Microsoft is reportedly discussing with Taiwan Semiconductor Manufacturing Co. securing production capacity for more than 300,000 next-generation Maia AI chips for delivery in 2027, which would expand its in-house AI accelerator program.

Microsoft is also reportedly preparing to unveil Maia 300 this fall, after earlier rolling out Maia 200, its second-generation AI accelerator.

The reported reservation would use the same Taiwan Semiconductor Manufacturing Co. capacity used by Nvidia GPUs and AI chips designed by cloud providers.

Advanced packaging and high-bandwidth memory (HBM) remain supply constraints, although some analysts said chip-on-wafer-on-substrate (CoWoS) capacity had started easing by early 2026.

The expansion would still leave Microsoft behind the more mature chip programs at Google and Amazon.

A market analysis estimated that 70% of Azure AI workloads still ran on Nvidia hardware as of late 2025, suggesting Maia remains a supplement rather than a replacement.

🔗 Source: Reuters