Chinese tech company and EV maker Xpeng has achieved a domestic first: a robotaxi built entirely with in-house technologies has entered series production. But rolling a vehicle off a production line might well be the easy part. The real test starts in the second half of 2026.
Xpeng has officially rolled off the assembly line the first mass-produced unit of its robotaxi in Guangzhou. The company describes it as the first time in China that an automaker has achieved series production of a robotaxi through full-stack, in-house development. This means that chips, AI models, vehicle platforms, and manufacturing all originate from within the same organization.
The achievement comes from Xpeng not solely being a car manufacturer but also a tech company. It can rely on its ecosystem to provide a comprehensive solution for its mobility concepts.
Configured for passengers
The Xpeng Robotaxi is built on the same GX platform that underpins the company’s consumer flagship SUV. Interestingly, both vehicles share identical core hardware: four proprietary Turing AI chips delivering 3,000 TOPS of on-board computing power, the VLA 2.0 autonomous driving system, and an aviation-grade safety redundancy architecture with six layers.
What changes is the cabin. Where the GX is a luxury vehicle that owners drive themselves, the Robotaxi reconfigures the interior around the passenger experience: privacy glass, zero-gravity comfort seats, rear entertainment screens, and an AI voice assistant. Three body variants are planned: a five-, six-, and a seven-seater.
The platform-sharing strategy is deliberate. By validating the same hardware across millions of consumer vehicles first, Xpeng cuts both development time and cost. It gives Xpeng a structural advantage over companies building dedicated robotaxi chassis from scratch.
Just vision
There is no lidar on the vehicle, and unlike most self-driving systems currently in operation, it does not rely on preloaded maps of every street it drives on. Instead, the Robotaxi navigates purely by looking at the world through cameras, the way a human driver does.
The brain behind that process is Xpeng’s VLA 2.0, short for Vision-Language-Action model. Think of it as a single AI agent that has been trained to watch the road, understand what it sees, and decide what to do next, all in one continuous flow.
Traditional autonomous driving systems split those tasks across separate software modules. VLA 2.0 reduces the entire chain to a single model. The great advantage is reaction time: under 80 milliseconds, or roughly the blink of an eye. It’s also twelve times faster than Xpeng’s previous system.
Anywhere, without mapping
There’s also a practical advantage: because the car reads its surroundings in real time rather than matching them against a preloaded map, it can theoretically be deployed in any city without requiring months of prior mapping work.
Pilot operations are planned for the second half of 2026. The goal of removing the onboard safety driver is planned for early 2027. Xpeng is entering a domestic sector that is dominated by Baidu, Pony.ai, and WeRide. Now that the tech and manufacturing are in place, the question is whether Xpeng can build a robotaxi business.
