Microsoft Moves Maia 200 AI Accelerator Into Scaled Production

Microsoft Moves Maia 200 AI Accelerator Into Scaled Production

Microsoft moved its second generation Maia 200 artificial intelligence accelerator into scaled production, announcing the step at Ignite 2026. The custom silicon runs Microsoft’s largest frontier AI models and delivers up to 1.7 × higher token throughput per watt than leading GPU solutions. The chips are already deployed in data centers across the United States, Europe and Australia, supporting live workloads such as Microsoft Copilot.

TSMC fabricates the Maia 200 on a 3 nm process optimized for AI inference. The architecture features FP4 and FP8 tensor cores, 216 GB of HBM3e memory and 272 MB of on chip SRAM. A single chip reaches more than 10 petaFLOPS at FP4 precision, and up to 6 144 accelerators can be linked over Ethernet to form large processing clusters.

Power efficiency drives the shift to custom silicon. Microsoft measured a 1.7 × gain in tokens per watt for native AI features in Microsoft Teams, Copilot and Cowork, reducing electricity use without slowing response times. GitHub Copilot and Microsoft Foundry will transition to the chips in upcoming phases. By 2027, Microsoft plans to expand Azure Maia deployments to multiple gigawatts of capacity, making the platform its primary inference infrastructure.

About the author

mgtid
mgtid
Owner of Technetbook | 10+ Years of Expertise in Technology | Seasoned Writer, Designer, and Programmer | Specialist in In-Depth Tech Reviews and Industry Insights | Passionate about Driving Innovation and Educating the Tech Community Technetbook

Join the conversation

Newsletter Subscription