AMD Helios rack-scale AI system lands Microsoft Azure as a customer

Microsoft will deploy AMD’s Helios rack-scale AI system on Azure, the two companies said July 20, 2026, adding the chipmaker’s most powerful accelerator platform to a growing customer list. That roster already includes Meta, OpenAI, Oracle and Tata Consultancy Services.

The expanded strategic partnership is AMD’s most direct challenge yet to Nvidia in rack-scale AI hardware, the tier of tightly integrated systems that hyperscalers now buy by the rack rather than the chip. Helios is AMD’s first such system, and Azure’s endorsement hands it a marquee cloud buyer alongside the model labs.

Each Helios rack packs 72 AMD Instinct MI455X GPUs with EPYC “Venice” CPUs and Pensando networking, which AMD says delivers up to 1.4 exaFLOPS of FP8 compute and 31 terabytes of aggregate HBM4 memory. Those are vendor figures and have not been independently verified. Azure is building three new virtual-machine series on the stack, including an ND MI455X v7 line aimed at large-scale AI inference.

AMD Chief Executive Lisa Su said the companies were extending their partnership across the full stack of AMD AI solutions on Azure.

Meta went further, committing to deploy up to 6 gigawatts of AMD GPUs over time, starting with 1 gigawatt on Helios racks later in 2026. Helios begins shipping to customers in the second half of the year, and AMD shares rose roughly 3 to 5 percent on the news.

The commitments still trail Nvidia’s installed base, and none of the customers has said how much Helios capacity will go live in 2026 versus later. The endorsements give AMD credibility at the rack level; converting them into booked, running capacity is the harder test.