After 30,000 robovans, Zelos bets on city-level physical AI with a rental platform

On September 10, autonomous delivery company Zelos held a strategy launch event in Guangzhou, announcing an upgrade toward what it calls city-level physical AI powered by its Zelos Brain platform. At the event it released Zelos Car Rental, a mini-program aimed at express logistics, retail distribution and short-haul urban delivery, and opened franchising for unmanned transport capacity with promotional terms for early partners.

Alongside the product news, Zelos signed cooperation agreements with partners including GAC Lyncan, Guangzhou Public Transport Group, Guanghuan Investment, Guangzhou Oweide, Shangyuan and Zelos Jijing, covering areas from vehicle manufacturing to urban services and industry-academia collaboration. Taken together, the moves trace a path the company has followed for five years: build the vehicle first, push it into more real-world scenarios, then turn proven technology and operations into infrastructure that connects cities and industries.

Zelos started with the RoboVan. Unlike Robotaxis aimed at passenger travel, RoboVans target the large pool of box-truck demand in urban logistics, which means coping with complex public-road conditions while meeting requirements for payload, range, reliability and operating efficiency. The report says Zelos launched the world's first RoboVan in 2023 and pushed it into public-road operation.

Scale is the foundation of that approach. According to Zelos's latest public figures, its fleet now exceeds 30,000 vehicles, operates routinely in more than 300 cities worldwide and has accumulated over 250 million kilometers of operation, serving customers such as China Post, SF Express, the Tongda couriers, Mengniu, New Hope and Zhongce Rubber, with the company describing its market share as leading. For L4 autonomy, every real-world kilometer tests both the driving system and the operating model behind it.

Safety is framed by Zelos as the precondition for scale. CEO Kong Qi said at the launch that safety is not something to solve after scaling but a capability that must exist from day one. The company feeds real-road experience back into its algorithms through a data loop, builds redundant safety designs around vehicle control and communications, and runs thousands of tests; the report says it set up a safety committee this year to advance safety management at both the organizational and technical levels.

The rental idea came from real users. The report describes a user in Kunshan who bought four Zelos vehicles in early 2025 to rent out; as demand grew he expanded from four vehicles to dozens and now more than 200, reorganizing idle time through shared bookings and time slots. Zelos then piloted the rental model with 300 customers across 45 Chinese cities, completing more than 1,500 rental orders and over 1 million kilometers of operation.

Building on those pilots, Zelos formally launched Zelos Car Rental with daily, monthly and annual options, letting companies buy transport capacity on demand rather than assets, and opened city-level franchising. In this model Zelos supplies the vehicles plus fleet management and intelligent dispatch platforms, while local partners contribute their logistics, retail and transport resources to put the vehicles into express, supermarket and commercial delivery scenarios.

The system rests on two pieces of infrastructure: the Zelos Brain and Zelos Inside. In the company's definition, the Zelos Brain is not merely autonomous-driving software but an AI capability hub for the physical world that accumulates and reuses scenario and L4 operating experience. Zelos Inside then exports that stack, including the driving system, AI algorithms, hardware and software, and operating know-how, to more automakers and industry partners; Zelos has previously worked with Dongfeng among others.

What to watch is how the city-level layer takes shape. Zelos is building three platforms: one to organize and dispatch capacity, one to monitor and guarantee vehicle health, and one that lets governments see, manage and trace unmanned operations. The business model is extending from selling equipment toward supplying transport capacity, and the competitive focus is shifting from a single vehicle's autonomy to city-scale capacity orchestration and regulatory coordination.