Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027
Microsoft plans multi-gigawatt Azure Maia deployments from 2027 as its main inference platform.
Microsoft executive Scott Guthrie said Azure Maia chips will be deployed at multi-gigawatt scale from 2027 and become the company’s main inference system. Maia 200 already runs Microsoft Copilot in the United States, Europe, and Australia, with GitHub Copilot and Foundry to follow by year-end; CEO Satya Nadella called the silicon the company’s biggest bet. The chips serve OpenAI, MAI, and open-source models, and Microsoft says Maia 200 pushes up to 1.7 times more tokens per watt than leading GPUs in Teams. Maia 300, shipping in 2027, packs eight HBM4 stacks totaling 288 GB at 18 TB/s and is rated at 950 watts, with a claimed 40 percent more tokens per megawatt and up to seven times the tokens of a Maia 200 rack in a 96-chip system.