AMD announced an expanded collaboration with Hewlett Packard Enterprise to advance open, rack-scale AI infrastructure using AMD’s new “Helios” architecture. HPE will be one of the first OEMs to adopt Helios, an open, full-stack AI platform that integrates AMD EPYC CPUs, AMD Instinct GPUs, Pensando networking, and the ROCm software ecosystem. The system is designed to simplify deployment of large AI clusters with high-bandwidth, low-latency Ethernet connectivity.

HPE will incorporate purpose-built HPE Juniper Networking switches, developed with Broadcom, to support Ultra Accelerator Link over Ethernet, providing optimized networking for large-scale AI workloads. HPE plans to offer the Helios rack-scale platform globally beginning in 2026.

The companies also announced that HLRS in Germany will deploy “Herder,” a new HPE Cray GX5000 supercomputer powered by AMD Instinct MI430X GPUs and next-generation AMD EPYC “Venice” processors. Scheduled for delivery in late 2027, Herder will support HPC, AI, and hybrid workflows for European scientific and industrial research.

AMD and HPE emphasized that their long-standing partnership continues to drive advances in high-performance computing, open standards, and scalable AI infrastructure.