Microsoft said on October 12 that its Azure Maia 200 accelerator is already serving production workloads, including Microsoft Copilot. It plans to begin multi-gigawatt deployments of its Maia family in 2027 and says Maia 300 production deployments are also planned for that year.
An in-house inference system for several services
Microsoft says Maia 200 is deployed across the United States, Europe and Australia, and that GitHub Copilot and Microsoft Foundry are expected to follow by the end of 2026. Maia 300 is designed as a system spanning accelerators, memory, networking and dataflow, with a 96-accelerator rack configuration.
The company estimates Maia 300 will deliver 40% more tokens per megawatt than leading GPU-based systems and up to seven times the rack-scale token throughput of Maia 200. Those are Microsoft's design and workload claims; the announcement does not provide a third-party production benchmark.