GuShiio.com ① The computing power structure has further evolved, and the scale of intelligent computing power has continued to expand;
② The speed of implementation applications is accelerating, and a large number of companies are actively “embracing” the big model;
③ Diversified demands put forward new requirements for the development of computing power.
In 2024, the market size will reach US$19 billion, a year-on-year increase of 86.9%
What signals will the growth of smart computing power exceed 70%(see highlights and readings for high-quality development)
core reading
With the development of large models and generative artificial intelligence, the scale of my country’s intelligent computing power has grown rapidly. Driven by a series of favorable measures, all parties have vigorously promoted the layout and construction of computing power resources, and the implementation and application speed of the intelligent computing power industry has been accelerating, reflecting the urgent needs of the market and the active layout and investment of China companies in the global artificial intelligence competition.
Special effects, rendering in movies, and generative artificial intelligence, Face Recognition, instant translation, etc. commonly used in life are all inseparable from the support of intelligent computing power. Currently, large models and generative artificial intelligence are constantly increasing the demand for computing power.
Recently, International Data Corporation (IDC) and Inspur Information jointly released the “2025 China Artificial Intelligence Computing Power Development Evaluation Report.” The report shows that the growth rate of my country’s smart computing power is higher than expected. In 2024, my country’s smart computing power will reach 72.53 trillion cycles per second (EFLOPS), a year-on-year increase of 74.1%; the market size will reach 19 billion US dollars, a year-on-year increase of 86.9%.
The computing power structure has further evolved, and the scale of intelligent computing power continues to expand
At the construction site of a large-scale intelligent computing center in Jinan, Shandong Province,”big” containers slowly rose and moved under the traction of a crane, and finally landed at a predetermined position. The intelligent computing center is like a “building block” that is quickly spliced and completed.
This Inspur Information Cell Computer “computing power factory”, which is made up of 119 prefabricated containers, took only 120 days from the beginning of construction to the official commissioning. Zhao Shuai, deputy general manager of Inspur Information, said: “We pre-integrate and pre-test the electromechanical equipment in the factory, load it into containers, then hoist and stack it, and finally connect it to the mains power, network and water source to run it independently.”
As an important foundation for the development of the digital economy, computing power is mainly divided into types such as supercomputing power, general computing power and intelligent computing power. Among them, intelligent computing power is specially designed for artificial intelligence applications and supports the training and operation of artificial intelligence algorithm models. At present, the demand for smart computing power in all walks of life is increasing day by day.
The “computing power factory” built in 120 days is a microcosm of the rapid growth of my country’s smart computing power. “In 2024, the growth rate of smart computing power in China will be more than three times that of general computing power in the same period.” Zhou Zhengang, vice president of IDC China, introduced.
In recent years, from the state to local governments to various business entities, the layout and construction of computing resources have been vigorously promoted. Last year’s “Government Work Report” proposed,”Build digital infrastructure appropriately ahead of schedule, accelerate the formation of a national integrated computing power system, and cultivate a computing power industry ecosystem.” The “Action Plan for the High-Quality Development of Computing Infrastructure” issued by six departments including the Ministry of Industry and Information Technology proposes to “gradually and reasonably increase the proportion of intelligent computing power.” Driven by a series of favorable measures, the scale of my country’s smart computing power has continued to grow, the service model has continued to improve, the software and hardware ecosystem has been gradually constructed, and a green and low-carbon development pattern has accelerated.
Zhou Zhengang said: “It is expected that in the next two years, the scale of China’s smart computing power will continue to maintain rapid growth, with 2025 increasing by 43% compared with 2024, and 2026 will double that of 2024.”
“By 2027, the proportion of inference computing power in the domestic artificial intelligence computing power architecture will increase significantly, which will drive further changes in the computing power structure. This will help promote the commercialization of large artificial intelligence models.” said Wei Kai, director of the Institute of Artificial Intelligence at China Academy of Information and Communications.
