On October 20th, at the Bund Summit in Shanghai, former governor of the People’s Bank of China, Zhou Xiaochuan, pointed out that once a massive investment in artificial intelligence partially fails, it may lead to financial turmoil or even crisis, and policies should be prepared in advance to handle it. On the same day, the National Data Group was established in the Xiong’an New Area to strengthen the integration of data resources and coordinated infrastructure.
Previously, on October 16th, Chinese leader Xi Jinping inspected Ulanqab in Inner Mongolia, affirming the prospects of the computing power industry while also cautioning against disorderly competition in the development of “new quality productivity,” expressing concerns about the phenomenon of “rising and falling in unison.”
Within just four days, three events have occurred successively, indicating the direct need for policy coordination. They all point to a common issue: China’s development in artificial intelligence is facing risks of excessive investment and redundant construction, with accumulating financial risks following suit.
According to data from the National Bureau of Statistics of the Communist Party of China, by the end of 2025, China’s intelligent computing power had reached 1.59 million PFLOPS; by the end of July 2026, it had increased to 2.45 million PFLOPS, representing a growth of approximately 54.1% in seven months.
In simple terms, computing power refers to the “brainpower” and “processing speed” of a computer. 1 PFLOPS = 1,000 trillion calculations per second, with PFLOPS akin to horsepower for a car, serving as a unit to measure the computational capacity of a computer.
By the end of 2025, the number of in-use computational facilities nationwide exceeded 13.73 million standard racks, with 42 Wanka intelligent computing clusters already established.
Ulanqab serves as a significant example of this computational construction. By August 2026, the region had signed 109 projects for computational centers, with a computing power scale of 190,000 P, where intelligent computing power accounted for over 90%.
The area boasts advantages such as green energy, cool climate, and proximity to the Beijing-Tianjin-Hebei market. However, the extensive project layout still needs to undergo scrutiny for market demand and investment returns.
Data centers are characterized by enormous investment amounts, rapid equipment depreciation, and high energy costs. If market demand falls short of expectations, some projects may face insufficient returns and asset depreciation.
The accelerated growth in computational power has contributed to the formation of an investment bubble. However, substantial simultaneous investment from local governments, state-owned enterprises, and financial institutions can lead to redundant construction and resource misallocation. Xi Jinping’s concern about “rising and falling in unison” reflects the contradiction between the AI strategy he has personally built and the market’s excessive investment.
While the government can determine the direction of industrial development, it might not fully control the investment impulses of local governments and the market.
On October 20, 2026, the National Data Group was established in the Xiong’an New Area, directly managed by the State-owned Assets Supervision and Administration Commission of the State Council of the Communist Party of China. Major contributors include central enterprises such as China Telecom, China Unicom, China Mobile, State Grid, and Southern Power Grid.
In recent years, China has focused on large-scale construction of AI computational power. Now there is a further enhancement with the establishment of a national-level platform to strengthen the coordination of computational power, data, networks, and power supply.
Communication networks, computational infrastructure, power supply, and data resource development were previously managed by different enterprise systems. Through the national-level platform’s strengthened coordination, there are conditions to improve cross-regional resource allocation, data circulation, and infrastructure operation efficiency.
The National Bureau of Statistics of the Communist Party of China mainly handles policy planning and coordination, while the National Data Group focuses on resource integration and market-oriented operation. With concentrated power, clear division of labor, enhanced supervision, efforts are made to prevent severe overinvestment, excessive construction, and control market risks.
Simultaneously, Zhou Xiaochuan warned of the potential consequences of “significant failures in artificial intelligence investments leading to financial turmoil.”
Zhou Xiaochuan categorizes AI investments into three types: large models, computational power, and data centers; AI applications; and supporting industries such as semiconductors, power, and liquid cooling.
He believes that the key to driving productivity enhancement lies in the application of artificial intelligence, rather than merely expanding computational facilities.
According to the “National Data Resource Survey Report (2025),” in 2025, China’s total data production volume reached 52.26ZB (1ZB = 1 trillion GB), with a 27.28% year-on-year increase. The total data storage volume reached 2.53ZB, with a 21.05% year-on-year increase.
The growth in data demand does not necessarily imply synchronous increases in computational needs. Techniques such as model distillation, lightweighting, and algorithm efficiency improvements may alter the future structure of computational demand, resulting in the premature devaluation of some existing equipment.
This implies that advancements in artificial intelligence technology do not guarantee that all related investments will be profitable.
Zhou Xiaochuan issues a warning: “Considering the massive flow of funds into artificial intelligence and related infrastructure, if a portion of it fails, it may trigger financial turmoil and lead to a crisis.”
The fundamental dilemma faced in China’s artificial intelligence development is the conflicting urgency of technological competition with its current economic structure.
Slowing down investments may result in missing out on the opportunities brought by a new technological revolution, yet continuing to increase investments to prevent falling behind may lead to overcapacity, resource waste, and financial risks.
This is the typical dilemma known as the “Strategic Prisoner’s Dilemma” – where everyone understands that cooperation may be better, but fearing that inaction will lose to the opponent, everyone invests heavily, resulting in everyone paying a considerable price in the end.
Xi’s concerns about unregulated development, Zhou Xiaochuan’s alerts regarding financial risks, and the establishment of the National Data Group, reflect China’s approach to this complex issue from the perspectives of industry investment, financial stability, and resource coordination.