Microsoft Plans to Deploy Over 1 Million New Chips to Support Internal AI Services

The Maia 300 chip is Microsoft’s third-generation self-developed AI chip, designed for executing AI-intensive computing tasks such as programming and representing user execution of complex business tasks. Its previous generation product is currently used to run Microsoft’s own AI models and OpenAI models provided through Microsoft Azure cloud computing services.

Microsoft’s Maia 300 chip uses a different design approach than its competitor NVIDIA, which typically uses a single large calculation chip and a high-bandwidth memory chip. Microsoft’s chip is composed of multiple small chiplets, with the core calculation processor using TSMC’s 2nm process technology and other chiplets using a 3nm process technology for cost savings.