Helios is a purpose-built rack-scale solution designed to power hyperscale and sovereign AI infrastructure, integrating AMD EPYC CPU, AMD Instinct GPU, and AMD Pensando networking into a cohesive system optimized for large-scale inference and training, scalable deployment, and operational efficiency.

The Supermicro AMD Helios Rack-Scale Solution is not just a rack of GPUs: it is a purpose-built system designed around repeatable 4-GPU compute trays, UALoE for rack-scale scale-up and standards-based Ethernet for cluster-scale scale-out, and hardware-rooted security to every layer of the AI rack.

6th Gen AMD EPYC CPUs deliver high core counts, exceptional memory bandwidth, and enterprise grade reliability to help sustain maximum performance across the entire Supermicro AMD Helios Rack.

AMD Instinct MI455X GPUs, built on the AMD CDNA™ architecture, offer massive HBM4 memory capacity, extreme bandwidth, and multi petaflop performance-enabling frontier class training, distributed inference, and high efficiency scaling across the entire Supermicro AMD Helios Rack.

AMD Pensando™ DPUs provide front-end network acceleration, storage offload, and inline security at line rate to keep every accelerator continuously fed across the entire Supermicro AMD Helios Rack.

AMD Pensando™ AI NICs deliver 800 Gbps open Ethernet connectivity, up to 2.4 Tbps of scale-out bandwidth per GPU, and UEC 1.0-compliant RDMA for high-throughput GPU-to-GPU communication.

Fill out the form to learn more about the new Supermicro AMD Helios Rack!