Microsoft plans multi-gigawatt rollout of next-gen Azure Maia AI accelerator racks

Microsoft has unveiled plans to roll out a multi-gigawatt rollout of its next-generation Azure Maia AI accelerator racks. The company's Azure Maia 200, introduced earlier this year, is already powering production workloads. A successor, Azure Maia 300, is expected to debut in 2027. Microsoft describes Azure Maia 300 as a complete intelligent system designed to handle new AI workloads.

This system is created with an inference-first architecture optimized for long-context, memory-intensive, and decode-heavy workloads. Each Maia 300 will feature eight HBM4 memory stacks, delivering up to 18 TB/s of peak memory bandwidth. Microsoft envisions deploying these racks-scale systems comprising up to 96 accelerators. The company emphasizes the necessity of a system-level approach, integrating software, silicon, physical rack infrastructure, networking, and datacenters co-designed to function as a single platform for cloud-scale agentic inference.

This marks a shift from Microsoft's 2024 AI Access Principles, which highlighted partnerships across its technology stack. The growing emphasis on in-house integration differs from Microsoft's stance in 2024, where the company criticized Google's vertical integration, citing it as a competitive advantage. Microsoft now boasts its own AI chips, rack-scale designs, and internally developed MAI models, enhancing its control over the technology stack.

The company aims to scale Azure Maia to multi-gigawatt deployments beginning in 2027 and intends to share more details on its strategy at Microsoft Ignite 2026 in San Francisco.