Microsoft Unveils Maia 300 Chip to Reduce Reliance on Nvidia GPUs
Microsoft is announcing its third-generation Maia in-house AI chip, the Maia 300, which the company designs to be the "primary AI inference system" for its products and services. The chip improves the cost and efficiency of running AI workloads in the cloud while handling the latest reasoning models and agentic tasks more capably. Production rollout begins in 2027.
Microsoft claims the Maia 300 delivers 40 percent more tokens per megawatt compared to leading GPU-based systems, according to Scott Guthrie, executive vice president of Microsoft's cloud group. In a full 96-accelerator configuration, the chip achieves up to 7x the rack-scale token throughput of the Maia 200.
These efficiency improvements allow Microsoft to rely mainly on its own AI chips instead of hardware from Nvidia and AMD. The company plans to scale the Azure Maia family to multi-gigawatt deployments starting in 2027, following the Maia 200, which already powers production AI workloads including Microsoft Copilot and GitHub Copilot.