China's most infamous ghost town is now training ground for driverless trucks (restofworld.org)

Ordos, the once-empty “ghost city” in northern China’s Kangbashi district, has been repurposed into a large-scale sandbox for autonomous vehicles. Local government has designated 355 km of real-world roads and installed more than 2,518 roadside devices (including lidar and roadside units) to enable vehicle-to-infrastructure and vehicle-cloud integration. The city’s one-sixth share of China’s coal reserves supplies steady freight demand, attracting firms such as KargoBot (Didi-backed, >300 trucks) and U.S.-listed WeRide plus partnerships with Baidu and Huawei. KargoBot now runs convoys of 2–6 trucks with only a safety operator in the lead vehicle, reporting ~300 million yuan in revenue last year and targeting >500 million yuan this year—evidence that industrial AV commercialization can scale in low-risk environments. Ordos illustrates China’s strategic approach to convert surplus infrastructure into testbeds that accelerate commercialization of freight and mining autonomy, while building rich roadside sensing networks for real-time data. However, the city’s low population and sparse, predictable traffic limit exposure to chaotic urban behaviors—pedestrians, cyclists, and dense rush-hour interactions—so firms must augment Ordos data with tests in crowded cities to generalize AI models. For commercial freight AVs, Ordos is ideal; for consumer robotaxis, it’s only a starting point. The broader implication: well-instrumented, low-risk real roads can rapidly advance industrial AV deployment, but end-to-end autonomy still requires varied, high-complexity environments.