Microsoft announced Maia 300 on October 12 and plans production deployments in 2027. Its official announcement says Maia 200 already serves workloads including Copilot. The new design is intended for AI inference, including long-context reasoning and multi-step agent tasks.
Reuters independently reported the announcement on October 12 and interviewed Microsoft hardware executive Rani Borkar. Maia 300 uses a chiplet design, with racks containing 96 accelerators. Microsoft plans to expand the Maia family to multi-gigawatt deployments starting next year.
Microsoft expects 40% more tokens per megawatt than leading GPU-based systems. Reuters noted that it did not name the comparison systems. This is a company performance claim, not an independently reproduced benchmark. The announcement does not mean Maia 300 is already deployed or available to every Azure customer.