The company is expected to unveil the Maia 300 AI chip this fall, potentially as soon as September, according to The Information, which cited people with direct knowledge of the plans.
The planned launch comes as Microsoft steps up efforts to develop its own AI hardware and reduce its dependence on expensive processors supplied by Nvidia, which currently dominates the market for chips used to train and run advanced AI models.
Microsoft first entered the custom AI chip race with the launch of its Maia AI processor in November 2023. However, its efforts have so far progressed more slowly than those of some major technology rivals, including Alphabet and Amazon, both of which have been expanding the use and production of their internally developed AI processors.
Google, for instance, began recognizing revenue from direct sales of its custom AI chips, known as Tensor Processing Units (TPUs), during the quarter that ended in June. Amazon, meanwhile, has reported growing adoption of its own AI processors, including its Trainium chips.
Microsoft is also seeking to secure substantial manufacturing capacity for the Maia 300. According to The Information, the company has been negotiating with Taiwan Semiconductor Manufacturing Company (TSMC) for capacity to produce more than 300,000 units of the chip for delivery in 2027.
The software giant is reportedly targeting an even larger production scale in the longer term, with an ambition to secure capacity for more than 1 million Maia 300 chips. However, the availability of components and continuing negotiations over manufacturing capacity could affect the scale and timing of the programme.
Beyond increasing production, Microsoft is also looking to persuade major cloud-computing customers to adopt its chips. Anthropic, the artificial intelligence company behind the Claude family of AI models, is among the major cloud customers Microsoft is reportedly hoping to attract.
"Microsoft continues to invest in custom silicon as part of our long-term AI infrastructure strategy. While we don't share production volumes, the figures reported don't reflect the scale of our program," said Andrew Wall, general manager for Microsoft's Azure Maia.
TSMC could not be reached for comment outside regular business hours.
Microsoft unveiled its second-generation Maia 200 in January, with the processor manufactured by TSMC using its 3-nanometer technology. The company designed the chip with a substantial amount of SRAM, a type of high-speed memory that can offer performance benefits for AI systems handling large volumes of requests simultaneously.
The development of the Maia 300 underscores the growing importance of custom silicon as major technology companies seek greater control over the hardware powering their AI services. Developing proprietary chips could allow companies such as Microsoft to improve performance for specific workloads while potentially reducing costs and dependence on third-party suppliers.
For Microsoft, the push is particularly significant as demand for AI computing continues to grow across its cloud business. The success of Maia 300 will therefore depend not only on the chip's technical capabilities but also on Microsoft's ability to manufacture it at scale and convince customers and developers to use it.