HPE will be one of the first system makers to adopt AMD’s “Helios” architecture, which integrates AMD CPUs, GPUs and networking gear with a purpose-built HPE Juniper scale-up switch built on Broadcom. The whole thing is pitched as a high-bandwidth Ethernet monster for AI clusters that want to behave themselves.

“Helios” bundles AMD Epyc CPUs, Instinct GPUs, Pensando networking kit, and the ROCm software stack into a single, tidy platform designed to boost performance, efficiency, and scalability. The idea is to cut out the deployment faff for anyone trying to assemble gigantic AI clusters in research labs, the cloud, or cramped enterprise basements.

AMD chair and chief executive Dr Lisa Su said: “HPE has been an exceptional long-term partner to AMD, working with us to redefine what is possible in high-performance computing.”

She added that “With ‘Helios’, we’re taking that collaboration further, bringing together the full stack of AMD compute technologies and HPE’s system innovation to deliver an open, rack-scale AI platform that drives new levels of efficiency, scalability, and breakthrough performance for our customers in the AI era.”

HPE president and chief executive Antonio Neri said: “For more than a decade, HPE and AMD have pushed the boundaries of supercomputing, delivering multiple exascale-class systems and championing open standards that accelerate innovation,” before promising faster deployments and less risk for cloud providers.

AMD claims “Helios” will pump out up to 2.9 exaflops of FP4 performance per rack using Instinct MI455X GPUs, next-gen Epyc “Venice” CPUs and Pensando Vulcano NICs tied together with ROCm. It sits on the OCP Open Rack Wide design, which lets partners stitch together new systems without losing months to wrangling metal and cables.

HPE’s flavour of “Helios” includes an Ethernet switch and software tuned for the platform. The switch uses the Ultra Accelerator Link over Ethernet standard, underscoring AMD’s fondness for open networking rather than pricey black-box gear. HPE expects to flog “Helios” worldwide in 2026.

HLRS’s new “Herder” supercomputer runs on AMD Instinct MI430X GPUs and Epyc “Venice” CPUs inside HPE’s Cray GX5000 chassis. The system is billed as a heavy hitter for hybrid HPC and AI workflows, replacing the current “Hunter” machine in late 2027.

HLRS director Prof Dr Michael Resch said: “The pairing of AMD Instinct MI430X GPUs and EPYC processors within HPE's GX5000 platform is a perfect solution for us at HLRS,” adding that the setup will support both simulation and the rising tide of machine learning. He said the platform will give users room to develop new hybrid workflows that are only possible with next-generation hardware, their surname said.

Herder is scheduled for delivery in the second half of 2027 and should start earning its keep by year’s end.