On 4 November 2026, at the internet entrepreneurs' forum held as part of the World Internet Conference in Wuzhen, NIO founder and chairman Li Bin stated that today it is intelligent electric vehicles that are the most widely deployed form of embodied artificial intelligence in the world. Later, on the social network X, he summed up the idea briefly: smart electric vehicles are becoming the largest sector of AI application. An important detail: this refers to the scale of deployment, not to the number of deliveries.
The logic of his argument is simple. Each such vehicle combines computing power, sensors and large models that close the full loop: the car perceives its surroundings, makes a decision and executes it itself. Li Bin made the "largest sector" assessment himself, and there is as yet no independent calculation that would confirm it in figures.
To gauge the scale, it is useful to look at the fleet as a whole. According to the China Academy of Information and Communications Technology, by August 2026, about 61 million passenger cars in China had L2-level driver assistance functions, and about 15 million of them are equipped with navigation autopilot (NOA). That is roughly a quarter of the L2 fleet. It is important to understand that the figures describe cars already driving on the roads, not sales for the current year.
NIO itself reports that its intelligent assistance system, built on a world model architecture and closed-loop reinforcement learning, is installed on more than 950 thousand electric vehicles. According to the company's own data, the system has accumulated over 9,14 billion kilometres of safe driving and has helped prevent or mitigate 9,07 million potential accidents. NIO's internal statistics also say that the safe mileage under joint human-machine control is eight times greater than under fully manual driving. These figures have not yet been verified by an independent audit. For comparison: in September, speaking in Geneva at a WTO forum, Li Bin cited 950 thousand vehicles and 7,7 billion kilometres, meaning that in a month and a half the mileage grew by a noticeable amount.
There is also context that is hard to miss. In October, NIO delivered 34 246 vehicles, which is 15,2% fewer than a year earlier, marking the first year-on-year decline since February 2024. Claims of AI leadership come against the backdrop of stalling sales, and the company is essentially betting that the software side will become its main advantage.
Regulators are approaching this picture cautiously. From 1 January 2027, a mandatory national safety standard for combined driver assistance systems (GB 47955—2026) comes into force, and for L3 and L4 level systems the standard will take effect from 1 July 2027. At the same time, the Ministry of Industry and Information Technology has set the goal of mass deployment of autonomous vehicles by 2030. In other words, China is expanding pilot zones but tightening requirements for safety, software updates and the admission of vehicles to the roads.
For drivers, this means that the car is gradually turning from mere transport into a device that learns from data from millions of trips. Li Bin links this idea to safety: he cites World Health Organization data according to which about 1,16 million people died on the world's roads in 2025, and drivers spend more than 1,5 billion hours behind the wheel every day. According to his logic, this burden and the number of accidents can be reduced not by replacing the human but by supporting his decisions. But for now this is a promise, not a measured result across the entire industry.
The main question for the near future is how quickly such systems will become available outside China and to what extent their real-world performance will be confirmed by independent data rather than manufacturers' internal statistics.
