“We have not yet established a standard and comprehensive unified market. We also face the dilemma between lack of computing power and low utilization efficiency.” Yu Xiaohui, director of the Academy of Technology and Technology (CAICT), was quoted as saying by Xinhua News Agency on Monday.
Is China's industrial policy on the best path?Some advisers in Beijing have strong opinions.
Is China's industrial policy on the best path?Some advisers in Beijing have strong opinions.
According to data released by the International Data Corporation (IDC), the growth rate of China's public computing investment declined from 66.55% in 2019 to 13.2% in the first half of 2023.
Public computing services save businesses, governments, and individuals money by eliminating the need to purchase or maintain computing infrastructure.
Meanwhile, according to US-based consulting firm Gartner, the proportion of data centers that provide computing services primarily with support from private companies and local governments will increase from 5.3% in 2020 to 20.7% in 2023. increased.
According to an article in Caijing Magazine from early March, data security concerns in China have led local governments and businesses to turn to locally deployed private computing platforms, in contrast to the U.S. model of focusing on public cloud services. is often selected.
Xiang Guangkun, a professor specializing in AI, robotics and smart sensors at Beijing's Beihang University, said, “The government should improve the overall performance of the country, especially by getting adequate financial support for the industry and cultivating more relevant talent.'' We need to accelerate the creation of new public computing services.” he told the Post.
“The fragmentation of computing power increases the cost for companies to train AI models. This is not conducive to the development of China's AI industry,” Xiang said.
The Chinese government aims to raise China's share of AI computing power from 25.4% in mid-2023 to 35% by the end of next year.
Meanwhile, according to the Chinese Academy of Engineering, only 38% of data center capacity was used in 2022, far below the global level of 60%.
CAE acknowledged in June that China faces a “significant computing power gap,” particularly in AI computing.
China's private companies fed up with catastrophic payment delays and 'red tape excuses'
China's private companies fed up with catastrophic payment delays and 'red tape excuses'
The country lacks a unified, open platform for sharing computing resources and operations among regional computing centers, impeding the efficient flow of large amounts of data across regions. He said there was.
Industry officials say unregulated competition and uncoordinated local investment could also be hurting the country's catch-up efforts.
“Many local governments do not consider their relative competitiveness when making economic plans,” Lu Ming, director of the Institute of National Economics at Shanghai Jiao Tong University, told Shanghai-based publication The told the paper. “Some people lack the awareness to build a digital economic ecosystem.”
Last week, state media outlet Securities Daily warned of an influx of hot money into the AI field, as food producers and textile manufacturers are also entering the field.
“[These companies] “They have to face the pressure of large initial investments, while also taking into account the slow conversion of computing power into cash and the uncertain stage of development of downstream application areas.” “Investments in the computing industry should be made with careful consideration and within capacity planning.”
As tech regulations tighten, the U.S. government is restricting China's access to high-performance chips that are key to AI and model training.
“China has spent the past decade adapting to the computing ecosystem built by the U.S.'s Nvidia chips. And the chip ban has undone all of this effort. “It's like asking someone to suddenly switch to a domestic computer system. The two systems are not compatible,” said Li Yanwei, a technology consultant working in the smart computing industry in Shenzhen.
Therefore, there is increasing urgency for China to devise its own computing power system. Analysts say this will require AI industry leaders to work together to design such a platform.
China is spending billions of dollars on its national computing network.A data manager explains why
China is spending billions of dollars on its national computing network.A data manager explains why
Beihang University's Xiang called for joint action by China's big tech companies to form a domestic industrial ecosystem, and argued that the government should steer such cooperation.
“The best way [to form a national computing service] “China still allows cooperation from high-tech giants, and since these tech giants are also competitors in the same market, the Chinese government will think about how to proceed with such cooperation without harming the interests of the companies. “There is a need,” he said.
According to a report released by CAICT in July, more than 80% of computing power providers in China said losses due to the lack of unified computing standards accounted for more than 10% of their costs.
In the past three years, the variety of AI chips in China has increased more than five times, but there are fewer than 100 computing power companies that are compatible with the various chips, which greatly reduces the application of computing power. The report states that it is limited to .