# AMD Helios: 72 GPUs and 31 TB HBM4 Mark a Giant Leap in AI AMD introduces Helios as a competitor to Nvidia. - **Source:** [The Next Web](https://brifing.si/source/afe24b20-b275-4c19-b9a6-8bb3c8babae7) - **Published:** 2026-07-20 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/80716cea-01fa-482a-8515-5605f34b77d9 ## What is AMD Helios? AMD Helios is an AI system that packs 72 MI455X GPUs, 31 terabytes of HBM4 memory, and 2.9 exaflops of FP4 inference power into a single rack. It’s AMD’s first rack-scale AI system and a direct response to Nvidia’s Vera Rubin NVL72. ## Technical Specifications Helios includes 18 compute trays, each housing four MI455X accelerators and a sixth-gen EPYC “Venice” CPU. The system leverages open standards like UALink and Ultra Ethernet Consortium specifications. ## Production and Launch Engineering samples of AMD Helios will ship in the second half of 2026, with mass production starting in Q2 2027. This is part of AMD’s multi-billion-dollar investment in AI infrastructure. ## Memory and Performance Each MI455X GPU features HBM4 memory with 19.6 TB/s bandwidth. The full rack delivers 260 TB/s of scale-up bandwidth and 43 TB/s of scale-out bandwidth, crucial for frontier model training. ## Software Support Helios comes with the ROCm software stack, supporting popular AI frameworks like PyTorch, TensorFlow, and JAX. This allows developers to transition from Nvidia’s CUDA ecosystem without rewriting code. ## Competition in AI AMD Helios aims to challenge Nvidia’s dominance in AI computing. While the hardware specs are competitive, the ecosystem’s success will determine Helios’ future. ## Sources covering this story - https://thenextweb.com/news/amd-helios-mi455x-72-gpu-rack-nvidia-rival — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._