Microsoft Unveils In-House AI Chip Maia 300, Key Chiplets Use TSMC 2nm

According to Central News Agency, citing Reuters, Microsoft announced the launch of its third-generation in-house AI chip, Maia 300, with plans to deploy more than 1 million units next year to support internal AI services. The chiplets responsible for key computing functions use TSMC's 2nm process. Maia 300 can handle code writing and execute complex business tasks on behalf of users. The previous generation chip has been used to run Microsoft's internal AI models, as well as OpenAI models offered through Microsoft's Azure cloud computing services. Microsoft has taken a different technical approach from Nvidia. Nvidia's AI accelerators use a single large computing chip surrounded by high-bandwidth memory chips; Maia 300's core computing processor is composed of chiplets, with the chiplets for key computing functions using TSMC's 2nm process, while other chiplets use the lower-cost 3nm process to handle data transfer out of the processor and to other locations within the system. Maia 300 will be installed in racks, with each rack housing 96 chips and consuming approximately 167 kW of power. In comparison, Nvidia's latest GB300 AI server rack is configured with 72 GPUs and consumes approximately 142 kW. Microsoft said it plans to begin large-scale deployment of Maia 300 starting next year, with the scale reaching several gigawatts. Microsoft previously stated that it expects capital expenditures to exceed $175 billion this year.