Microsoft (MSFT.O) launched its second-generation AI chip, Maia 200, along with software tools aimed at competing with Nvidia’s developer ecosystem. The chip goes live this week at a data center in Iowa, with a second location planned in Arizona.
Competing with Nvidia
Cloud giants like Microsoft, Google, and Amazon are building their own AI chips, increasingly rivaling Nvidia. Google, in particular, is working with major Nvidia customers such as Meta Platforms to close software gaps between its chips and Nvidia’s offerings.
New Software Tools
Microsoft will also provide software to program Maia 200, including Triton, an open-source tool developed with OpenAI. Triton competes directly with Nvidia’s CUDA, widely seen as one of Nvidia’s strongest advantages.
Advanced Hardware Features
The Maia 200 is manufactured by Taiwan Semiconductor Manufacturing Co using 3-nanometer technology and high-bandwidth memory, though it’s slightly slower than Nvidia’s latest chips. Microsoft added SRAM memory, which can speed up AI workloads, especially for chatbots handling many users. Other companies like Cerebras Systems and Groq use similar memory designs for faster performance.
