Microsoft Unveils Maia 300 AI Chip for Azure Infrastructure

Microsoft introduced the Azure Maia 300 AI chip at the OCP Global Summit 2026 in San Jose, aiming to expand its custom AI infrastructure and reduce dependence on third-party processors from Nvidia and AMD, according to company announcements.

The Maia 300 will join Microsoft's Maia 200 and Maia 100 chips in multigigawatt deployments beginning in 2027, according to Microsoft. The company said Maia 300 is designed for AI inference workloads involving long-context reasoning and multi-step agent tasks.

Microsoft said the Maia chips will become the "primary AI inference system for Microsoft products and services."

According to Microsoft, Maia 300 is expected to deliver 40% more tokens per megawatt compared with leading GPU-based systems. A 96-accelerator rack configuration is designed to provide up to seven times the rack-scale token throughput of Maia 200.

Microsoft said Maia 200 is already serving production workloads, including Microsoft Copilot, OpenAI models, Microsoft's MAI models, and open-source model families. The company said GitHub Copilot and Microsoft Foundry support are planned to follow by the end of the year.

Microsoft's existing Maia 200 deployments across the United States, Europe, and Australia deliver up to 1.7 times more tokens per watt for built-in AI features in Microsoft Teams, Copilot, and Cowork experiences compared with leading GPU infrastructure, according to the company.

Microsoft described Maia 300 as an inference-first system with a focus on balancing compute, memory bandwidth, networking, and power efficiency. The chip uses eight HBM4 stacks, providing 18.0 TB/s of peak memory bandwidth within a 950-watt thermal design power.

The company said the Maia 300 system uses a multi-chiplet architecture and software-defined dataflow design to improve efficiency. Microsoft also said the rack system combines 96 accelerators, 27.6 TB of HBM capacity, and 1.7 PB/s of aggregate HBM bandwidth.

Microsoft is among major technology companies developing custom AI processors as cloud providers seek more control over infrastructure. Amazon, Google, and Meta have also developed or are developing their own chips, while AI labs including OpenAI and Anthropic are working with specialized hardware strategies.