Strengthening Azure’s AI Computational Power

The global demand for high-performance AI computing continues to surge, driven by the emergence of sophisticated open models like Kimi K3. To address this growing need, Microsoft and AMD have announced a strategic partnership that will see the widespread deployment of AMD’s Helios rack-scale AI accelerators across Microsoft’s data centers. This initiative is designed to boost AI FLOPS for both internal development and the expanding ecosystem of Azure AI customers.


Strategic Benefits for Azure Ecosystem

By incorporating AMD’s latest hardware, Microsoft aims to provide its users, including research labs and enterprise clients, with the necessary compute power for intensive AI training and inference tasks. The integration will specifically support managed compute services for corporate clients utilizing Microsoft Foundry to deploy their AI applications.

This commitment represents a significant milestone for AMD as it continues to challenge Nvidia’s dominance in the data center GPU market. Following similar high-profile collaborations with Meta and OpenAI, this deal further cements AMD's position as a key supplier in the rapidly evolving infrastructure race.


Technical Capabilities of the Helios System

The Helios architecture is engineered to compete directly with high-end solutions like Nvidia’s Vera Rubin NVL72 system. Key technical highlights include:

  • GPU Integration: A unified system featuring 72 next-generation Instinct MI455X GPUs.
  • Memory Capacity: An aggregate of 31.1TB of HBM4 memory.
  • Performance: Capable of delivering up to 1.4 exaFLOPS of FP8 compute and 2.9 exaFLOPS of FP4 performance for optimized AI workloads.
  • Bandwidth: Designed for 260 TB/s of internal scale-up bandwidth, matching current industry leaders.

New VM Series and Infrastructure Enhancements

In addition to the Helios accelerators, Microsoft is broadening its use of AMD hardware within Azure. The company plans to introduce two new virtual machine series based on AMD’s upcoming sixth-generation Epyc Venice CPUs:

  • HDv2 series: Tailored for agentic AI tasks and complex data pipelines.
  • HXv2 series: Optimized specifically for semiconductor design workflows.

Furthermore, Microsoft will continue to leverage AMD Pensando DPUs to enhance its Azure Boost offerings, significantly accelerating storage and networking operations to ensure high efficiency across its infrastructure.