The MetaX GPGPU provides mixed-precision computing acceleration for AI training. The high-capacity storage, new-generation high-speed IO interfaces, and multi-GPU interconnection technologies allow the efficient release of GPU computing power. Our self-owned MXMACA software stack provides comprehensive solutions for deep learning, parallel communication, virtualization, and cluster management to help customers quickly build large-scale AI training cluster environments.