Unsurprisingly, AI has emerged as the hottest industrial topic at this year’s Two Sessions. Deputies have offered a wide range of insightful views and forward-looking proposals from diverse perspectives. Nevertheless, regardless of how AI evolves, its development remains heavily reliant on computing power—precisely one of the key bottlenecks currently facing China’s AI industry.

Consequently, the development of computing power has naturally become a focal topic of intense discussion among deputies and committee members. From the construction of intelligent computing centers to breakthroughs in domestic substitution, from green development to industrial applications, a series of proposals and recommendations are outlining the future blueprint for China’s computing power industry.

01

A Dual Leap from Quantity to Quality

During the Spring Festival, “DeepSeek’s servers are busy” has become a famous internet meme. Huang Qunhui, Member of the National Committee of the Chinese People’s Political Consultative Conference It points out that large models such as DeepSeek have enormous computational power requirements, yet they face the contradictory realities of insufficient computing capacity and network congestion. Accordingly, it calls for accelerating the nationwide deployment of an integrated computing power network to break the bottleneck of computing power shortages.

Zhang Yunquan, Member of the National Committee of the Chinese People’s Political Consultative Conference It was also pointed out that, although models such as DeepSeek achieve remarkable efficiency gains through “achieving great results with modest resources,” general-purpose large models still adhere to the principle that model parameter scale expansion is positively correlated with computing power demand. At present, China’s domestic intelligent computing centers still lag behind international cutting-edge levels in terms of computing density and efficiency. To address this, he recommended establishing a national-level ultra-large-scale computing cluster, promoting the construction of super-intelligent integrated computing centers based on domestically produced chips, and enhancing overall computing resource utilization.

Cao Peng, a member of the National Committee of the Chinese People’s Political Consultative Conference and Chairman of the JD.com Group Technical Committee Accordingly, a proposal was submitted on “advancing the development of heterogeneous computing infrastructure,” highlighting the urgent need to upgrade such infrastructure. He pointed out that the high computational costs and escalating technical complexity associated with large-model applications necessitate the promotion of AI’s widespread adoption through independent R&D and resource-sharing mechanisms. This represents not only a “hardcore” upgrade of the underlying infrastructure but also a profound reevaluation of the efficiency of the computing-power industry.

In the global AI race, achieving independent and controllable computing infrastructure has become an unavoidable topic. Liu Qingfeng, a deputy to the National People’s Congress and Chairman of iFlytek At the Two Sessions, a proposal was put forward to accelerate the development of domestically produced, independently controllable large-scale models and their industrial ecosystems on indigenous computing platforms, thereby reducing reliance on foreign technologies.

02

Green Development and Industrial Empowerment

The high energy consumption of the computing power industry has long been a major pain point. Yong Shiping, a member of the Chengdu Municipal Committee of the Chinese People’s Political Consultative Conference We propose an integrated development model of “green power + computing power” and recommend establishing a cross-regional direct-link mechanism for “green-power enclaves” to deeply integrate Chengdu’s computing centers with clean energy. By implementing coordinated policies that align computing and power, we can enhance the sustainable development capacity of the computing industry and promote its long-term, stable growth.

Mao Yaping, Member of the Jiaxing Municipal Committee of the Chinese People’s Political Consultative Conference She proposed the “Computing Power Plus” and “Data Times” strategies to promote the digital–physical integration of the manufacturing sector. She recommended coordinating computing–power synergy, developing data application scenarios, and innovating mechanisms for integrating education, science, and industry. She also called for strengthening industry planning as a guiding force, introducing data-industry policies, and formulating a five-year plan for the data industry along with detailed policy guidelines.

Currently, large-scale models have been widely adopted in sectors such as healthcare and finance and are receiving strong support from both the government and enterprises. The deep integration of computing power with industry has become a new strategic lever for China’s high-quality economic development.

According to incomplete statistics from IDC Circle, as of 2025, China has commissioned a total of 150 intelligent computing center projects (for details, see: Core Internal Publication of the Computing Power Industry | Layout, Leading Institutions, and Domestic Chip Architecture Analysis of 150 Commissioned Intelligent Computing Center Projects in 2025 ), there are also some voices within the industry suggesting that intelligent computing centers are oversupplied. However, as evidenced by the proposals put forward by various committee members and delegates, China’s computing power industry is currently in a state of supply falling short of demand, with the primary challenges being the scarcity of high-quality computing resources and the mismatch between computing capacity supply and demand.

China has long been known as the “infrastructure superpower,” and now the world is entering the age of artificial intelligence. Accordingly, China must also usher in an era of “computing power supremacy.” Only by truly building a nationwide intelligent computing network—making computing power as readily accessible and ubiquitous as water and electricity—can AI be genuinely made available to every citizen and benefit society as a whole.