Core Insights - The National Data Bureau has reported significant progress in the construction of national data infrastructure over the past year, aiming to create a new infrastructure system focused on releasing the value of data elements, supported by networks and computing power [1] Group 1: Infrastructure Development - A guideline for national data infrastructure construction has been released, establishing a "three-step" approach to advance the initiative [1] - The optimization of computing resource layout shows that the operational costs of data centers in the western region are approximately 50%-70% lower than those in the eastern region, indicating a substantial cost advantage [1] - Eight national hub nodes have been established with a total intelligent computing scale of 620,000 PFLOPS, accounting for about 80% of the national total [1] Group 2: Collaborative Framework - The National Data Bureau has organized trials in 18 cities, including Beijing and Shanghai, focusing on six technical routes to promote collaborative data infrastructure [1] - A three-tier technical architecture has been constructed to facilitate coordinated data infrastructure across regions, with seven regional functional nodes and 25 city business nodes established, covering 16 provinces [1] - By the end of this year, the number of city business nodes is expected to expand to around 50, covering 80% of provinces [1] Group 3: Computing Resource Monitoring - An integrated computing network experimental verification platform has been established, creating a "dynamic map" of national computing resources based on 63 indicators [2] - The platform enables monitoring and scheduling of computing resources across eight hubs, ten clusters, and several non-hub provinces, covering 901 computing facilities [2] - It can schedule 40,000 PFLOPS of multi-source computing power across different entities and regions, offering over 40 products and services related to general computing, intelligent computing, and supercomputing [2]

Core Viewpoint - The article highlights significant developments in the computing power sector, emphasizing the government's push for high-quality development and the need to address the current supply-demand imbalance in computing resources [1][4][5]. Group 1: Government Initiatives - The State Council recently released a top-level document on "Artificial Intelligence+", focusing on strengthening the overall coordination of intelligent computing power [1][6]. - During the 2025 Digital Expo, the National Data Bureau launched a nationwide integrated computing power network monitoring and scheduling platform, aiming to enhance resource allocation and efficiency [6][7]. - A joint initiative was proposed to strengthen the collaborative development of computing power hubs, emphasizing the need for unified technical standards and regional cooperation [7]. Group 2: Industry Development - As of June 2023, China has 10.85 million operational computing centers, with intelligent computing power reaching 788 EFLOPS [4]. - The computing power industry is experiencing a structural mismatch in supply and demand, with industry insiders expressing concerns about the quality of current resources [4][8]. - The government is signaling the importance of improving the quality of computing power supply, with a focus on avoiding disorderly competition and ensuring sustainable development [5][6]. Group 3: Challenges in Computing Power - The industry faces challenges such as insufficient core supply, communication bottlenecks, and complex scheduling of computing resources [8]. - By 2025, the demand for intelligent computing in China is expected to reach 300E, but the supply distribution remains uneven [8][14]. - The mismatch in computing resources is particularly evident in the government cloud sector, where various factors contribute to the imbalance [9][12]. Group 4: Perspectives on Resource Allocation - Some industry experts argue that the current supply-demand imbalance is a normal market adjustment and not necessarily a true mismatch [14][15]. - The lack of effective commercial closure for many large model enterprises contributes to underutilization of computing resources [15]. - There is a consensus among industry players on the need for high-quality development and various proposed solutions to optimize resource utilization [16][17].