Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs

Microsoft announced its third-generation in-house AI chip, Maia 300, positioning it as the primary AI inference system for Microsoft's products and services. The company says the chip improves the cost and efficiency of running AI workloads in the cloud and is better suited to the latest reasoning models and agentic tasks. Production rollout is slated to begin in 2027.
Scott Guthrie, executive vice president of Microsoft's cloud group, said the company expects Maia 300 to deliver 40 percent more tokens per megawatt compared to leading GPU-based systems. That metric matters because power and cooling, not just raw compute, increasingly set the ceiling on how much inference a datacenter can serve.
Microsoft framed the design around efficiency and did not detail pricing, availability beyond the 2027 production window, or which specific services will run on the chip first. The Verge's excerpt also references a "full 96-acceler" configuration without completing the figure, so the exact rack or system specification is not established here.
Why it matters: This is a direct move by Microsoft to reduce reliance on Nvidia GPUs for inference, the workload that dominates ongoing serving costs once models are deployed. For developers and product teams, broader first-party silicon could eventually shift the economics and capacity available for AI features on Azure, though the 2027 timeline means no near-term change. Treat the efficiency figure as Microsoft's own projection until independent benchmarks exist.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Microsoft is announcing its third-generation Maia in-house AI chip today, designed to be the "primary AI inference system" for Microsoft's products and services. Maia 300 not only improves the cost and efficiency of running AI workloads in the cloud, it's also more capable of handling the latest reasoning models and agentic tasks. Maia 300 will […]