Microsoft Plans September 2026 Launch of Maia 300 AI Chip, Targets Big Output Boost

The company is in talks with TSMC over a capacity contract for the delivery of more than 300,000 chips in 2027, an order-of-magnitude increase from the current Maia 200 generation, which has produced only tens of thousands of units. Microsoft ultimately hopes to secure over one million units of Maia 300 capacity, but the actual scale remains uncertain due to component supply and capacity negotiations. Microsoft is also in discussions with large cloud customers such as Anthropic about using Maia chips, and plans to shift more internal AI workloads to its self-developed silicon while continuing to rent Nvidia chips to Azure cloud customers.

Microsoft's progress in self-developed chips lags some competitors. The Maia 200 was delayed in 2025 after early testing missed internal targets, and has since been deployed in only a few data centers. As of the end of last month, the chip was still running in just two data centers within the United States. Among competitors, Google and Amazon's AI chips have been adopted by multiple large customers, with Google even starting to sell its TPU chips to external customers. Microsoft Chief Executive Satya Nadella said in June 2026 that he planned to extend Maia 200 to more data centers, including overseas deployments, but as of end of last month, actual deployment remained limited to two U. S. facilities.

Cost advantages are a core driver of Microsoft's self-developed chip strategy. Last month, the company disclosed to investors that Maia 200, when running OpenAI and Microsoft's own models, has operating costs 30% to 40% lower than Nvidia's leading chips. Data centers, which typically require several gigawatts of power, can house millions of AI chips.

Microsoft's self-developed chip strategy is tied to its concerns over chip costs and control of AI models. In a 2022 email later made public through judicial proceedings, Nadella mentioned the high cost of running OpenAI models on Azure and noted that the company could control neither chip costs nor the AI models themselves. Since then, Microsoft has pursued both self-developed chips and proprietary models. Currently, Microsoft still relies primarily on Nvidia chips to run most of its AI software, including the MAI models that power Copilot.

Microsoft's other self-developed chip line, the Cobalt central processing unit, has recently gained more customer adoption. Last week, Microsoft announced that large customers such as OpenAI and Adobe have adopted Cobalt in more than 25 data centers worldwide. Cobalt is a traditional CPU, distinct from AI-specific chips, but demand for CPUs is also rising amid the current growth in AI infrastructure needs. Cobalt's adoption supports Microsoft's overall self-developed silicon strategy.