'Helios AI platform' and EPYC processors set to power Azure's next AI push

To meet surging compute demand and the rise of agent-driven AI workloads, Microsoft is taking the next step in expanding its Azure infrastructure. The company plans to strengthen its AI fleet with AMD's latest AI and high-performance computing technologies. Building on their existing partnership, Microsoft will bring AMD's Helios AI platform and next-generation EPYC data centre processors to Azure.

These technologies will power three upcoming Azure offerings: HDv2 virtual machines for AI data processing, HXv2 virtual machines for electronic design automation (EDA), and ND MI455X V7 virtual machines for AI inference workloads.

The new Azure HDv2

Azure HDv2 virtual machines are Microsoft's latest offerings, designed from the ground up to eliminate bottlenecks such as workload coordination and limited CPU compute capacity. Developed in collaboration with AMD, they expand Azure's portfolio of purpose-built infrastructure for demanding CPU-intensive AI workloads, including data preparation, search, reinforcement learning and large-scale agent coordination.

More from Tech

HDv2 virtual machines feature nearly 500 physical 6th Gen AMD EPYC CPU cores, 4TB of RAM, 32TB of local NVMe storage, and 400 Gb Azure Boost networking, making them well suited for some of Microsoft's most demanding AI customers.

The new Azure HXv2

Azure's existing HX virtual machines gained popularity for their 3D V-Cache technology, which saw significant adoption among silicon design firms developing next-generation AI chips. Building on that success, Microsoft has unveiled the new Azure HXv2 virtual machines, powered by AMD's 6th Gen EPYC processors, to meet growing demand for AI and high-performance computing workloads.

Building on the existing HX platform, HXv2 delivers faster single-threaded performance, larger memory configurations of up to 4TB RAM, increased cache per core and 800 Gb InfiniBand connectivity. These improvements make it well suited for AI chip design, scientific simulations and other compute-intensive applications.

Production-scale AI inference: ND MI455X V7

The ND MI455X V7 virtual machines are designed for modern AI services, including reasoning, search and agentic workloads. Powered by AMD's rack-scale solution, they expand Azure's infrastructure options for large-scale AI inference and are designed to deliver strong performance and greater efficiency for demanding AI workloads.

Together, these new Azure offerings give customers greater flexibility to choose the computing infrastructure best suited to their AI workflows. Microsoft said customer choice remains a core design principle for Azure, and its expanded partnership with AMD is expected to drive the next wave of innovation needed to support rapidly growing AI computing demands.