Microsoft plugs AMD's newest chips into Azure to handle the AI compute crunch

Key points

• Microsoft announced three new Azure virtual machine types in 2025, all built around AMD's newest processor technology.
• HDv2 virtual machines pack nearly 500 CPU cores and 4 terabytes of RAM, aimed at feeding data to large AI training and agent systems.
• HXv2 virtual machines reach clock speeds above 5 GHz and carry 50 percent more cache per core than the previous generation, targeting chip designers.
• The ND MI455X v7 virtual machine uses AMD's Helios rack-scale system to run AI inference, the step where a finished AI model answers real user queries.
• AMD's CTO Mark Papermaster confirmed AMD uses Azure HX machines itself to design future EPYC and Instinct chips.

The cloud is running out of headroom. More AI models, more software agents acting on users' behalf, and a relentless appetite for raw computing power have put pressure on every part of the infrastructure stack. Microsoft's answer, announced on the official Microsoft AI blog, is three new flavours of Azure virtual machines, each built around AMD's latest silicon.

Think of a virtual machine as a rented computer that lives inside a massive Microsoft data centre. You pay for what you use, and Microsoft handles the physical hardware. The new machines give companies different tools for different jobs.

The data feeder: Azure HDv2

AI models learn from data, and agents, meaning AI software that carries out multi-step tasks on a user's behalf, need a constant supply of it. HDv2 exists to keep that pipeline full. It uses AMD's 6th-generation EPYC processors, a family of high-core-count server chips, and squeezes in nearly 500 cores, 4 terabytes of RAM and 32 terabytes of fast local storage into a single virtual machine. That is a lot of horsepower dedicated purely to moving and organising information.

The chip designer's workhorse: Azure HXv2

Designing a new processor requires simulating billions of transistors in software before a single piece of silicon gets made. That simulation work is brutally demanding. HXv2 targets exactly those customers, using AMD's 3D V-Cache technology, a trick that stacks extra memory directly on top of the processor to cut the time data spends travelling, and pushing clock speeds past 5 GHz. The previous HX machines launched in 2023 already attracted silicon design firms; HXv2 pushes further with 50 percent more addressable cache per core and up to 4 terabytes of RAM. AMD itself uses the HX series to design its own next-generation chips.

The inference engine: ND MI455X v7

Inference is the moment a finished AI model does its actual job: answering a question, generating an image, running a search. The ND MI455X v7 uses AMD's Helios rackscale platform, a system that links multiple processors across an entire server rack, to handle that work at production scale.

What does this mean for ordinary users?

Directly, nothing changes in your day-to-day apps. But the companies that build those apps now have more options for running AI cheaply and efficiently, which over time tends to mean faster responses and lower costs passed on to customers. For businesses building their own AI tools on Azure, these machines offer more specialised choices rather than forcing every task onto the same general-purpose hardware.