Microsoft has officially announced the development of its third-generation in-house AI processor, the Maia 300. The company intends for this hardware to serve as the primary system for AI inference across its various products and services. By moving toward proprietary silicon, Microsoft aims to decrease its reliance on third-party hardware providers, specifically Nvidia, which currently dominates the market for AI-capable graphics processing units (GPUs).

According to Microsoft, the Maia 300 is engineered to optimize both cost and energy efficiency for cloud-based AI workloads. The company claims the chip is specifically tailored to handle complex reasoning models and agentic tasks, which are becoming increasingly central to modern AI development. Microsoft projects that the Maia 300 will achieve a 40 percent improvement in tokens per megawatt compared to existing GPU-based systems currently in use. The hardware is scheduled for a production rollout starting in 2027. While the announcement highlights a strategic shift in Microsoft's hardware supply chain, the long-term impact on the broader semiconductor market remains to be seen as the company continues to balance its internal chip development with its ongoing partnerships with major GPU manufacturers.