Maia 300 reveal could land in September as Microsoft books TSMC capacity for 2027

Microsoft is hoping to publicly launch its Maia 300 AI accelerator as soon as September, according to a report by The Information.

The company is also in talks with TSMC to secure more than 300,000 units by next year.

Microsoft to reveal Maia 300 during the fall

Microsoft has contacted Taiwan Semiconductor Manufacturing Company (TSMC) to reserve capacity for the chips, which are due to be delivered in 2027. The order alone would exceed the tens of thousands of Maia 200 parts produced so far.

Microsoft’s longer-term target is to exceed one million units, but packaging negotiations and supply of components may cap the actual number the company gets.

The Maia 200 slipped after early tests failed to meet internal goals. Since then, it has reached only a handful of data centers.

Microsoft CEO Satya Nadella told investors on the company’s fiscal Q4 2026 earnings call on July 29 that the chip runs 30% better on performance per dollar than existing hardware and is being scaled to support OpenAI and Microsoft’s own MAI models.

Back in January 2026, when Cryptopolitan reported the launch of the Maia 200, Microsoft described it as a second-generation part for inference based on TSMC’s 3-nanometer process.

Microsoft’s executive vice president for cloud and AI, Scott Guthrie, called it “the most efficient inference system that Microsoft has ever built.”

Microsoft wants to reduce reliance on Nvidia

Nadella has said openly that a goal is to cut reliance on Nvidia. Microsoft says its chips can run its own and OpenAI workloads at lower cost. It is expanding internal use through Azure AI Foundry and Copilot while selling the hardware to large outside cloud customers.

Earlier this month, Anthropic confirmed it is building its own semiconductor team to design custom chips for its Claude models. That makes it a potential buyer of the Maia 300 and a future rival in custom silicon chips.

Last year, in-house chips from Google, Amazon, Meta, and OpenAI were projected to make up 45% of the AI-chip market by 2028, compared with 37% in 2024.

Microsoft’s fiscal Q4 2026 revenues were $90.0 billion, up 18% year-over-year, with Azure and other cloud services up 43%.

TSMC’s N3 node and CoWoS advanced packaging will face supply constraints through 2027. That’s around the same time Microsoft wants to ramp up Maia 300 chip production.

If you're reading this, you’re already ahead. Stay there with our newsletter.