Microsoft unveils Maia 300 and targets multi-gigawatt inference

Microsoft said on Oct. 12 that Azure Maia 300 will begin production deployments in 2027 and that the Maia family will scale to multi-gigawatt deployments that year as its primary inference system. It expects a 96-accelerator rack to deliver up to 7x Maia 200 token throughput and 40% more tokens per megawatt than leading GPU systems. CNBC reported a Microsoft general manager declined to set a firm date for that multi-gigawatt scale, and Maia chips are not offered for hourly rental.