The computing town in Hangzhou, Zhejiang Province, with the integration of production, learning and research of computing research institutes and supercomputing enterprises as the starting point, will help promote the innovation, transformation and development of the city. The picture shows a view of China (Hangzhou) Suoli Town from the air. Long Wei has a floating point computing capacity of 10 billion to 100 billion times per second, and a petabyte (PB) of data storage scale... Many people may not have many concepts about these indicators, but when it comes to special effects, rendering and face recognition, intelligent customer service, voice translation and other applications in the Spring Festival movie, you should not be unfamiliar. Behind the visible application is the invisible intelligent computing power. Based on the latest AI theory and adopting the leading AI computing architecture, the Intelligent Computing Center (hereinafter referred to as "Intelligent Computing Center"), which provides computing power services, data services and algorithm services, plays an important role in promoting the development of AI and digital economy. Recently, the Intelligent Computing Center Innovation and Development Guide (hereinafter referred to as the Guide) released by the National Information Center and Inspur Information shows that more than 30 cities across the country are currently building or proposing to build intelligent computing centers, with the overall layout mainly in the eastern region and gradually expanding to the central and western regions. The innovative development of the intelligent computing center is expected to become a new engine driving the rapid development of AI and related industries. Not long ago, the Artificial Intelligence Supercomputing Center (Phase I) project located in Ningbo High-tech Zone, Zhejiang Province, was officially put into operation, providing 100P (i.e. PFLOPS, 1PFLOPS=10 billion floating point operations/second) semi-precision artificial intelligence computing power; In Tianjin, the first phase of the intelligent computing center project was completed at the end of last year, covering more than 850 enterprises and research institutes, and reaching an intention to cooperate with 120P of computing power space; Chengdu Intelligent Computing Center, officially launched in May last year, has recently signed a number of projects, focusing on smart medical, smart office and other application scenarios... Computational resources are an important base for the development of the digital economy. With the vigorous development of the digital economy, digital new things, new formats and new models promote the diversified development of application scenarios, the scale of computing power continues to expand, and the demand for computing power continues to rise. According to the data released by the Ministry of Industry and Information Technology a few days ago, in 2022, the total scale of data center racks in use nationwide will exceed 6.5 million standard racks; In the past five years, the average annual growth rate of the total scale of computing power has exceeded 25%. When computing power is applied in thousands of industries, different precision computing power needs to be "adapted" to various application scenarios. Especially with the rapid development of artificial intelligence technology, the computing power structure has also evolved, and the demand for intelligent computing power is increasing day by day. "The concept of general computing is different from intelligent computing, supercomputing and other special computing. It refers to the general computing power provided by the original CPU and GPU working together, in which the CPU undertakes the main computing work, and GPU is used for image processing, which has a wider range of applications. However, in the AI computing scenario, its computing efficiency is low, and the same number of floating point operations requires more computing resources." Alibaba Cloud technical experts said, Intelligent computing is based on heterogeneous computing resources. It is usually oriented to the needs of AI training and reasoning. Because of its specificity, it is oriented to AI