LANZHOU, April 28 (Xinhua) -- On the ridges of Huanxian County in northwest China's Qingyang City, wind turbines spin above the loess hills and rows of solar panels stretch across the slopes. The electricity generated there is flowing not only to households and factories, but increasingly to server racks hundreds of kilometers away.
For Qingyang, formerly known as a producer of oil, gas and coal, this represents a shift in how its resources are being used. In the age of AI, energy, land and networks are becoming part of a new industrial equation.
Behind chatbot responses, image generation and voice interaction are physical infrastructures: data centers, reliable electricity, ample land, cooler climates for operation, and high-speed transmission networks. As AI applications spread, regions far from China's coastal tech hubs are gaining strategic significance.
The immediate driver is the surge in demand for tokens -- the basic units that AI models use to process information.
According to the National Data Administration, China's average daily token calls have exceeded 140 trillion in March this year, which represents a surge of more than 1,000 times compared with 100 billion at the beginning of 2024, and an increase of over 40 percent compared with 100 trillion at the end of 2025.
That growth is helping Qingyang play a significant role in China's wider push to build the infrastructure needed for a digital and smart economy.
"Qingyang has unique regional advantages for developing the computing-power industry," said Zhou Jijun, a leading official of the city.
China launched its "East Data, West Computing" project in 2022, aiming to channel data and computing demand from economically developed eastern regions to western areas with greater advantages in energy availability and land resources. Qingyang is one of the eight national computing hubs and 10 data center clusters in the country's integrated computing-power network.
Its role is to serve computing demand from regions including Beijing-Tianjin-Hebei, the Yangtze River Delta and the Guangdong-Hong Kong-Macao Greater Bay Area.
Qingyang's appeal is partly geographic. It is within 2,000 km of Beijing, Shanghai, Guangzhou and Shenzhen, and its relatively cool climate and stable geological conditions make it suitable for data center development.
But its strongest advantage may be energy supply. Intelligent computing centers are heavy electricity users, with industry data showing that power can account for more than half of their operating costs.
A green power aggregation pilot project, developed by a Qingyang-based new energy subsidiary of Gansu Electric Power Investment Co., Ltd., recently connected its first units to the grid. The project sends clean electricity from wind and solar facilities in Huanxian to the city's "East Data, West Computing" industrial park.
"About 55 percent of the park's electricity can be guaranteed through green power aggregation, with the rest supplemented through the power market," said Wang Jiaxue, head of the company. "The final electricity price for users will be no more than 0.4 yuan (about 6 U.S. cents) per kilowatt-hour."
Together with industrial land priced at 200,000 yuan per mu (about 0.067 hectare) and supporting infrastructure, lower power costs are helping Qingyang attract businesses with high demand for real-time computing.
Inside the Qingyang industrial park, rows of standard racks stand in orderly lines, filled with servers whose indicator lights flicker as data flows through the system. The park is connected to other national computing hubs through high-speed network links, allowing it to provide computing and application services across the country and beyond.
So far, the park has built 102,000 standard racks, with computing power reaching 142,000 PFlops. PFlops (petaflops) is a unit used to measure computing speed for AI and data-intensive tasks, where 1 PFlop equals one quadrillion calculations per second. Nearly 600 digital enterprises have established a presence in Qingyang, including companies in AI, internet services and cloud computing.
Kingsoft Cloud was among the early major companies to enter the park. Yang Jinwei, general manager of Kingsoft Cloud's northwest region, said the company's integrated computing-power dispatching platform has already been connected to the Qingyang data center. The company also plans to support the expansion of WPS AI services and explore business scenarios such as large models for autonomous driving.
That local push now fits a broader national priority -- treating computing power as core infrastructure for the smart economy. China's 2026 central government work report, for the first time, called for creating "new forms of a smart economy," with hyper-scale intelligent computing clusters and coordinated computing-power and electricity development listed among key tasks.
Dong Zongmou, deputy director of the management committee of the Qingyang industrial park, said the change has been visible this year, with more companies coming to discuss cabinet leasing and data center construction.
"There have been many phone inquiries and on-site visits recently, and the park is also accelerating construction to meet market demand," he said.
For Qingyang, the bigger question is not just how many server racks it can build, but whether computing power can help foster a broader local industrial ecosystem.
In Ningxian County, workers at Jiuzhou Wei'an (Gansu) Technology Development Co., Ltd. are assembling large drones. The company's products are designed for agriculture, forestry, emergency rescue and urban governance.
Zhang Bailong, deputy general manager of the company, said drone research, testing and operations all require computing support. The company can now produce nearly 2,000 drone bodies a month and plans to introduce production lines for drone motors and permanent magnet motors this year.
In Huanxian, EAST Digital Equipment Manufacturing (Huanxian) Co., Ltd. has built intelligent production lines for charging piles, energy storage systems and digital energy micro-modules, drawing on local computing resources as well as wind and solar power advantages.
"We completed the transition from project construction to production within one year," said Kou Long, head of the company. "By the end of 2025, the project had achieved an output value of 590 million yuan, created jobs and filled a gap in local manufacturing."
AI-enabled uses of computing power are also spreading into traditional sectors. In the Qingyang section of Changqing Oilfield, drones equipped with infrared thermal imaging devices are replacing manual pipeline inspections. At an intelligent apple sorting and storage center, AI can detect sugar content, color and fruit diameter for a single apple within 0.5 seconds. Meanwhile, an AI system for traditional Chinese medicine is providing auxiliary diagnostic support to grassroots medical institutions.
These are just a few examples of how Qingyang is working to move beyond hosting data centers and apply computing power to manufacturing, energy, agriculture and healthcare.
As demand grows around the production, use, distribution and settlement of tokens, the city will also need stronger support in talent training, financing and technical services.
For Qingyang, the test is whether its advantages in energy, land and computing infrastructure can be turned into lasting industrial capacity. The AI boom has opened a new window for the city, but the transition is still unfolding. ■
