Microsoft announced on October 12, 2026 that it will deploy its homegrown Maia 300 AI inference chip at gigawatt scale starting next year. The chip is built on TSMC's N2P node and delivers up to 7x the rack-scale token throughput of the Maia 200 at 96 accelerators, with Microsoft claiming 40% more tokens per megawatt than leading GPU systems. Nadella announced it himself, calling it Microsoft's biggest bet on custom silicon, and production starts in 2027. Microsoft will use it for internal workloads like Copilot and GitHub and will not rent the compute to third parties. Reporting since says Microsoft tried to convince Anthropic to become a Maia customer and that no Azure customer has signed on to rent the…