Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027
decoder ·
Microsoft has announced plans to scale its proprietary Maia AI chips to multi-gigawatt levels starting in 2027. According to executive Scott Guthrie, these chips will become the primary inference system for the company's products. The current Maia 200 chip is already powering Microsoft Copilot in the US, Europe, and Australia, with GitHub Copilot and Foundry expected to follow by year-end. CEO Satya Nadella described this initiative as the company's biggest bet on its own silicon. The successor, Maia 300, is scheduled for shipment in 2027 and features eight HBM4 memory stacks totaling 288 GB of memory. Microsoft claims the new architecture offers significant efficiency gains, handling up to 1.7 times more tokens per watt than leading GPUs in Teams applications. A full technical report detailing these advancements will be released at Ignite 2026.
decoder