BlockBeats News, July 20th, AMD officially launched its first AI-focused rack-scale system, Helios, and announced that Microsoft will deploy the system in Azure data centers, becoming another key customer following Meta, OpenAI, Oracle, and other enterprises.

Helios is seen as AMD's key product in recent years to challenge Nvidia's AI infrastructure dominance. It is also the first AI rack solution directly targeting Nvidia's Grace Blackwell and Vera Rubin systems, with deliveries expected to start later this year.

Microsoft stated that it will leverage AMD Helios to support cutting-edge model inference, AI application development, and Azure AI services. Microsoft will also introduce two types of compute instances based on AMD's latest 'Venice' CPU for AI agent and semiconductor design scenarios.

AMD mentioned that Helios integrates the company's in-house GPUs, CPUs, network chips, and software ecosystem, offering a lower cost per token. The system features the MI400 series GPU, with each compute tray hosting 4 Instinct GPUs and 1 EPYC CPU.

Currently, Meta plans to deploy a Helios system with a maximum capacity of 1GW. Companies like OpenAI and Oracle have also announced the adoption of similar solutions. AMD stated that 8 out of the top 10 AI companies globally are running their workloads on its Instinct GPUs.