Microsoft Unveils Maia 300 AI Chip, Plans 2027 Rollout to Replace Nvidia Dependency

Microsoft announced its third-generation Maia 300 in-house AI chip on October 12, 2026, positioning it as the "primary AI inference system" for its products and services.

Scott Guthrie, EVP of Microsoft's cloud group, stated that Maia 300 is expected to deliver 40 percent more tokens per megawatt compared to leading GPU-based systems. In a full 96-accelerator configuration, it achieves up to 7 times the rack-scale token throughput of the previous generation, Maia 200.

Built on TSMC's N2P process node with a new chiplet architecture separating compute, memory, and I/O, the chip aims to optimize silicon area for specific functions. Microsoft plans to scale Azure Maia family to multi-gigawatt deployments starting in 2027.

While Maia 200 currently supports production workloads like Copilot, Maia 300 is designed for emerging tasks such as long-context reasoning and agentic workflows. Microsoft AI CEO Mustafa Suleyman emphasized that owning the platform down to the silicon is critical for competing with leaders like OpenAI and Anthropic.