Woofun AI reports that AMD has officially launched Helios, its first rack-scale AI system, with Microsoft confirming deployment in Azure data centers. This adoption adds Microsoft to a customer base that already includes Meta, OpenAI, and Oracle. The system is designed to compete directly with Nvidia’s Grace Blackwell and Vera Rubin architectures, with deliveries scheduled to begin later this year.

Microsoft plans to utilize Helios for advanced model inference and AI application development, alongside introducing compute instances based on the 'Venice' CPU. AMD highlights that Helios integrates in-house MI400 series GPUs, EPYC CPUs, and network chips to reduce cost per token. Each compute tray contains four Instinct GPUs and one EPYC CPU. Meta intends to deploy systems with up to 1GW capacity, while AMD notes that eight of the top ten global AI firms run workloads on its Instinct GPUs.