Microsoft is developing a new line of processors intended for artificial intelligence. According to Reuters, the company plans to launch the Maia 300 in the second half of the year, possibly as early as September, aiming to increase its internal production capacity and reduce dependence on processors supplied by Nvidia.
This strategy also covers the cloud computing sector. Microsoft is negotiating with TSMC to manufacture over 300,000 units of this new chip, with delivery expected in 2027, and plans to considerably increase this volume.
Microsoft's entry into the AI chip segment occurred in November 2023 with the presentation of the first Maia. Since then, competitors such as Google and Amazon have advanced in expanding their own solutions.
This move aims to mitigate dependence on Nvidia processors while internally developed corporate chips gain relevance in the cloud computing market. Google has already started generating revenue from the direct sale of its artificial intelligence chips, known as Tensor Processing Units (TPUs), while Amazon demonstrates growing adoption of its processors, including Trainium.
More specifically, Reuters reports that Microsoft is negotiating with TSMC the production of over 300,000 units of the Maia 300, scheduled for delivery in 2027. Furthermore, the company intends to persuade major cloud computing clients, such as Anthropic, to use this chip.
The long-term goal established is to secure capacity to exceed 1 million units. However, this plan may be restricted by component availability and capacity agreements made with TSMC.
Among the central aspects of this strategy are: the Maia 300 will be the successor to the Maia 200, the second generation introduced by Microsoft in January. This component was produced by TSMC using 3-nanometer technology. Additionally, the chip was equipped with a substantial amount of SRAM, a type of memory that can confer speed benefits to AI systems that need to process a large volume of user requests.
