Microsoft Expands Azure AI and HPC Infrastructure With AMD
Microsoft Expands Azure AI and HPC Infrastructure with AMD
As AI workloads scale faster than single infrastructure approaches can support, Microsoft is evolving Azure’s infrastructure by expanding its AI fleet with AMD’s most advanced AI and high-performance computing (HPC) solutions, according to the official Microsoft Blog announcement. Surging compute demand and growing agentic workloads are driving the need for greater specialization across the cloud stack, prompting the introduction of new virtual machine offerings built on AMD’s latest silicon.
The Tech TL;DR:
- New AMD-Powered Instances: Microsoft introduced Azure HDv2, HXv2, and ND MI455X v7 virtual machines to target heavy data processing, semiconductor design, and large-scale AI inference.
- Hardware Specs: HDv2 VMs pack nearly 500 physical 6th Gen AMD EPYC CPU cores, 4 terabytes of RAM, and 400 Gb Azure Boost networking.
- Ecosystem Integration: The HXv2 series leverages 3D V-cache technology and 800 Gb InfiniBand to optimize Synopsys AI-powered electronic design automation (EDA) workflows.
Architectural Requirements for Agentic Workloads
Modern AI systems require dense, power-efficient CPU compute to handle data preparation, search, and agent coordination at scale. AI accelerators depend on underlying CPU infrastructure to keep data pipelines full. Without sufficient throughput, training jobs starve and agentic workflows stall.
To eliminate these hardware bottlenecks, the newly announced Azure HDv2 virtual machines feature nearly 500 physical 6th Gen AMD EPYC CPU cores, 4 terabytes of RAM, 32 terabytes of local NVMe storage, and 400 Gb Azure Boost networking. These configurations are engineered to process massive data volumes and reinforce distributed reinforcement learning loops.
Optimizing Silicon Design with Azure HXv2 and Synopsys
Developing next-generation AI silicon requires immense compute power for electronic design automation (EDA) and register-transfer level (RTL) simulation. Building on the Azure HX instances launched in 2023, the new HXv2 virtual machines incorporate AMD’s 3D V-cache technology to boost single-threaded performance and memory efficiency.
HXv2 VMs feature 176 AMD 6th Gen EPYC CPU cores running at clock frequencies exceeding 5 GHz, alongside 50% more addressable cache per core and memory footprints scaling up to 4 terabytes of RAM. The inclusion of 800 Gb InfiniBand supports large-scale MPI-based simulations for engineering analysis.
“Engineering teams are pushing the limits of simulation, chip design and scientific computing,” said Mark Papermaster, Executive Vice President and CTO at AMD, noting that Azure HX serves as a critical platform for scaling complex EDA workloads.
This hardware optimization extends directly to software tooling. Shankar Krishnamoorthy, Chief Product Development Officer at Synopsys, stated that the collaboration enables customers to run AI-powered EDA solutions on Azure, extending workloads beyond traditional infrastructure constraints to maximize design quality and accelerate development schedules.
Scaling Production Inference via AMD Helios
For reasoning, search, and production-scale inference, Microsoft is deploying the ND MI455X v7 virtual machine series. Powered by the AMD Helios rackscale solution, this infrastructure addition diversifies Azure’s heterogeneous compute options, allowing enterprise engineering teams to select tailored hardware tiers for specific machine learning pipelines.
Strategic Deployment and Future Outlook