Microsoft Unveils Maia 200 Chip to Scale AI Inference Workloads

Microsoft (NASDAQ: MSFT) announced the launch of its newest in-house AI chip, Maia 200, positioning it as a high-performance “silicon workhorse” built to scale AI inference, which is the process of running models in production, faster and more efficiently.

The Maia 200 follows Microsoft’s Maia 100, which debuted in 2023, and comes as major AI players increasingly focus on inference as a growing driver of cost and infrastructure strain. While training large models remains expensive, the ongoing compute demand to deploy and serve them at scale has become a critical expense line item as AI products mature.

Microsoft said Maia 200 is designed to run powerful AI models with improved performance and lower disruption, helping reduce the power burden of modern AI workloads. The company said the chip contains more than 100 billion transistors and delivers over 10 petaflops of performance in 4-bit precision, along with roughly 5 petaflops in 8-bit performance, marking a significant jump from the previous generation.

Microsoft added that a single Maia 200 node is capable of running today’s largest models while leaving room for larger systems in the future. The launch underscores a broader shift across Big Tech toward custom-designed chips that can reduce reliance on Nvidia’s GPUs, which remain central to the AI boom but have become an increasingly expensive and strategically important supply chain.

The company’s move mirrors efforts from other hyperscalers, including Google’s Tensor Processing Units and Amazon’s Trainium accelerators, which are aimed at offloading some AI workloads from traditional GPU infrastructure.

Microsoft also highlighted competitive performance comparisons in its announcement, saying Maia 200 delivers three times the FP4 performance of third-generation Amazon Trainium chips, while also exceeding the FP8 performance of Google’s seventh-generation TPU. Microsoft said Maia 200 is already being used internally to support workloads tied to its Superintelligence team, and is also helping power Copilot, its AI assistant platform.

The company added that it has invited a range of external users, including developers, academics, and frontier AI labs,  to access the Maia 200 software development kit for use in their own AI workloads.

About Microsoft

Microsoft Corporation (NASDAQ: MSFT) is a global technology company that develops software, cloud services, and hardware products. The company’s offerings include the Windows operating system, Azure cloud platform, and AI-driven tools such as Copilot. Microsoft is also expanding its custom silicon strategy through chips like Maia, designed to support large-scale AI workloads across its infrastructure and enterprise customer base.

At the time of publishing, Stocks.News holds positions in Microsoft, Amazon, and Google as mentioned in the article.