Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs

Microsoft unveiled its third-generation Maia artificial intelligence chip on Monday, positioning the Maia 300 as the primary AI inference system for its products and services. The company said the chip is built to cut the cost and power needed to run AI workloads in its cloud while handling newer reasoning models and agentic tasks. Microsoft plans to begin rolling the Maia 300 out to production workloads in 2027.

Scott Guthrie, executive vice president of Microsoft's cloud group, said the company expects the Maia 300 to produce 40 percent more tokens per megawatt than leading GPU-based systems. In a full 96-accelerator configuration, he said, it delivers up to seven times the rack-scale token throughput of the Maia 200. Those gains are expected to let Microsoft lean mainly on its own chips rather than those from Nvidia and AMD. Guthrie said Microsoft will scale its Azure Maia family to multi-gigawatt deployments starting in 2027, making Maia the primary AI inference system for Microsoft products and services.

The Maia 300 follows the Maia 200, which Microsoft announced only in January. The Maia 200 already runs production AI workloads across the company, including Microsoft Copilot, GitHub Copilot, Microsoft Foundry, and Microsoft's own AI models. Guthrie described the Maia 200 as the best performing AI infrastructure in Microsoft's fleet when measured by real-world metrics for AI services. He said the Maia 300 is a complete AI system built for emerging workloads such as long-context reasoning, latency-sensitive conversational experiences, and sustained multi-step agentic tasks.

To reach its efficiency targets, Microsoft adopted a new chiplet architecture that separates compute, memory controller, and I/O functions. The chip is built on TSMC's N2P process node. According to Microsoft's Azure hardware and infrastructure team, the design increases the silicon area devoted to compute performance and lets each function use the technology best suited to its role.

Microsoft AI CEO Mustafa Suleyman tied the chip to the company's own model development, which could help it work toward catching up with Google DeepMind, OpenAI, and Anthropic. He said Microsoft has to work across the full stack down to the silicon, and that at that scale it needs a platform it owns and co-designs with its models. Running Microsoft AI models on top of Maia is a critical focus for his team, he said.

More AI news from TechManNews.