Key Highlights
- Maia 300 enters production in 2027, with 96 chips per rack, double the number for the Maia 200.
- A rack can run up to seven times more AI tokens, with 40% more tokens per megawatt than GPUs.
- Analysts at a brokerage estimate Maia represents about 10% of Microsoft’s AI chip spending by the first quarter of 2027.
- Unlike Google’s and Amazon’s chips, Microsoft’s accelerators cannot be rented by the hour by other companies.
From Nvidia Reliance to Homegrown Chips
Microsoft (NASDAQ:MSFT) is ramping up use of its own AI chips after years of reliance on Nvidia (NASDAQ:NVDA) graphics processing units. Starting with the third-generation Maia AI accelerator, unveiled on Monday, the company said it intends to rely mainly on its own silicon to handle incoming requests and to make Maia its primary architecture for running AI models.
Microsoft is targeting multi-gigawatt deployments of chips in the Maia family, according to a blog post from an executive vice president. A general manager declined to give a timeline. He said Microsoft reserves some optionality because the industry is changing fast and the company wants to deploy the right silicon for the right models at the right time.
What Maia 300 Offers
Production of the Maia 300 starts in 2027. Microsoft expects to pack 96 chips into each data center rack, double the number for the Maia 200. The new version can run up to seven times more AI tokens per rack, with 40% more tokens per megawatt when compared with GPUs.
The general manager said the industry needs new architectures that make better use of available power, since power translates directly into operating cost in a data center. A technical fellow at Microsoft named parts of Microsoft 365 Copilot among the workloads that gain speed on Maia 300.
A Chip Line With a Track Record
Microsoft introduced its first AI chip, the Maia 100, in 2023 and presented the Maia 200 in January. Chief executive Satya Nadella told analysts in July that the Maia 200 delivers 30% better performance per dollar than the latest hardware in Microsoft’s fleet and was already handling OpenAI models. The company’s finance chief said last month that using its own chips has helped margins over the course of a decade. For general-purpose work, Microsoft already uses its own central processing units alongside silicon from Intel (NASDAQ:INTC) and AMD (NASDAQ:AMD).
Why the Market Wants Choice
Nvidia’s GPUs have been the standard for AI workloads and fill Microsoft’s existing data centers. They also train models for OpenAI and Anthropic. As demand for advanced processors keeps growing, the leading model developers have embraced silicon from Amazon (NASDAQ:AMZN), Advanced Micro Devices, Cerebras and Google (NASDAQ:GOOGL) and, in recent months, Microsoft.
Nvidia stays part of the picture. Its chief executive, Jensen Huang, spoke last week at an event where Microsoft announced a Surface laptop with an Nvidia processor, and the general manager said Microsoft partners with Nvidia and AMD for model training. Analysts at a brokerage estimate that Maia accounts for about 10% of Microsoft’s AI chip spending by the first quarter of 2027.
Unlike Google’s tensor processing units or Amazon’s Inferentia and Trainium chips, Microsoft’s accelerators cannot be rented by the hour by other companies. One industry analyst noted that nobody outside Microsoft has tested them yet.
Silicon Markers to Follow
The points to track are the 2027 production start for Maia 300, Maia’s share of Microsoft’s AI chip spending, any timeline for the multi-gigawatt target, and whether Microsoft opens Maia to outside customers.






Please wait processing your request...