“Yu Xiaohui, the head of the China Academy of Information and Communications Technology (CAICT), highlighted the absence of a unified and diversified industry, along with the challenges posed by a shortage of computing power and low performance in usage,” as reported by Xinhua on Monday.
The professional landscape in China is under scrutiny. Various experts in Beijing have weighed in on the matter.
Recent data from the International Data Corporation (IDC) reveals a decline in China’s public computing investment growth rate, dropping from 66.55 percent in 2019 to 13.2 percent in the first half of 2023.
One solution offered by a technology company is the streamlining of costs for businesses, governments, and individuals by removing the necessity to procure or maintain computing equipment.
Contrastingly, data centers, often supported by private enterprises and local governments to furnish processing services, have witnessed an increase from 5.3 percent in 2020 to 20.7 percent in 2023, according to Gartner, a US-based consultancy.
In China, the preference for directly deployed private technology platforms by local governments or firms stems from concerns about data security. This stands in contrast to the US approach, which emphasizes public cloud services, as outlined in an article by Caijing Magazine in early March.
Shan Guangcun, an AI, robotics, and smart sensors specialist at Beihang University in Beijing, emphasized the necessity for the government to accelerate the establishment of public computing services nationwide. This includes nurturing more skilled professionals and providing financial backing to the industry.
Shan expressed concerns about the fragmentation of computing power, stating that it could escalate costs for companies training AI models, thereby impeding the advancement of China’s AI sector.
Beijing aims to increase China’s share of AI computing power from 25.4 percent in mid-2023 to 35 percent by the end of the following year.
Despite these ambitions, only 38 percent of data center capacity was utilized in 2022, significantly below the global average of 60 percent, according to the Chinese Academy of Engineering.
Acknowledging the substantial computing power gap, particularly in AI computing, the Chinese Academy of Engineering admitted the challenges faced by China.
The absence of a cohesive and open platform for sharing computing resources among local centers in China has hindered the smooth flow of vast amounts of data across regions, according to industry insiders.
Disorderly competition and disjointed local investments may further jeopardize the nation’s progress, as noted by Lu Ming, the director of Shanghai Jiao Tong University’s Institute for National Economy.
State-backed media outlet Securities Daily cautioned against the influx of speculative capital into the sector, highlighting challenges faced by companies venturing into AI, such as high initial investments and slow monetization of computing power.
Amid escalating tech restrictions, Washington’s limitations on China’s access to high-performance chips, crucial for AI and model training, have posed additional hurdles.
In response to these challenges, there is a growing urgency for China to develop an indigenous computing power system. Collaboration among AI industry leaders is deemed essential in designing such a platform, according to analysts.
Shan from Beihang University advocated for coordinated efforts among China’s tech giants to establish a domestic industrial ecosystem, emphasizing the government’s role in facilitating such collaborations.
A report by CAICT in July highlighted the significant losses incurred by Chinese computing-power providers due to the absence of unified computing standards, amounting to over 10 percent of their costs.
While the variety of AI chip types in China has surged over the past three years, the limited compatibility among the fewer than 100 computing-power companies poses a significant constraint on the application of computing power, according to the report.
