Microsoft's In-House Chip Maia 200 Enters Mass Production, Surpasses GPU Infrastructure in Performance per Watt
On November 17, Microsoft announced at Ignite 2026 that Maia 200, its second-generation in-house AI accelerator, has entered large-scale production and has begun supporting its largest-scale frontier AI models in data centers. The chip focuses on token throughput per watt. In built-in AI features such as Teams, Copilot, and Cowork, it delivers up to 1.7x improvement over leading GPU infrastructure, producing nearly twice the output at the same power, and is designed for AI inference.