Microsoft on Monday announced Maia 300, the third generation of its custom artificial intelligence chip, as the company moves to run more of its AI services on its own hardware instead of processors from Nvidia and AMD.
The company said the chip will start rolling out to production workloads in 2027. It plans to scale its Azure Maia systems to deployments of several gigawatts, making Maia the primary system for running AI in Microsoft products and services.
Scott Guthrie, executive vice president of Microsoft's cloud group, said Maia 300 is expected to deliver 40 percent more tokens per megawatt than leading GPU-based systems. In a full rack of 96 accelerators, he said, it can deliver up to seven times the throughput of the previous Maia 200.
Maia 300 follows the Maia 200, which was announced in January and already runs production services including Microsoft Copilot, GitHub Copilot and Microsoft Foundry. Microsoft said the new chip is designed for newer kinds of AI work, such as long-context reasoning, fast conversational replies and multi-step agentic tasks.
The chip is built on TSMC's N2P manufacturing process and uses a chiplet design that separates computing, memory control and input-output functions. Microsoft said this leaves more of the silicon free for processing power.
Microsoft AI chief executive Mustafa Suleyman said the company needs a platform it owns and designs together with its own AI models in order to compete at the scale of rivals such as Google DeepMind, OpenAI and Anthropic.