Microsoft launches Maia 300 for Azure cloud AI workloads

Microsoft introduced the Microsoft Azure Maia 300, its largest in-house compute processor for Azure Cloud deployments. CEO Satya Nadella said Microsoft plans to scale the Maia chip family to multi-gigawatt deployments and make it the company’s primary AI inferencing system for its own products and services.

Microsoft is deploying the Maia 200 across the US, Europe, and Australia regions, where it is delivering 70% more tokens per Watt across Copilot and Teams than AI GPUs. Maia 300 is expected to provide a further 40% uplift in tokens per Watt over GPU-based solutions.

The Microsoft Azure Maia 300 uses a multi-chiplet tile-based design with a main compute die built on TSMC N2P (2 nm), alongside two I/O dies on an older node and four DDR RDIMM memory tiles for system-level main memory. The accelerator also includes 288 GB of 9 GT/s HBM4 memory across eight stacks, offering 18 TB/s of bandwidth, while the full SoC has a maximum TDP of 950 W.