Microsoft, AMD Introduce New Azure VMs for AI, HPC, and Chip Design

Microsoft has expanded its Azure AI and high-performance computing (HPC) infrastructure through a deeper partnership with AMD to meet the growing demand for AI workloads. The expanded collaboration introduces new AMD-powered Azure services designed to accelerate large-scale AI inference, agentic AI, data processing, and semiconductor design while improving performance and scalability.

Microsoft is continuing to build a diverse Azure infrastructure that combines technologies from partners such as AMD alongside Microsoft’s own custom silicon. This approach aims to improve performance, cost efficiency, and energy usage for different AI workloads.

Specifically, Microsoft has introduced three new Azure offerings powered by AMD’s latest AI and datacenter technologies to support a wide range of advanced computing workloads. These include Azure HDv2 virtual machines, Azure HXv2 virtual machines, and Azure ND MI455X v7 virtual machines. The new offerings give customers purpose-built infrastructure tailored to different stages of the AI development and deployment lifecycle.

Azure HDv2: Designed for AI data systems

Azure HDv2 is designed to handle the data-intensive tasks that are essential for modern AI systems. These virtual machines provide the processing power required for activities such as data preparation, search, reinforcement learning, and AI agent orchestration. Azure HDv2 comes with around 500 physical 6th Gen AMD EPYC CPU cores, up to 4 TB of memory, 32 TB of local NVMe storage, and high-speed Azure networking to keep large AI workloads running efficiently while reducing data bottlenecks.

Azure HXv2: Built for chip design and scientific computing

Azure HXv2 is aimed at organizations running complex engineering and computing workloads that demand both speed and scalability. This new generation delivers enhanced single-threaded performance, larger memory options, and increased cache capacity to help accelerate electronic design automation (EDA) tasks such as chip verification and simulation. It is also useful for scientific research, engineering analysis, and other high-performance computing apps that require large-scale processing across multiple systems.

Azure ND MI455X v7: Focused on AI inference

Lastly, Azure ND MI455X v7 is designed for organizations that need to run AI models at production scale, particularly for applications involving reasoning, search, and AI agents. These virtual machines, which are powered by AMD’s Helios AI platform, are built to deliver the performance and efficiency required for demanding inference workloads.

Microsoft is also strengthening the underlying network infrastructure that supports Microsoft Azure. The company plans to deploy AMD Pensando data processing units (DPUs) across its cloud platform and integrate AMD’s networking technology with Azure Boost. This combination is intended to enhance the way network traffic is managed, which improves performance, efficiency, and scalability while reducing processing overhead.