Microsoft unveiled its Maia 300 artificial intelligence chip on October 12, with production deployments planned for 2027. The company said its Maia family would expand to multi-gigawatt scale and become the main inference system for its products and services. Yahoo Finance reported that the launch took place at the OCP Global Summit in San Jose.
The announcement extends a programme already serving real workloads. Microsoft said Maia 200 supports Microsoft Copilot and its MAI model team, running models from OpenAI, Microsoft and open-source families. GitHub Copilot and Microsoft Foundry are expected to follow by the end of this year, according to the company.
Microsoft expects Maia 300 to deliver 40% more tokens per megawatt than leading systems based on graphics processors. It also claims up to seven times Maia 200's token throughput per rack with a full configuration of 96 accelerators. These are company performance claims, rather than independent test results presented in the announcement.
An Azure technical post says the design separates computing, memory control and input-output functions across chiplets, with computing components built on TSMC's N2P process. Each accelerator has eight HBM4 memory stacks providing 288 gigabytes of capacity and 18 terabytes per second of bandwidth. The rack combines 96 accelerators within a 167-kilowatt power envelope.
The design targets long-context reasoning, responsive conversations and AI agents carrying out multiple steps. Microsoft says coordinating software, memory, networking and cooling can improve the amount of useful work delivered within data centres' power limits. Yahoo Finance described the launch as part of an effort to reduce reliance on Nvidia and AMD processors.
Production use of Maia 300 remains scheduled for next year. Microsoft said it would share more about its AI strategy at the Ignite event in San Francisco and online next month. Its stated goal is to expand access to AI while making the infrastructure more efficient and affordable.
Comments