At yesterday's (November 17) Ignite 2026 event, Microsoft announced that Maia 200, its second-generation in-house AI accelerator, has officially entered volume production and has begun powering Microsoft's largest frontier AI models in its data centers. The chip's most impressive scorecard is tokens processed per watt: compared with "leading GPU infrastructure," it delivers up to 1.7x higher token throughput per watt in built-in AI features such as Microsoft Teams, Copilot, and Cowork—the same power, nearly double the output.
How much is packed into one chip


