Microsoft unveiled its third-generation in-house AI accelerator, Maia 300, on October 12, while announcing that Maia 200 is now in production across data centers in the US, Europe, and Australia. Maia 200 delivers up to 1.7 times more tokens per watt than leading GPU infrastructure and currently powers Microsoft Teams, Copilot, and Cowork, with GitHub Copilot and Microsoft Foundry expected to follow by year-end. Maia 300, slated for production deployment in 2027, promises 40% more tokens per megawatt than leading GPU-based systems and up to seven times the rack-scale token throughput of Maia 200 in a full 96-accelerator rack. Microsoft said it plans to scale the Azure Maia family at multi-gigawatt scale in 2027, according to a blog post by Executive Vice President Scott Guthrie.
Microsoft's plan to deploy custom silicon at multi-gigawatt scale signals a major shift in how hyperscalers manage AI infrastructure costs and performance, reducing reliance on Nvidia and AMD GPUs for internal workloads. Unlike AWS and Google Cloud, Microsoft has not made Maia available to external customers via Azure instances, meaning efficiency and cost gains accrue internally rather than being passed through as a competitive cloud offering. The scale of the planned rollout has direct implications for GPU procurement volumes and data center power planning across Microsoft's global footprint.
Microsoft announcing a new custom AI chip (Azure Maia 300) is a major product development signal for hyperscaler compute strategy and data center infrastructure planning. This specific chip announcement does not appear in the already-published list.