Microsoft is preparing to launch its Maia 300 AI chip as it scales custom silicon production and seeks to reduce its dependence on Nvidia processors.
Microsoft is preparing to introduce its next-generation Maia 300 artificial intelligence (AI) chip this autumn, potentially as early as next month, as the company steps up efforts to develop its own AI computing hardware and reduce dependence on Nvidia processors.
According to a report by The Information, Microsoft is working to significantly increase production of the Maia 300 and is in discussions with Taiwan Semiconductor Manufacturing Company (TSMC) to secure manufacturing capacity. The company is reportedly seeking capacity for more than 300,000 chips for delivery in 2027.
Microsoft also aims to eventually secure production capacity for more than 1 million Maia 300 chips. However, component availability and ongoing negotiations with TSMC could affect the scale and timing of production.
The move comes as Microsoft seeks to catch up with Google and Amazon, which have expanded their own AI chip programmes. Google has begun generating revenue from direct sales of its Tensor Processing Units (TPUs), while Amazon is seeing increasing adoption of its internally developed AI processors, including Trainium.
Microsoft launched its first Maia AI accelerator in November 2023 and introduced the second-generation Maia 200 in January 2026. The Maia 200 is manufactured by TSMC using a 3-nanometre process and features a large amount of static random access memory (SRAM), designed to improve performance for AI workloads handling high volumes of user requests.
The company is also looking to encourage major cloud customers, including Anthropic, to adopt its custom chips as production scales up.
Microsoft said its custom silicon programme remains part of its long-term AI infrastructure strategy, while declining to disclose production volumes.
The Maia 300 rollout could strengthen Microsoft’s efforts to build a broader in-house AI hardware ecosystem and reduce its reliance on Nvidia’s high-cost AI processors.
