Microsoftmay unveil Maia 300, the next generation of its own AI chip, as early as September. According to The Information, as cited by Reuters, the company is in talks with TSMC to secure production capacity for over 300,000 chips, with delivery scheduled for 2027. Ultimately, the programme is expected to exceed one million units, although Microsoft has disputed the reported production figures.
This marks the next stage in Microsoft’s efforts to build its own hardware infrastructure. In January, the company launched Maia 200, manufactured using TSMC’s 3 nm process and designed primarily for inference – that is, the day-to-day operation of already trained AI models. Microsoft states that the chip offers a performance-to-cost ratio that is over 30 per cent better than the company’s most recent hardware.
The Maia 300 is part of a broader shift in the market. In the second quarter of 2026, Google began recognising revenue from direct sales of its TPU chips, whilst Amazon is developing the Trainium family and reporting growing interest in its own AI processors.
For Microsoft, therefore, the stakes go beyond simply competing with Nvidia. Its own chips could reduce the cost of running Azure, increase the availability of computing power and limit its reliance on a single supplier. However, this does not signal a swift end to the dominance of Nvidia’s GPUs. Microsoft continues to develop an infrastructure based equally on its own chips and on processors from Nvidia and AMD.
The key test for the Maia 300 will not be performance alone, but the ability to mass-produce the chips and convince major cloud customers to adopt the new architecture.
