Maia 100 is Microsoft's first custom AI accelerator designed for Azure, built to handle large-scale AI workloads. It maximizes performance, scalability, and flexibility through the collaborative optimization of hardware and software. Featuring TSMC N5 process technology and COWOS-S interconnect technology, Maia 100 provides bandwidth of up to 1.8TB/s and has a capacity of 64GB, supporting a thermal design power (TDP) of up to 700W, while operating efficiently at 500W. The accelerator integrates high-speed tensor units, vector processors, DMA engines, and hardware semaphores, supporting various data types and tensor partitioning schemes. Additionally, it features Ethernet interconnect for large-scale AI model support. The Maia SDK offers a rich set of components for rapid deployment of PyTorch and Triton models and ensures efficient data handling and synchronization through a dual programming model.