In order to solve the problem that embedded devices are unable to run artificial intelligence and other computationally intensive tasks due to the poor performance, while high-performance computing devices can not be deployed on the application site because of the limitation of volume and power consumption, this paper presents a hybrid architecture with embedded and cloud computing storage. The system based on a robust network communication architecture can timely transmit data collected by embedded devices to the cloud, and can properly dispatch cloud computing node cluster for high-speed computing. After the actual test, the system fully meets the enormous requirements of computing resources for the embedded device to run artificial intelligence, greatly reducing the computational load of the embedded terminal. This hybrid architecture provides a large amount of computing resources for embedded devices, avoids resource waste and maintenance difficulties caused by the local deployment of computing resources, as well as effectively reduces the overall capital investment of the device