AMD has expanded its strategic partnership with Microsoft to cowl GPUs, CPUs, networking, and software program throughout Azure infrastructure. The collaboration consists of Microsoft’s deliberate deployment of AMD’s Helios Rackscale Answer for AI inference workloads supporting frontier fashions, Azure AI companies, and buyer functions. The deployment makes Microsoft the primary hyperscaler to publicly decide to Helios at scale, the clearest sign but that AMD’s rack-scale platform is touchdown as a manufacturing different to NVIDIA’s rack-scale methods relatively than a reference design.
Helios is AMD’s rack-scale AI platform, integrating Intuition MI455X GPUs, sixth Gen AMD EPYC Venice CPUs, Pensando networking know-how, and the ROCm software program stack. The platform is designed to supply an open structure for large-scale AI coaching and inference, positioning AMD to supply an alternate infrastructure stack for cloud suppliers and organizations constructing massive mannequin environments.
AMD’s companion weblog frames Helios as a system-level play relatively than a element sale: a completely built-in platform designed round AMD compute, networking, software program, energy, and cooling, constructed to span coaching, inference, fine-tuning, and agentic workloads relatively than optimizing for any single one. The division of labor is specific: AMD provides the silicon foundations, whereas Microsoft wraps them in Azure’s cloud operations, developer companies, safety, and AI platform companies; the businesses’ shared argument is that rack-scale infrastructure turns into extra priceless when it arrives inside a whole cloud platform prospects can devour at scale. The timing is not any accident both: the announcement lands simply forward of AMD’s Advancing AI 2026 occasion, the place Microsoft and AMD are co-presenting classes on sovereign AI, AI economics, and production-scale infrastructure.
For Azure, the deployment targets inference workloads throughout Microsoft companies and enterprise functions. Frontier mannequin builders will have the ability to use AMD-powered Azure infrastructure for coaching and deployment, whereas enterprise prospects can handle manufacturing AI workloads by means of Azure Foundry Managed Compute.
The businesses are additionally extending their CPU collaboration with two new Azure VM households based mostly on sixth Gen AMD EPYC Venice processors. Azure HDv2 is meant for agentic AI and knowledge pipeline workloads, whereas Azure HXv2 targets semiconductor design use circumstances. These cases broaden Azure’s current AMD EPYC footprint throughout AI, data-intensive, and engineering workloads.
Networking is one other element of the partnership. Constructing on Microsoft’s current broad deployment of AMD Pensando DPUs, Azure is extending them into AMD AI backend networking infrastructure and choose Azure companies, and the businesses are integrating Azure Enhance with AMD applied sciences to enhance networking efficiency, effectivity, and connection processing at cloud scale. That issues most for AI clusters, the place east-west site visitors between GPUs grows sooner than anything within the knowledge middle.
AMD and Microsoft characterised the expanded association as an effort to supply scalable AI infrastructure spanning mannequin improvement, inference, knowledge preparation, search, and reinforcement studying. For AMD, the settlement broadens the function of its GPU, CPU, DPU, and software program portfolio inside a serious hyperscale cloud surroundings.
AMD expects to start transport Helios methods to prospects, together with Microsoft, throughout the second half of 2026.
