Over 10 years in the making: China launches 2,000km-wide AI computing hub
Implications are revolutionary for extremely high real-time demands such as AI large model training, telemedicine, director Liu Yunjie says
Zhang Tongin Beijing, SCMP
Published: 12:00pm, 10 Dec 2025
China has activated what may be the world’s largest distributed AI computing pool, alongside a high-speed data network developed over more than a decade, according to a state newspaper.
The official Science and Technology Daily said the optical network connected distant computing centres, so they could work together almost as efficiently as a single giant computer.
The 2,000km-wide (1,243-mile) computing power pool formed via this network could achieve 98 per cent of the efficiency of a single data centre, Liu Yunjie, a member of the Chinese Academy of Engineering and chief director of the project, was quoted as saying in the report last Thursday.
China’s top computing centres are scattered across the country but this pool would allow them to operate as a unified system, working together seamlessly to fast-track the development of the most powerful AI models and other cutting-edge technology.
“The implications of this dedicated data highway are revolutionary for scenarios with extremely high real-time demands, such as AI large model training, telemedicine and the industrial internet,” Liu told the daily.
The Future Network Test Facility (FNTF) is China’s first major national science and technology infrastructure project in the information and communication sector. After more than a decade of development, it officially began operations on December 3.
The facility’s role in advancing the development of artificial intelligence is particularly significant, offering substantial savings in both time and economic cost, according to Liu.
“Training a large model with hundreds of billions of parameters typically requires over 500,000 iterations. On our deterministic network, each iteration takes only about 16 seconds. Without this capability, each iteration would take over 20 seconds longer – potentially extending the entire training cycle by several months,” he said.
“The facility is ideally positioned to serve the national ‘East Data West Computing’ project.”
Launched in 2022, the East-West project’s goal is to send data processing work from the busy east to the west where there is more clean energy available. For this to work, the network needs to be extremely reliable so that computing power can be turned on and off as easily as water or electricity.
FNTF was first outlined in China’s “Medium- and Long-Term Plan for the Construction of Major National Science and Technology Infrastructure” in 2013. Today, the facility spans 40 cities with a total optical transmission length exceeding 55,000km – enough to circle the equator 1½ times.
Operating around the clock, the platform can simultaneously support 128 heterogeneous networks and run 4,096 service trials in parallel, boasting high throughput, high reliability and deterministic transmission capabilities.
The project team led the development of 206 international and domestic standards, obtained 221 invention patents and created the world’s first distributed large-scale network operating system.
“In the future, this network will be opened to sectors like industrial manufacturing, energy, power and the low-altitude economy,” Liu said.
The network would also help advance scientific research for humanity. At last week’s launch ceremony, a massive 72 terabyte data set generated by FAST, the world’s largest single-dish radio telescope, was transmitted across 1,000km in just 1.6 hours from a university in Guizhou province in the southwest to another in Wuhan, a city in central China.
By comparison, sending the same volume of information over the regular internet would reportedly have taken about 699 days.
The high-speed transfer was enabled by a “deterministic network” channel. This technology employs a suite of advanced techniques to provide dedicated bandwidth with absolute guarantees of ultra-low latency and near-zero packet loss.
During the demonstration, researchers activated two parallel channels between Guizhou and Wuhan: one 50 gigabits per second deterministic network and one 42Gbps regular network. When extra digital traffic was added to mimic congestion, the regular network slowed down dramatically to below 1Gbps, while the deterministic network steadfastly maintained its full 50Gbps capacity.
A normal network is like a busy highway – traffic is unpredictable and delays are common. A deterministic network, however, works like a precise train timetable where every data packet is scheduled to travel a set route and arrive exactly on time, without delays or loss.
Reflecting on the project’s broader impact, Wu Hequan, a member of the Chinese Academy of Engineering who took part in its evaluation, told Science and Technology Daily that the technologies developed for the facility had already provided support for the research and development of 5G Advanced and 6G technologies.
“Going forward, both research institutions and enterprises will be able to test various new technologies on this platform,” he added.
Over 10 years in the making: China launches 2,000km-wide AI computing hub | South China Morning Post |
