Microsoft Unveils Maia 300 AI Chip in Bid to Loosen Nvidia's Grip
Microsoft introduced the Azure Maia 300 at the OCP Global Summit 2026, claiming 40% more tokens per megawatt than leading GPU-based systems as it seeks to cut its reliance on Nvidia and AMD.
The Maia 300 joins Microsoft's earlier Maia 200 and Maia 100 chips. Microsoft said the Maia line will become the "primary AI inference system for Microsoft products and services." The company did not say when the chip will be generally available or who will manufacture it, according to a Yahoo Finance report cited by Slashdot.
Microsoft said the Maia 300 delivers 40 percent more tokens per megawatt than "leading GPU-based systems," a comparison aimed squarely at Nvidia's data-center processors. A rack of 96 Maia 300 chips offers seven times the rack-scale token throughput of Microsoft's Maia 200, the company said.
The Maia 200 is already running Microsoft production workloads, including Microsoft 365 Copilot, along with OpenAI models, Microsoft's own MAI models and open-source models. Microsoft said the Maia 200 provides 1.7 times more tokens per watt for AI features in Teams, Copilot and Cowork than other GPU-based systems.
Microsoft has designed its own silicon while remaining one of Nvidia's largest customers, and the Maia 300 announcement places the company more directly against the chipmaker whose GPUs dominate AI training and inference. AMD, which also supplies accelerators to cloud providers, is affected by the same push toward in-house processors.
The performance figures are Microsoft's own, as carried in the Yahoo Finance report, which is the only source for the announcement details. No third-party benchmark data accompanied the claims.