Microsoft unveils Maia 200 AI chip to rival Nvidia

Microsoft unveiled the Maia 200 AI chip with 140B transistors, 216GB HBM3e memory, and 30% better performance per dollar, aiming to rival Nvidia and boost cloud AI efficiency.

Microsoft has introduced the Maia 200, its second-generation AI chip, designed to challenge Nvidia’s dominance in enterprise AI acceleration and enhance its cloud computing capabilities. The Maia 200 boasts up to 140 billion transistors, 216GB of HBM3e memory, and is manufactured using TSMC’s advanced 3-nanometer process. Delivering up to 10 petaflops of performance, the chip offers a 30% improvement in performance per dollar over previous models, making it Microsoft’s most efficient and cost-effective inference processor yet. It is tailored for demanding workloads, including OpenAI’s GPT-5.2, Microsoft 365 Copilot, and Foundry applications, and is already operational in Microsoft data centers. This strategic move aims to reduce Microsoft’s reliance on Nvidia, with plans for broader availability to customers and developers. The announcement has been well received by investors, as reflected in a rise in Microsoft’s share price.