The demand for AI computing power has increased sharply, and Shanghai Lingang will build a tens-billion-scale computing power industry

(Editor/Lu Dong)

The integrated innovation of large models and large computing power is causing a major shift in the production paradigm, pushing scientific research and industrial applications towards the era of general artificial intelligence (AGI) driven by intelligent computing. In the early stages of rapid technology iteration, how to build a new generation of infrastructure, lower the application threshold, shorten the research and development cycle, and improve innovation efficiency has become a new challenge that the government and upstream and downstream industries need to jointly solve.

"At present, when artificial intelligence companies come to Lingang, they do not simply focus on support policies such as subsidies, but ask whether they can solve the computing power needs." On June 2, at the Shanghai Lingang New Area Intelligent Computing Conference, Lu Yu, director of the High-Tech Department of the Lingang New Area Management Committee, said.

On the same day, the "Action Plan to Accelerate the Construction of Computing Industry Ecosystem in Lingang New Area" (hereinafter referred to as the "Plan") was officially released. The goal is that by 2025, Lingang New Area will form a system that focuses on intelligent computing and basic computing power. A multi-computing power supply system coordinated with supercomputing power, the total computing power exceeds 5EFLOPS (FP32), AI computing power accounts for 80%, and the overall scale of the computing power industry exceeds 10 billion yuan.

Conference scene Source: Observer Network

Observer Network learned from the conference that as of now, the Lingang New Area has 8 data centers including SenseTime Artificial Intelligence Computing Center (AIDC), Youfu Networks, and Information Feiyu, with a total computing power of more than 3EFLOPS (FP32). Intelligent computing power accounts for nearly 80%, and the total computing power accounts for nearly 20% of Shanghai. At the same time, the computing power industry in the new area has made corresponding arrangements in the fields of upstream software and hardware, midstream data centers, dispatching platforms, and downstream applications.

Xu Li said at the meeting that computing power is the energy source of the new era, and competitiveness depends on computing power to some extent. In the era of large AI models, the amount of model parameters will increase at an exponential rate, and the amount of data will also increase massively with the introduction of multi-modality, which will lead to a sharp increase in the demand for computing power.

"In the past, we talked about Moore's Law, and the integration density of transistors doubled every 18 months. However, in the past 5-6 years of artificial intelligence, the consumption of computing power by the best top models in the industry has doubled every 4.6 months. Double." Yang Fan, co-founder of SenseTime and president of the large device business group, said.

He said that GPT-3 used 10,000 Nvidia V100 GPUs to build a complete model, which consumed about 600P of half-precision computing power (FP16). This computing power scale is very large, because in many places in China today, the smart computing power scale of the entire city may only be 100-300P.

The "Action Plan for Accelerating the Construction of Computing Industry Ecosystem in Lingang New Area" mentioned that in order to accelerate the construction of computing industry ecology and achieve the 2025 goal, Lingang New Area will next improve its computing power supply capacity and improve the operation of computing network capabilities to improve computing power innovation capabilities; optimize the allocation of computing power resources, build a standardized and efficient trading system, and promote the open sharing of computing power resources; consolidate the computing power support base, develop basic computing power software, accelerate algorithm development and application, and encourage the orderly flow of data. Build an industrial standard system; accelerate the implementation of multi-industry and multi-scenario applications.

At this conference, the Lingang New Area Intelligent Computing Industry Alliance was officially established. Alliance members represent enterprises providing computing power such as intelligent computing power, basic computing power, and supercomputing power centers, including GPU, FPGA, ASIC, etc. Computing power chip companies, as well as large model, AI for science and other computing power demand companies, a total of 25 companies, as well as a total of 3 universities and research institutes including China Academy of Information and Communications Technology East China Branch, Xi'an University of Electronic Science and Technology, and University of Electronic Science and Technology of China.

Observer Network learned that computing power providers include SenseTime Technology, China Telecom, China Unicom, China Mobile, Youfu Networks, etc., and computing power chip companies include Muxi Integrated Circuit, Suiyuan Technology, Tianshu Zhixin, Hanbo Semiconductor, Kunlun Core, etc., and computing power demanders include ByteDance, Alibaba Cloud, etc.

Lingang New Area Intelligent Computing Industry Alliance was established

At the meeting, SenseTime was awarded the title of “New Area Intelligent Computing Industry Chain Leader” enterprise. Xu Li, Chairman and CEO of SenseTime, delivered a keynote speech at the conference. Focusing on the "value of computing power", he shared how SenseTime's large devices have been used in many fields such as weather prediction, autonomous driving, remote sensing, games, and large language models in the past six months. A series of achievements achieved.

Xu Li revealed that as of May this year, SenseTime has served more than 40 core customers, including more than 10 large model customers, covering fields such as intelligent driving, biopharmaceuticals, chip design, smart business, and university scientific research. , and has achieved large model delivery in more than 20 landing scenarios.

Also as a provider of computing power, Zang Yunfeng, chief technology officer of Youfu Network, revealed that Youfu Lingang Cloud Computing Data Center is one of the five major computing power centers in Shanghai. Relying on Shanghai Lingang Cloud Computing Data Center, Youfu has expanded Investing in the construction of large-scale intelligent computing capability base and computing power cluster, it is expected to form E-level computing power base and computing power cluster capabilities by the end of 2023.

At the same time, at this conference, China Telecom Shanghai Company and the Lingang Management Committee announced the launch of the construction of China Telecom Lingang Public Intelligent Computing Service Platform and Domestic GPU Joint Innovation Base. The base will promote the commercialization and industrialization of domestic computing power. The first batch of domestic GPU partners to settle in the base include Huawei, Suiyuan Technology, Tianshu Zhixin, Cambrian, Muxi Integrated Circuit, etc.

The construction of domestic GPU joint innovation base started

Recently, as AI large models continue to be popular, Shanghai, Beijing, Shenzhen and other places have successively issued policy documents supporting the artificial intelligence industry, proposing new plans for the supply and layout of computing power. For example, on May 30, "Several Policies and Measures for Shanghai to Increase Efforts to Support the Development of Private Investment" proposed to support the extensive participation of private enterprises in the construction of artificial intelligence infrastructure such as data and computing power.

The "Guiding Opinions on Promoting Unified Scheduling of Computing Power Resources in Shanghai" issued recently also proposes to carry out a survey of the city's computing power infrastructure and computing power resource output capabilities to form a computing power list. Based on the base number of computing resources, promote leading enterprises to access the city's artificial intelligence public computing service platform, build an integrated computing power dispatching service system and platform basic framework, and achieve unified orchestration of computing resources.

Tang Wenkan, deputy director of Shanghai Economic and Information Technology Commission, pointed out in his speech at this conference that during the "14th Five-Year Plan" period, Shanghai will strengthen the coordination, dispatch and sharing of computing resources in the city, improve the utilization rate of computing resources, and accelerate the circulation of data elements. , fully releasing the value of data, we hope that the Lingang New Area will take the establishment of the Intelligent Computing Industry Alliance as an opportunity to explore the collaboration of all factors upstream and downstream of the industrial chain to form a new hot spot in the digital economy.

The above is the detailed content of The demand for AI computing power has increased sharply, and Shanghai Lingang will build a tens-billion-scale computing power industry. For more information, please follow other related articles on the PHP Chinese website!