Microsoft to deploy new AMD Helios AI on Azure, next-gen Epyc-powered virtual machines

Microsoft is expanding its partnership with AMD as it looks to beef up Azure’s AI infrastructure with the chipmaker’s latest technology. The announcement sees Microsoft becoming one of the first cloud providers to deploy AMD’s new Helios Rackscale Solution, designed to handle large-scale AI inference workloads for frontier AI models, Azure AI services and enterprise customers.

Helios is the company’s next-generation AI rack-scale platform that combines Instinct MI455X GPUs, 6th Generation Epyc “Venice” processors, Pensando networking and the ROCm software stack into a single integrated system. AMD says Helios will start shipping to customers, including Microsoft, in the second half of 2026.

For Microsoft, the deployment is aimed at supporting AI inference rather than model training. The company says Helios will be used to run its own frontier AI models, power Azure AI services and provide AI compute for Azure customers building their own applications. Azure’s new virtual machines, the HDv2 series, will also be powered by 6th Generation Epyc “Venice” processors.

On top of that, Microsoft is expanding its deployment of AMD Pensando data processing units (DPUs), while integrating AMD silicon with Azure Boost, Microsoft’s in-house networking acceleration platform. The companies say the combination will improve networking performance, efficiency and connection processing across Azure’s cloud infrastructure.

AMD and Microsoft have spent years building high-performance infrastructure together, and today we’re extending that partnership across the full stack of AMD AI solutions on Azure.

Microsoft’s new AMD deployments mark an important milestone as we deliver leadership compute solutions to Azure customers and scale the next generation of AI infrastructure together,” – Dr. Lisa Su, AMD Chair and CEO

Microsoft says the expanded partnership will provide Azure customers with more hardware choices for AI workloads ranging from data preparation and search to inference and reinforcement learning.

Read more of our articles below!