Microsoft brings AMD Helios rackscale to Azure for inference

On July 20, 2026, AMD announced an expanded strategic partnership with Microsoft spanning AMD GPUs, CPUs, networking and software on Azure. The centerpiece: Microsoft will deploy the AMD Helios Rackscale Solution to power frontier-model AI inference for Microsoft, its AI customers and Azure AI services.

Helios is an open, integrated rackscale platform combining AMD Instinct MI455X GPUs, EPYC "Venice" CPUs, Pensando networking and the ROCm software stack for large-scale training and inference.

Key points

• AMD begins shipping Helios to customers, including Microsoft, in H2 2026.
• Azure adds two VM series on 6th Gen EPYC "Venice": HDv2 (agentic AI and data pipelines) and HXv2 (semiconductor design).
• Microsoft broadens Pensando DPU deployment across AI backend networking and select Azure services, and integrates AMD silicon with Azure Boost for cloud-scale networking performance.
• Frontier model builders can train and serve on AMD-powered infrastructure; enterprises can run production AI workloads via Azure Foundry Managed Compute.
• Lisa Su (AMD) and Satya Nadella (Microsoft) both framed the move as giving customers more performance, scale and choice.

FAQ

What changes for developers running AI on Azure? Another infrastructure option: once Helios lands, inference workloads can run on the AMD stack (Instinct + ROCm) rather than a single GPU vendor.

When is it available? AMD says Helios ships in H2 2026; per-service Azure availability is announced separately by Microsoft.

Does code need porting? The release gives no migration detail; AMD's software layer is ROCm, so CUDA-centric teams should budget porting effort.

Full detail is in AMD's official press release (see source).