XPeng’s LiDAR-free robotaxi is leaving the demo phase for the harder fleet test
A factory-floor reveal of a sleek XPeng-style robotaxi leaving an illuminated Guangzhou production line, with visible camera/radar sensor discipline but no roof LiDAR dome.📷 AI-generated image / TECH&SPACE
- ★XPeng’s robotaxi was produced in Guangzhou as a full-stack in-house vehicle and autonomy project.
- ★The vehicle targets L4 autonomy, uses four Turing AI chips delivering 3,000 TOPS, and removes LiDAR from the sensor stack.
- ★Pilot operations are reportedly planned for the second half of 2026, with early 2027 targeted for fully autonomous service.
XPeng has moved its robotaxi from announcement gravity into production reality. According to Electrek’s report, the first mass-produced unit has rolled off the line in Guangzhou, making XPeng the first automaker in China to claim mass production of a robotaxi built through full-stack, in-house development.
That phrase matters more than the usual launch-day varnish. Robotaxis are no longer judged only by whether they can complete a tidy route in a demo zone; they are judged by whether the hardware, software, manufacturing process, maintenance model, and safety case can survive repetition. A vehicle coming off a production line is not proof of deployment success, but it is a more serious signal than another edited city-drive video.
The technical bet is sharp. XPeng says the vehicle is engineered to L4 autonomous driving standards and uses four self-developed Turing AI chips delivering 3,000 TOPS of compute, while avoiding LiDAR entirely. That makes the stack more tightly controlled, but also puts more burden on cameras, radar, software perception, mapping assumptions, and validation discipline.
Guangzhou is only the start: the real test is fleet operation, supervision and edge cases
A closer operational scene showing the LiDAR-free robotaxi reading rain, scooters, lane markings and construction cones in a dense Chinese urban test zone.📷 AI-generated image / TECH&SPACE
The LiDAR-free choice is the detail worth watching. Removing LiDAR can reduce cost and simplify industrial scaling, but it also narrows the margin for perception errors in rain, glare, construction zones, unusual road behavior, and the ordinary messiness of dense traffic. Robotaxi systems do not fail in the showroom; they fail at the lane merge nobody designed a slide for.
XPeng’s roadmap reportedly points to pilot robotaxi operations in the second half of 2026, with fully autonomous operations targeted for early 2027. Those dates are close enough to be operational commitments, not distant mood lighting. The practical questions are route design, remote assistance, fleet uptime, cleaning, charging, insurance, and how the system behaves when passengers and traffic do things that are perfectly legal and deeply unhelpful.
If XPeng can turn this into a repeatable service, the implications are industrial rather than theatrical: purpose-built autonomous fleets, tighter integration between carmaker and software stack, and a clearer path from advanced driver assistance to paid mobility services. The real signal here is not that the robotaxi exists. It is that XPeng now has to prove production volume can coexist with safety validation, city permission, and the public’s low tolerance for beta software on public roads.
