The annual meeting of the 2024 Zhongguancun (000931) Forum was held from April 25 to 29. With the theme of "Innovation: building a better World", hundreds of top experts and thousands of speakers talked about opening up and innovation, jointly seeking development opportunities, and contributing wisdom and strength to building a better world around scientific and technological innovation, leading social progress, improving people's livelihood, and sustainable development.
During the parallel forum "overseas Chinese Innovation and Development Forum" held on the 29th, Yang Fan, co-founder of Shangtang Technology and president of the large installation Business Group, was interviewed by a reporter from the Daily Economic News (hereinafter referred to as NBD).
NBD: in the context of the booming of big models, will there be a bubble in the intelligent computing center as a new infrastructure?
Yang Fan: the development of large models has promoted a large demand for computing power. As an infrastructure to provide this kind of computing power, the construction and development of the Intelligent Computing Center has received great attention. In 2023, the total scale of China's computing power will reach 1 per second.Castlecrash9.7 trillion floating-point operations, ranking second in the world, with an average annual growth rate of nearly 30% in the past five years.
At present, the intelligent computing center, as a key infrastructure carrying computing power in the intelligent era, has entered a high-speed construction cycle under the dual drive of policy and market, and the application prospect and enabling potential are gradually highlighted. It is estimated that the scale of intelligent computing power in China will be close to 2800 EFLOPS by 2028 (EFLOPS means 10 billion floating-point operations per second), showing the broad prospects of the intelligent computing center industry.
The investment in the construction of the Intelligent Computing Center is huge, and a number of large-scale projects are under way. Take Shangtang as an example, the computing power of SenseCore Shangtang device management has realized the unified scheduling of the national network, and new computing nodes have been expanded in Shanghai, Shenzhen, Guangzhou, Fuzhou, Jinan, Chongqing and other places, with a total computing power scale of 12000 petaFLOPS (1 petaFLOPS, that is, 1000 trillion operations per second), and more than 45000 GPU (graphics processors), of which Lingang AIDC (artificial Intelligence Computing Center) is 8400 petaFLOPS. It far exceeds the total calculation power of 3740 petaFLOPS of the project plan.
At present, the state has issued more than a dozen related policies to guide the development of the center. Among them, the "New data Center Development three-year Action Plan (2021-2023)" proposes to accelerate the deployment of high-performance, intelligent computing centers. The National Informatization Plan of the 14th five-year Plan emphasizes the overall construction of computing power and algorithm centers for block chains and artificial intelligence. The Digital economy Development Plan of the 14th five-year Plan proposes to promote the orderly development of intelligent computing centers and create a new intelligent infrastructure with the integration of intelligent computing power, general algorithms and development platforms.
Computing power is a new type of productivity in the era of digital economy. The new quality productivity has the characteristics of high-tech, high efficiency and high quality, with the substantial increase of total factor productivity as the core symbol. Computing power has become an important strategic resource of the country, which is closely related to the strategic emerging industries and future industries mentioned in the new quality productivity. To a certain extent, a country's numeracy level represents its scientific and technological innovation ability and international competitiveness.
NBD: users in various industries have different needs for computing power. How do you think we should lower the threshold for the use of computing power?
Yang Fan: at the level of computing power, the new AI infrastructure reconstructs the supply system of computing power. By providing large-scale, efficient and intensive computing infrastructure services, it can greatly improve the utilization ability of computing power in various industries, reduce the use cost of traditional self-building computing power, and improve the supply efficiency of computing power. In addition, through years of understanding and precipitation of AICastlecrashWe have developed different levels of software products and service systems to help you develop and use artificial intelligence models at a lower threshold and at a lower cost.
Daily economic news
