“We have yet to set up a unified market that is standard and inclusive. And we are also facing a dilemma between computing power shortage and low efficiency in usage,” Yu Xiaohui, head of the China Academy of Information and Communications Technology (CAICT), was quoted as saying by Xinhua on Monday.

Is China’s industrial policy on best path? Some Beijing advisers offer hot takes

According to data released by International Data Corporation (IDC), the growth rate of China’s public computing investment declined from 66.55 per cent in 2019 to 13.2 per cent in the first half of 2023.

A public computing service enables cost savings for businesses, governments, and individuals by eliminating the need to purchase or maintain computing infrastructure.

Meanwhile, data centres, which are mainly backed by private companies and local governments to provide computing services, saw an increase from 5.3 per cent in 2020 to 20.7 per cent in 2023, according to Gartner, a US-based consulting company.

In China, local governments or firms often opt for locally deployed private computing platforms due to concerns about data security, in contrast with the US model to focus on public cloud services, according to an article by Caijing Magazine in early March.

“The government must push faster to build up public computing services across the nation, especially by cultivating more related talents with adequate financial support for the industry,” Shan Guangcun, a professor specialising in AI, robotics and smart sensors at Beihang University in Beijing, told the Post.

“The fragmentation of computing power will increase the cost for companies to train AI models, this is not conducive to the development of China’s AI industry,” Shan said.

Beijing is looking to raise China’s share of AI computing power to 35 per cent by the end of next year from 25.4 per cent in mid-2023.

On the other hand, only 38 per cent of the capacity of data centres was used in 2022, far below the global level of 60 per cent, according to the Chinese Academy of Engineering.

CAE conceded in June that China was being challenged by “an extremely large computing-power gap”, especially in AI computing.

China’s private firms fed up with crippling payment delays, ‘red-tape excuses’

It said the country was lacking a unified and open platform for sharing computing resources and operability among local computing centres, hindering the efficient flow of massive quantities of data across regions.

Industry insiders have also said disorderly competition and uncoordinated local investment may also be compromising the nation’s catch-up efforts.

“Many local governments have not considered their relative competitiveness when doing their economic planning,” Lu Ming, director of Shanghai Jiao Tong University’s Institute for National Economy, told a Shanghai-based publication, The Paper, earlier this month. “Some lack the awareness for building a digital economy ecosystem.”

State-backed media Securities Daily last week warned of the inflow of hot money into the sector, as food producers and textile makers are also marching into the AI sector.

03:48

Young Chinese singles turn to AI-generated partners

Young Chinese singles turn to AI-generated partners

“[These companies] have to face the pressure of hefty initial investments, while also considering the slow conversion of computing power into cash and the uncertain developmental stage of downstream application areas,” it said. “They need to invest in the computing industry within their capacities planning, after careful consideration.”

Amid heightened tech curbs, Washington has restricted China’s access to high-performance chips, which are key for AI and model training.

“China has adapted to the computing ecosystem created by the US’ Nvidia chips over the past decade. And with the chip ban, all of this effort is shattered. It is just like asking long-term Windows users to suddenly switch to a domestically produced computer system – the two systems are not compatible,” Li Yangwei, a technical consultant working in the smart computing industry in Shenzhen.

It thus raises the urgency for China to come up with an indigenous computing power system. To that end, it needs its AI industry leaders to work together on designing such a platform, according to analysts.

China is spending billions on a national computing network. Its data chief says why

Shan of Beihang University called for joint actions from China’s leading tech firms to form a domestic industrial ecosystem, and the government should navigate such cooperation.

“The best way [to form a national computing service] is still letting tech giants cooperate, and Beijing needs to figure out how to navigate such collaboration without hurting companies’ interests, as these tech giants are also competitors in the same market,” he said.

According to the CAICT’s report published in July, more than 80 per cent of Chinese computing-power providers said that losses due to the lack of unified computing standards have accounted for more than 10 per cent of their cost.

In the past three years, the AI chip types in China have increased by more than five times, yet there are fewer than 100 computing-power companies that are compatible with different chips, which has severely constrained the application of computing power, the report said.