Microsoft will unveil its next-generation Maia 300 AI chip in September 2026 and is negotiating with Taiwan Semiconductor Manufacturing Co (TSMC) to produce more than 300,000 units for delivery in 2027 [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]. Microsoft aims to eventually produce over 1 million Maia chips, but availability of components and TSMC manufacturing capacity may limit output [2, 4, 8, 10].
The Maia 300 chip uses TSMC's advanced 3-nanometer process and CoWos packaging technology to boost AI performance, incorporating a large amount of SRAM memory to accelerate AI workloads [2, 8, 10]. This follows Microsoft's earlier launches of the first Maia chip in November 2023 and the second-generation Maia 200 chip in January 2026, also made on TSMC's 3nm process [2, 5, 10]. However, the Maia 200 rollout was delayed and remains deployed only in limited data centers due to poor internal test results [8].
Microsoft is developing its own AI chips partly to reduce dependence on Nvidia's costly AI processors [1, 2, 3, 5, 6, 8, 9, 10]. It is courting major cloud customers, including Anthropic, to adopt the Maia 300 chip [2, 3, 5, 8, 9, 10]. The Azure Maia division general manager said, "The ultimate goal is to deploy Maia chips at a scale of gigawatts" [1, 4, 7]. Andrew Wall, Microsoft Azure Maia general manager, noted the company’s ongoing investment in custom chips as part of its long-term AI infrastructure strategy but declined to confirm production volumes, saying reported numbers do not reflect the full scale of Microsoft’s plans [10].
Microsoft lags competitors like Alphabet and Amazon that have already scaled custom AI chips for their cloud businesses [2, 7, 10]. If Microsoft expands Maia chip production, Taiwan’s semiconductor sector including TSMC and ASIC design firms like Global Unichip could benefit [10].
Microsoft plans to start delivering Maia 300 chips produced by TSMC in 2027 [1, 2, 3, 4, 5, 6, 7, 8, 9, 10].
