanded AMD–HPE partnership:

Strengthens AMD’s position in the high-end AI systems race
Provides governments and enterprises with Sovereign AI–ready infrastructure
Increases competition against NVIDIA-based rack-scale platforms
Deepens AMD’s ecosystem with Broadcom and Juniper networking technologies