Overview
- A deal announced Monday between AMD and Microsoft commits Azure to deploy the Helios rack‑scale AI system at scale to power frontier model inference, Azure AI services, and customer workloads.
- Helios is a full‑stack rack solution that bundles 72 Instinct MI455X GPUs, 6th‑gen EPYC 'Venice' CPUs, Pensando networking DPUs, and ROCm software, and AMD says customer shipments begin in the second half of 2026.
- Azure will add two new EPYC 'Venice' VM families — HDv2 for agentic AI and data pipelines, and HXv2 for semiconductor design — to give customers new instances tuned for those workloads.
- Microsoft will expand deployment of AMD Pensando DPUs and integrate AMD silicon with Azure Boost to accelerate networking and connection processing across its cloud fleet.
- Markets reacted immediately with AMD shares rising about 5% and analyst targets moving higher, but independent reports note higher rack sticker estimates and flag that large‑scale validation of performance, cost‑per‑token and supply remains the key near‑term test.