A large part of AI's global growth is achieved through cloud computing. When an AI application runs a deep learning model and continuously analyzes hundreds of billions of data, this process will require a lot of storage, data and computing, and AI will inevitably migrate to the cloud platform. The cloud provides a platform for deep learning of AI, which in turn improves the "intelligence" of cloud computing.

The leading cloud computing service provider provides an overall AI cloud solution for AI scenarios, as shown in the figure below. The features of the solution are as follows:

(1) It supports large-scale machine learning and deep learning applications. Users can implement data preprocessing, model training, application reasoning and other applications.

(2) Deploy a variety of typical deep learning frameworks and machine learning algorithm library environments, such as TensorFlow, Cafe-MPI, CNTK, mxnet, Torch, and so on, for the convenience of users.

(3) Deploy cloud and virtualization resource management platform, in-depth learning management AIStation platform and application analysis tool Teye to realize unified management, scheduling, monitoring and application feature analysis of resources.

(4) According to the requirements and computing characteristics of machine learning and deep learning application resources, configure different computing resources, such as data preprocessing CPU cluster, to achieve rapid preprocessing of training data; Provide high-density GPU clusters based on Nvidia P40, and realize rapid training for large-scale big data models; For the deep learning application of inference recognition, FPGA cluster is provided to realize high throughput and low latency processing. The entire cloud platform provides large-capacity, high-bandwidth, low-latency storage resources, high-speed 10-gigabit/IB network, and fast cloud services.