Microsoft announced Maia 300, its next-generation custom AI chip. Coverage reports multi-gigawatt deployments for inference workloads from 2027, with a claimed 40% inference efficiency gain.

Microsoft has announced Maia 300, the next generation of its AI inference platform, and plans to expand Azure Maia to multi-gigawatt deployments beginning in 2027. With the rollout, Maia 300 will become the primary AI inference system for Microsoft products and services.

A snapshot of what's new:

Built for emerging AI workloads: The Azure Maia system is designed around the real-world workloads Microsoft runs every day and new types of AI workloads, including long-context reasoning, latency-sensitive conversational experiences and sustained, multi-step agentic tasks.

40% more tokens per megawatt: Maia 300 is expected to deliver 40% more tokens per megawatt, allowing more AI compute to be provided with the same amount of power as demand for Copilot, AI agents and other AI services continues to grow.

Up to 7x higher token throughput: At rack scale inside a Microsoft datacenter, Maia 300 systems are expected to deliver up to seven times the token throughput of Maia 200, allowing AI responses to be delivered at greater volume and speed across Microsoft's cloud services.

Building on Maia 200: Maia 200 already delivers up to 1.7 times more tokens per watt for built-in AI capabilities used by hundreds of millions of people. Maia 300 builds on this foundation to enable more efficient coding with GitHub, more flexible AI model deployment through Microsoft Foundry, and faster interactions with Copilot, Teams and Dragon Copilot.

Learn more about Maia 300 in the Official Microsoft Blog and on the Maia 300 microsite.