AMD has introduced Helios, a next-generation AI rack designed for large-scale artificial intelligence deployments. The system is positioned as a high-end platform for frontier AI workloads and sovereign computing, putting it directly into the growing market for rack-scale AI infrastructure.
According to the announced specifications, the Helios AI Rack is built around 72 MI455X GPUs and 6th Gen EPYC processors with a combined 256 CPU cores. AMD is also highlighting a massive 31TB of HBM4 memory, signaling that the system is aimed at training and inference tasks that demand dense compute and high memory bandwidth.
The company’s latest move also underscores its effort to challenge NVIDIA’s dominance in rack-scale AI systems. By packaging GPUs, CPUs and memory into a single large system, AMD is targeting customers that want integrated infrastructure for advanced AI models and national or enterprise-level deployments.
Helios is expected to sit at the very top of the AI hardware market, with reported pricing in the $5 million to $5.5 million range. That places it firmly in the category of premium AI platforms built for organizations pursuing cutting-edge model development and other compute-intensive workloads.
