Microsoft to expand Azure Maia system to multi-gigawatt deployments (blogs.microsoft.com)

Microsoft has announced an expansion of its Azure Maia system, with the Maia 200 now in production and the upcoming Maia 300 set for a 2027 release. This multi-gigawatt deployment strategy marks a significant evolution in Microsoft’s AI infrastructure, positioning Maia as the primary inference system for the company's suite of products. Maia 200 currently powers workloads including Microsoft Copilot, leveraging insights from diverse AI models—OpenAI's, Microsoft's MAI models, and leading open-source frameworks—to enhance efficiency across enterprise applications. The Maia 300 system promises substantial advancements in handling next-generation AI workloads, such as long-context reasoning and complex agentic tasks. Key features include a software-defined dataflow architecture that optimizes data movement and maximizes compute efficiency, coupled with a novel multi-chiplet design that separates computing, memory, and I/O functions for enhanced scalability. With up to 7x the token throughput of Maia 200 and significantly improved power efficiency, Maia 300 is expected to revolutionize AI capabilities in cloud environments, supporting broader AI applications and enabling organizations to deploy more intelligent and effective systems. Microsoft's commitment to this infrastructure is central to its vision of "Frontier Transformation," which aims to democratize AI access while driving innovation across industries.