China's most infamous ghost town is now training ground for driverless trucks

A Chinese city that failed to attract people has become a sandbox for machines.

Ordos, a coal-rich northern Chinese city that never took off after prices crashed in 2012, has attracted more than a dozen autonomous vehicle companies seeking risk-free testing environments. Its wide, well-paved roads and minimal traffic provide a real-world laboratory where self-driving vehicles can operate without endangering pedestrians or navigating complex traffic.

Ordos’ transformation from property disaster to tech testing ground reflects China’s broader strategy of turning infrastructure failures into technological opportunities. The city holds one-sixth of China’s coal reserves, providing the heavy cargo that now trains its fleet of autonomous trucks.

“Being able to be known as an autonomous driving city has become an extremely valuable brand,” Liu Lidan, a researcher at China’s Anbound think tank, told Rest of World. “It represents hundreds of billions of investment opportunities and helps the city to attract investment and talent.”

Ordos’s transformation came two decades after city planners began developing Kangbashi, a new district within Ordos designed as a showcase for the region’s coal wealth. Built starting in 2004, this 32-square-kilometer satellite city was envisioned as a futuristic metropolis complete with stadiums, museums, and luxury apartments.

The 2012 global coal price crash and China’s property bubble left more than 70% of Kangbashi’s properties unfinished. Ordos New Media Center

The 2012 global coal price crash and China’s property bubble left more than 70% of Kangbashi’s properties unfinished, according to Chinese media reports. Today, only 131,000 people live in the district, while greater Ordos has just 2.2 million residents spread across 86,000 square kilometers, roughly 25 people per square kilometer.

“There is a huge demand for cargo logistics in Ordos,” Chen Hongjun, the general manager of the city’s state-backed autonomous vehicle testing agency, was quoted in a July 2025 report. “It offers ideal conditions for autonomous driving for commercial use, like in freight and mining sectors, to achieve commercialization.”

China has made autonomous vehicles its top technological priority alongside robotics and AI development. Ordos’ failed 2010s development provides ideal conditions for testing autonomous trucks, robotaxis, and smart buses. The city government has designated 355 kilometers (221 miles) of roads specifically for autonomous driving tests, making it one of China’s 20 official test grounds for vehicle-cloud-integration systems.

Being known as an autonomous driving city has become an extremely valuable brand.”

Ordos has installed more than 2,518 roadside devices, including laser radar and roadside units, to facilitate real-time data exchange between vehicles and infrastructure. This system collects data from roads, traffic lights, and vehicles to provide traffic and road safety information for autonomous driving.

KargoBot, backed by China’s largest ride-hailing service Didi, has been testing trucks in Ordos since 2021. The company’s fleet of more than 300 trucks, the largest self-driving operation in Ordos, was trained in a region that has one-sixth of China’s coal reserves.

After years of testing, the company now operates in Ordos with a safety operator present only in the lead vehicle of convoys, which range from two to six trucks. KargoBot reported a revenue of 300 million yuan ($42 million) last year, and expects to exceed 500 million yuan by the end of this year from commodity transportation, primarily from its Ordos operations.

“Ordos is big, but not overly massive,” Wang Ke, KargoBot’s vice president, told Rest of World. “It also has one-sixth of all of China’s coal reserves, which provides an industrial setting for us to test our vehicles.”

U.S.-listed Chinese robotaxi company WeRide began testing in Ordos in 2023, joining Baidu and Huawei in partnerships with the local government. The companies are developing smart transportation systems that assist self-drive vehicles.

Kargobot has been testing and operating its self-driving cargo truck fleet in Ordos since 2021.

While governments worldwide are designating testing grounds to attract autonomous driving investments, each serves different purposes.

California’s GoMentum Station transformed a former naval weapons base with 20 miles of paved roads into the U.S.’s biggest segregated autonomous driving test site. South Korea built K-City, a 79-acre mock town featuring highways, downtown streets, and parking lots for testing.

Even within China, different cities serve different autonomous vehicle needs. Yizhuang, a Beijing suburb, hosts larger fleets of robotaxis and robobuses to align with its high-tech hub ambitions.

Ordos remains unique for its combination of industrial-scale commercial demand for cargo transport, and urban infrastructure without the population. Unlike purpose-built test tracks or carefully controlled zones, Ordos offers real city roads with real business applications but smaller pedestrian risks.

Still, testing environments like Ordos have inherent limitations. The emptiness that makes the city safe for testing fails to provide the unpredictable scenarios that autonomous vehicles need to master. These include pedestrians darting into traffic, cyclists weaving through lanes, and the chaos of rush hour.

Ordos is big, but not overly massive. It also has one-sixth of all of China’s coal reserves, which provides an industrial setting for us to test our vehicles.”

As companies operating there acknowledge, developing truly capable AI requires exposure to more challenging environments, not just empty boulevards. This empty-city advantage works perfectly for industrial autonomous vehicles but creates challenges for consumer-focused applications, according to  Kevin Mak, the U.K.-based principal analyst in the Global Automotive Practice of information platform  TechInsights.

“Much of the world is now looking at testing in their natural cities,” Mak told Rest of World. “The commercial opportunity to test and then deploy are still in large populated cities where mobility providers can generate revenue.”

Kargobot’s self-driving trucks deliver cargo at an electricity-generating plant in Ordos.

KargoBot has expanded business to seven Chinese cities to collect more diverse data despite maintaining its primary operations in Ordos. Henry Liu, director of the University of Michigan Transportation Research Institute, sees limited potential for robotaxi services in the city.

“For robotaxis to make money, there needs to be more potential passengers, the place needs to have a dense population in densely populated cities for [the companies] to commercialize,” said Liu, who runs MCity, one of the U.S.’ well-known autonomous driving test sites. 

For KargoBot, Ordos was only meant to be a starting point for its business. 

“Ordos provides a suitable setting for the initial launch of our vehicles and operations,” Wang said. “But for developing our technology, we actually need data from more complex environments to strengthen and challenge the AI ability.”

The transformation of Ordos reflects broader shifts in China’s economic strategy. Local governments are scrambling to position themselves within Beijing’s vision of a modernized, tech-driven economy through autonomous driving initiatives. Yet for Ordos, the same low population density that doomed Kangbashi’s original vision now limits the city’s potential as a comprehensive autonomous driving hub.

“We have to train in more challenging environments [than Ordos] because it is a bit like having to train yourself with college-level maths,” Wang said. “You do it so you can handle a high-school level exam with ease.”