Microsoft has unveiled remarkable strides in its in-house chip development endeavors. The Azure Maia 200 chip has entered mass production, and the company has also launched the cutting-edge Azure Maia 300. Microsoft has set ambitious plans to scale up the deployment of Maia series chips to multi-gigawatt levels starting from 2027. These chips are poised to become the cornerstone inference system for Microsoft's array of products and services. Presently, the Maia 200 is already being integrated into some of Microsoft's production workloads. There are plans to broaden its application to encompass more services by the end of this year, and it will provide support for multiple mainstream AI models. The Maia 300, tailored specifically for new AI workloads, is anticipated to offer a 40% increase in token output per megawatt when compared to leading GPU systems. Microsoft is ramping up its in-house chip development efforts to forge a closed loop between models and chips. This strategic move is perceived by industry outsiders as a means to cut costs and diminish reliance on external suppliers. At present, tech giants such as Google and Amazon are also expediting their in-house chip development. Notably, Amazon's relevant chip business has already been embraced by a multitude of AI startups.