The Formation of a National "Single Network" for Computational Power: Sichuan Hub Goes Online, Ushering in an Era of "On-Demand" Computing

At the 2025 Computing Power Internet Conference held in Chengdu, Sichuan, a nationwide computing power scheduling system was officially launched, marking a new phase for China's computing power development characterized by accessibility anytime, anywhere, and on-demand.

The "1 + M + N" architecture—comprising "1 National Computing Power Internet Service Platform, M regional computing power interconnection platforms, and N industry-specific computing power interconnection platforms"—was announced as having completed initial verification at the conference on December 24, 2025.

During the conference, the cross-domain system of the National Computing Power Internet Service Platform was officially released, alongside the launch of the Sichuan Hub Service Node of the National Computing Power Internet.

01 Architecture Implementation

The "1 + M + N" architecture of the National Computing Power Internet has transitioned from blueprint to reality.

The National Computing Power Internet Service Platform serves as the core, responsible for the comprehensive management of the computing power interconnection platform system. Multiple regional computing power interconnection platforms, covering key computing power clusters in Beijing, Shanghai, Sichuan, Hubei, and Guangdong, are conducting verification work for computing power interconnection.

Industry giants such as China Telecom, China Mobile, and China Unicom have leveraged such platforms to initiate the construction of experimental computing power internet networks, building a "single network" for computing power that covers the entire domain and operates across layers.

According to statistics, the experimental computing power internet system had cumulatively scheduled 20 million card-hours of computing power by 2025.

02 Sichuan's Breakthrough

The Sichuan Hub Service Node, acting as the computing power scheduling hub for Southwest China, achieved three landmark breakthroughs.

The Sichuan node took the lead in integrating computing resources from multiple companies including Sichuan Telecom, Sichuan Mobile, Sichuan Unicom, and Sichuan Shutian Information, addressing industry challenges such as resource fragmentation and unified allocation.

The platform launched the country's first agent-based computing product service "supermarket." Through an intelligent service system featuring unified encapsulation and elastic allocation, it promotes a shift for both supply and demand sides from the traditional infrastructure leasing model to a service-oriented computing power supply model.

The platform also achieved minute-level monitoring capabilities for key indicators such as computing resource scale, location, and availability, supporting Sichuan Province in accelerating the process of "connecting computing into a network."

03 Global Context

Globally, AI-driven demand for computing power is growing exponentially, and competition in computing power has evolved from individual technology contests to ecosystem synergy games.

Wu Hequan, an academician of the Chinese Academy of Engineering, pointed out that countries are actively building integrated computing power ecosystem systems through top-level design. Even major international tech giants have faced severe "computing power shortages" this year, establishing dedicated executive committees to coordinate computing resource allocation.

04 Market Scale

China's development in the computing power field has entered a critical stage of transition from large-scale deployment to integrated interconnection.

As of December 2025, the user base of China's Computing Power Internet at the consumer level was close to tens of millions. The utilization of public intelligent computing services reached 1.407 billion card-hours, and the service market scale for intelligent computing power exceeded 130 billion RMB.

In terms of demand structure, it is estimated that by 2026, China's demand for intelligent computing will account for 71% of new demand. The explosion of generative AI has made building smarter, greener, and more secure data centers an imperative for the upgrading of the computing power industry.

05 Technical Challenges

The realization of a computing power internet faces multiple technical challenges, particularly regarding heterogeneous computing power coordination.

With the surging domestic demand for computing power, factors such as tight supply of high-end computing power, diversity of domestic chips, and mixed expansion construction of intelligent computing centers have led to the formation of numerous heterogeneous computing power silos.

These heterogeneous computing resources include Graphics Processing Units (GPUs), Neural Processing Units (NPUs), and Field-Programmable Gate Arrays (FPGAs). Difficulties in management and low utilization rates have become common industry pain points.

Domestic manufacturers have begun exploring solutions, achieving the implementation of China's first hybrid training technology for four or more types of heterogeneous chips, and gradually progressing towards 10,000-card heterogeneous clusters.

06 Green Transition

The green transition of the computing power industry has become an irreversible trend.

The high-density nature of intelligent computing presents significant power challenges, making liquid cooling a necessity for high-density cabinets. Green computing power is shifting from a cost factor to a core competitive edge and a market access element, with new projects commonly requiring a proportion of green electricity.

The industry-proposed concept integrating Reliability, Agility, and Sustainability (RAS) is becoming an important guiding framework for data center construction in the era of intelligent computing. This systemic restructuring aims to make every watt of electricity carry more effective computing power.

07 Future Outlook

According to the plan, by 2026, China will establish a relatively comprehensive system of standards, identifiers, and rules for computing power interconnection.

By 2028, China aims to basically achieve standardized interconnection of national public computing power, gradually forming a computing power internet with capabilities for intelligent perception, real-time discovery, and on-demand access.

With the opening of high-speed regional data channels, data connectivity between Hong Kong and Guangzhou has been strengthened. These regional breakthroughs, combined with the national network, will progressively build a unified, open, and competitively ordered national computing power service market.

In the industrial sector, software companies have successfully integrated computing resources from ten global regions to build cloud-based R&D platforms, integrating hundreds of simulation tools. These platforms support high-intensity tasks such as automotive crash safety analysis and chip design verification, enabling global collaborative R&D.

Financial institutions are driving transformation powered by computing, supporting a large number of algorithm models and application scenarios. Their credit investigation report writing assistants can generate comprehensive enterprise research reports in a short time, significantly improving approval efficiency.

With the official launch of the cross-domain system of the National Computing Power Internet Service Platform, the vision of computing power being as convenient to use as water and electricity is accelerating towards reality.