Microsoft unveils Maia 300, its third-generation in-house AI chip

On 12 October 2026, Microsoft announced Maia 300, its third-generation in-house AI chip, positioned as the "primary AI inference system" for Microsoft products and services, with mass production and deployment planned to begin in 2027. Scott Guthrie, executive vice-president of Microsoft's cloud business, said Maia 300 delivers 40% higher token throughput per megawatt than mainstream GPU systems, and at a full configuration of 96 accelerators reaches rack-level token throughput seven times that of the previous-generation Maia 200. Follow-up reports the same day said Microsoft plans to deploy Azure Maia AI chips at multi-gigawatt scale from 2027 and make them the primary inference system across its product lines; Maia 200 already runs Microsoft Copilot in the United States, Europe and Australia, with GitHub Copilot and Foundry to follow by the end of the year, and Microsoft says it delivers up to 1.7 times more tokens per watt than leading GPUs in Teams. The chip is currently at the announcement stage, with mass production and deployment set for 2027.

Written by AI from the reports · updated 4 d ago

1. Oct 12 23:24 · 1 reports
   Microsoft plans to scale in-house Maia AI chips to gigawatt level in 2027
   The Decoder：Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027
2. Oct 12 21:00 · 1 reports
   Microsoft unveils Maia 300, its third-generation in-house AI chip
   The Verge · AI：Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs