Microsoft Plans to Deploy In-House AI Chip Maia 300 Next Year: Expected to Boost Token Output per Megawatt by 40% and Reduce Reliance on Nvidia GPUs
I'm LongbridgeAI, I can summarize articles.Microsoft plans to deploy its in-house AI chip, Maia 300, next year, aiming to boost token output per megawatt by 40% and reduce reliance on Nvidia GPUs. The chip is designed for AI inference, with Microsoft targeting gigawatt-scale deployment. While currently using Nvidia GPUs for most tasks, Microsoft seeks to improve efficiency and cut costs through custom hardware. Unlike Amazon and Google, Microsoft does not plan to rent out Maia compute power to cloud customers.
TradingKey - Microsoft (MSFT) stated that it plans to deploy its next-generation in-house AI chip, Maia300, in data centers starting next year and expand the deployment scale of the Maia series to several gigawatts, aiming to power its AI software primarily with custom hardware and reduce reliance on Nvidia GPUs.
The Maia chips are specifically designed for AI inference—using existing models to perform tasks such as summarizing meetings and answering research questions. Microsoft expects that compared with mainstream GPU systems, the Maia300 will generate 40% more tokens per megawatt.
Rani Borkar, an executive heading Microsoft's data center hardware and chip design team, stated, "We are confident that we have built the best product for these workloads." She noted that Microsoft will back this conviction with action by deploying Maia on a gigawatt scale starting next year.
Currently, most of Microsoft's AI inference tasks are still handled by Nvidia GPUs. By customizing chips for its own data centers, the company hopes to improve inference efficiency and reduce a portion of its spending on purchasing Nvidia GPUs.
Microsoft began deploying its previous-generation Maia200 earlier this year. Borkar said the chip is already being used to power AI software such as Copilot, running models from OpenAI, Microsoft's AI team, and open-source model families. Maia200 has been deployed in data centers in the U.S., Europe, and Australia, providing AI capabilities for products such as Microsoft Teams and Copilot.
Custom AI accelerators have become an important path for major cloud providers to reduce reliance on external hardware. Amazon and Google are also advancing their own strategies; having started earlier than Microsoft, both are considered leaders in the custom chip space. Microsoft's expanded deployment aims to accelerate the use of in-house hardware for its internal AI services.
However, Microsoft currently does not plan to directly rent out the underlying compute power of Maia chips to cloud customers. Borkar noted that this differs from Google and Amazon, which already offer compute capacity from their in-house AI chips to customers.