Microsoft to unveil Maia 300 AI inference chip in September, marking the company's strategy to reduce dependency on Nvidia GPUs.

Microsoft is planning to unveil its new Maia 300 AI chip this fall, possibly as soon as September, according to reports from The Information citing people with direct knowledge of the plans. The company is scaling up its efforts to reduce reliance on Nvidia and challenge rivals including Alphabet and Amazon.

Microsoft introduced its original Maia AI chip in November 2023 but has lagged behind competitors in ramping up production. The company is now in talks with chipmaker TSMC to secure manufacturing capacity for more than 300,000 units of the Maia 300 for delivery in 2027. Microsoft ultimately aims to secure capacity for more than 1 million Maia 300 chips, though component suppliers and ongoing capacity negotiations with TSMC could constrain these plans. The company's second-generation Maia 200 chip was unveiled in January and built by TSMC using its 3-nanometre process. The Maia 200 processor is capable of running OpenAI's GPT-5.2 model.

A key strategic opportunity lies in Anthropic. Microsoft has been in talks to supply its custom AI chips to the company, which could represent a significant win given Anthropic's scale. However, Anthropic has not yet closed a deal with Microsoft over Maia adoption. Historically, Anthropic has relied heavily on Nvidia GPUs to train and run generative AI models. The company signed a 10-year arrangement with Amazon Web Services in April worth more than $100 billion to use AWS's custom Trainium chips, and announced plans in October to use Google's Tensor Processing Unit chips.

Preliminary assessments of the Maia 300 claim the chip delivers more than a 30% improvement in token-per-dollar efficiency compared with the latest silicon currently in Microsoft's fleet.