The speed of implementation applications is accelerating, and a large number of companies are actively “embracing” the big model
At present, the speed of implementation and application in the smart computing power industry is accelerating, and a large number of companies are actively “embracing” large models. A survey conducted by IDC shows that 42% of enterprises have begun to conduct preliminary testing and key concept verifications of large models, and 17% have introduced technology into production and applied it in actual business.
The manufacturing industry has complex scenarios, numerous categories, and huge and fragmented data, which makes large models face problems such as difficulty in implementing and high costs in the manufacturing industry. “For example, in the field of quality inspection, it is very difficult for us to obtain all defect samples.” Jia Jiaya, founder of Simou Technology, introduced.
“In the past, collecting various defect types required the factory’s production line to operate for two to three months; now, with the deep learning capabilities of the industrial multimodal large model we released, it generally only takes a week or less.” Jia Jiaya said that this large model has generated hundreds of defect types and is widely used in fields such as consumer electronics and new energy.
The growing demand for large models and generative artificial intelligence is significantly promoting the rapid development of my country’s artificial intelligence computing infrastructure, prompting enterprises to further support complex artificial intelligence computing tasks by applying high-performance computing equipment, optimizing data center infrastructure, and improving storage and network capabilities.
In addition to building their own intelligent computing infrastructure, many companies also achieve intelligent transformation by purchasing cloud computing services. You Peng, president of Huawei Cloud Big Data and Artificial Intelligence, said that Huawei Cloud Pangu Model has been implemented in more than 30 industries and more than 400 scenarios, covering manufacturing, pharmaceutical research and development, coal mining, steel and other industry fields.
“Currently, the penetration of artificial intelligence in the fields of Internet, finance, manufacturing, etc. continues to increase, which not only reflects the market’s urgent demand for advanced artificial intelligence technology, but also reflects the active layout and investment of China companies in the global artificial intelligence competition.” Zhou Zhengang said.
Diversified needs put forward new requirements for the development of computing power
According to the set environment, lighting, materials and rendering parameters, the model in the three-dimensional scene is presented on the two-dimensional screen in a realistic way… In film and television works, the detailed details and wonderful pictures are all inseparable from rendering. The development of intelligent computing power has provided computing power guarantee for it.
“We have provided rendering computing power support for more than 1900 special effects shots in” Nezha: The Devil Boy Raises in the Sea “.” Liu Tingjuan, project manager of the Technology Research and Development Department of Gui ‘an New District Science and Technology Innovation Industry Development Co., Ltd., said that strong computing power support can not only shorten the film production cycle, but also make the quality more secure.
The computing source originates from the Supercomputing Center in Gui ‘an New District, where rows of server cabinets stand neatly and operate at high speed. In one second, data can be calculated 300 billion times here, and it can be directly connected to the network with 38 cities including Beijing, Shanghai, and Guangzhou.
Not only in the field of film and television rendering, the current diverse needs of different industries and application scenarios have put forward new requirements for the development of computing power.
“The financial industry needs a high-security and low-latency computing environment, the medical industry needs to process a large amount of medical imaging data, the manufacturing industry needs to achieve real-time production optimization, and the Internet industry needs to process large-scale user data and content recommendations.” Liu Jun, senior vice president of Inspur Information, gave an example that these requirements put forward high performance, low latency, high security, scalability and cost-effectiveness requirements for the underlying computing power architecture.”Through resource pooling, intelligent scheduling, multi-tenant support, Heterogeneous computing and edge computing technologies, artificial intelligence computing infrastructure can flexibly respond to these needs, thereby promoting the widespread application of artificial intelligence technology in various industries.”
Industry experts said that after a certain period of accumulation of computing power scale, the key to the development of China’s computing power industry lies in how to achieve the leap from quantitative change to qualitative change. While continuing to expand the scale of computing power and achieve a balance between supply and demand, it is also necessary to solve difficulties in applying computing power in various industries, such as mismatch of storage, computing, and network resources, management and operation and maintenance problems of heterogeneous smart computing, and deep integration of computing power and real application scenarios to achieve the transition from expanding the scale of computing power to releasing the value of computing power with high quality.
