Microsoft unveils Maia 300 to cut reliance on Nvidia

Microsoft has announced Maia 300, its latest in-house AI chip, with production deployment scheduled to begin in 2027. Designed for long-context reasoning, latency-sensitive conversations and sustained agentic tasks, the system is intended to become the primary AI inference platform for Microsoft products and services. Microsoft plans to scale its Azure Maia family to multi-gigawatt deployments starting in 2027.

Microsoft expects Maia 300 to deliver 40 percent more tokens per megawatt than leading GPU-based systems. In a full 96-accelerator configuration, the company says it delivers up to 7x the rack-scale token throughput of Maia 200. Those efficiency improvements are expected to let Microsoft rely mainly on its own chips rather than hardware from Nvidia and AMD. The new design uses TSMC’s N2P process node and a chiplet architecture that separates compute, memory controller and I/O functions, dedicating more silicon area to computation.

Its predecessor, Maia 200, already handles production workloads for Microsoft Copilot, GitHub Copilot, Microsoft Foundry and Microsoft’s own AI models. The new hardware also supports Microsoft’s efforts to catch up with Google DeepMind, OpenAI and Anthropic. Microsoft AI CEO Mustafa Suleyman says owning the platform and co-designing it with the company’s models is a critical focus.