In the laboratory of Pengcheng National Laboratory, researchers are observing a new type of optical fiber that is as thick as a human hair under a 60x microscope, and four parallel optical transmission channels are clearly visible. This new medium, which has a transmission capacity four times higher than ordinary optical fibers, is a microcosm of the nationwide integrated computing power network construction in 2026. From the domestic computing power cluster of Zhengzhou Supercomputing Center to the new node of intelligent computing in the Guangdong Hong Kong Macao Greater Bay Area, the network of national computing power is rapidly expanding.

Centralized expansion of core computing power nodes

Since 2026, China's computing power base has continued to consolidate. Zhengzhou's national supercomputing Internet core node has put into use the first national 100000 calorie artificial intelligence supercomputer cluster, which integrates scientific computing and intelligent computing capabilities. The peak computing capacity per second is equivalent to 200 years of continuous human computing. At present, it has provided support for more than 300 computing tasks in 26 fields, including new materials, innovative drugs, and artificial intelligence.

At the national hub nodes of the Guangdong Hong Kong Macao Greater Bay Area, the newly launched domestically produced 10000 card intelligent computing cluster is specifically targeted at the research and development needs of trillion level large models. It can compress the training cycle of trillion parameter large models from one year to six months, while supporting the full chain inspection and testing environment of large models. As of the end of June 2026, the scale of intelligent computing power in China has increased by 177% year-on-year, reaching 218.5 trillion floating point operations per second, and the support capability of computing hardware base has achieved a leapfrog improvement.

The computing power scheduling network continues to take shape

In response to the pain point of uneven distribution of computing power resources in different regions, the national level computing power dispatch channel, the Shenzhen Guizhou line, is accelerating its implementation, relying on new optical fibers and low latency optical direct connection to connect the high-speed transmission link between Shenzhen and Guizhou, the two major computing power hubs. At present, the new national integrated computing power network monitoring and scheduling test verification platform has entered the trial operation stage, and over 60% of the country's computing power has been included in the unified monitoring system.

Around the eight major national hub nodes of East Digital and West Computing, a millisecond level computing power service circle has been formed in the main regions of the east and west of the country. The precise matching ability of computing power continues to improve, covering hundreds of application scenarios such as artificial intelligence, industrial manufacturing, and scientific research. According to the public plan, China's computing power network construction will increase direct investment by 4 trillion yuan during the 15th Five Year Plan period, with a focus on areas such as computing power scheduling, computing power collaboration, and ultra large scale intelligent computing clusters.

From the breakthrough of a single computing power device to the interconnection of cross regional computing power networks, the construction of the national computing power network in 2026 is breaking away from the traditional path of single point expansion. This nationwide computing power network not only enables the efficient supply of green computing power from the west to the industrial demand in the east, but also opens up the underlying support channels for the intelligent transformation of thousands of industries, continuously empowering the high-quality development of China's digital economy.Editor/Cheng Liting