The first automaker in the industry with an intelligent driving mileage share exceeding 50% has emerged.

```

On May 8, Xpeng Group disclosed a set of data: One month after the rollout of the second-generation VLA intelligent driving system, the proportion of smart driving mileage in real driving scenarios exceeded 50% for the first time.

On models equipped with this system, the human driver behind the steering wheel is becoming the "second choice," with the AI driver taking over more than half of the driving mileage.

For car companies whose new systems can achieve over half intelligent driving mileage,it alsomeans that intelligent assisted driving is moving past the stage of "does it exist or not," entering a new round of competition — "should you use it, is it good, do you trust it."

This is a milestone. When users are truly willing to hand over more than half their journeys to AI, the competition in smart driving is no longer a contest of specs on a feature list.China's assisted driving industry is irreversibly entering a new phase of accelerated development characterized by computing power and scaled implementation.

Xpeng's "More Than Half"

Xpeng's May 8 disclosureshowsthat during the one month since the push of the second-generation VLA, the proportion of smart driving mileage has surpassed half, and core engineering quality indicatorsareimproving.

During the same period, the number of driver interventions per 100 km for the second-generation VLA dropped 25.87% month-on-month, and the number of trips using VLA for 100% of the journey increased 27.84% month-on-month. In the recent "May Day" holiday stress test, Xpeng's AI assisted driving daily usage rate reached as high as 93.21%, with a cumulative mileage of 84.46 million kilometers, and even a record single trip with intelligent driving of 5,441 kilometers in an extreme long-tail case.

In the past, the standard for assessing an assisted driving system was often "how many types of roads it can drive," but nowadays the core criterion has shifted to "how long users are willing to let it drive."

Nearly a 26% drop in driver interventions per 100 kmis a more intuitive indicator.

In real open roads, interventions often occur when the system encounters unrecognizable obstacles, plans unreasonable routes, or performs overly abrupt braking maneuvers.

The fall in intervention rates proves that large end-to-end models are rapidly improving their ability to generalize in complex road conditions. Nearly 30% growth in journeys using VLA 100% of the time indicates thatusers' attitude is changing, from "let's give it a try" to "turn it on as soon as you get in."

A Xpeng G7 owner shared with Wallstreetcn that, during his self-drive trip from Beijing to Gansu over the "May Day" holiday, he essentially always turned on assisted driving for the whole journey. Before using assisted driving, he felt it would be impossible for him to do the round-trip 3,000 km drive on his own. He even tried assisted driving in his hometown's countryside after the trip, and though it was slower, the experience was still smooth.

Hetrusted assisted driving because he himself works in the AI industry.He believes the current AI technology does make assisted driving quite smart — "Wherever ordinary people can drive, AI can drive too."

Trust is built on more than just algorithms.Xpeng Group's Chairman and CEO He Xiaopeng has repeatedly stated in earnings calls and livestreams that autonomous driving has moved from software engineering to a capital-intensive "computing power engineering" field.

He Xiaopeng has revealed that, to sustain rapid iteration of large models, Xpeng's annual R&D investment in autonomous driving remains at a high level of several billion RMBandthey will continue to increase computing power investment each year going forward.

It is understood that by 2026, physical AI-related R&D investment will reach 7 billion RMB. For a car company still delivering tens of thousands of units a year, this intensity of investment is considerable.

Of course, Xpeng invests not only in cars, but also in self-developed chips, foundational models, and AI infrastructure, seeking to bet simultaneously on the "computing power–algorithm–data" triangle in physical AI.

Industry Turning Point

Xpeng's "more than half" is a microcosm of the industry's acceleration phase. Looking at a longer timeline, the assisted driving penetration rate in China's passenger car market has drawn a steep upward curve in recent years.

Around 2021, highway NOA functions began to appear in vehicle models priced from RMB 200,000 to 300,000, mostly as differentiating flagship-level configurations, with limited user selection and actual usage rates.

The turning point came in 2023 and 2024, as city NOA became the focus of competition, with automakers waging "city-opening wars" — whoever could announce coverage of hundreds of cities first would dominate the launch event narrative.

Soon, the slogan "drive anywhere nationwide" became an experience users could actually perceive, and the price range for advanced intelligent driving also dropped noticeably.

Penetration rate data supports this trend. According to the Automotive Industry Economic and Technology Information Research Institute, the penetration rate of new cars with L2-level functions in China reached 64% in the first three quarters of 2025, and is expected to hit 66.1% by year-end.

In 2026, NE Times data shows that in February, penetration of L2 and above assisted driving vehicle models reached 69.28%.

L2 functions are becoming standard in new cars, and city NOA, which truly represents "the gold value of smart driving," is also quickly gaining share. According to the China Association of Automobile Manufacturers report, from January to November 2025, cumulative sales of passenger vehicles equipped with city NOA reached 3.129 million units, with penetration reaching 15.1% of insured passenger vehicles, up 5.6 percentage points from 2024.

Meanwhile, Guojin Securities forecasts city NOA hardware configuration penetration could reach 25% in 2026, with annual sales of about 5.45 million units. From 15% to 25%, city NOA is shifting from early adopters to mainstream users.

Of course,higher penetration rates don't necessarily mean users are actively using it. Only the mileage usage stats disclosed by each company prove that assisted driving is truly integrating into more people's daily driving.

In April, Huawei Qiankun ADS announced cumulative assisted driving mileage exceeded 10 billion kilometers, with whole-vehicle installations exceeding 1.7 million units. Li Auto reached 1.505 million assisted driving users by the end of 2025, with total mileage over 6 billion kilometers.

As penetration rateand user usage rateclimb to new heights, the dimension of competition is bound to change. In other words, the industry is continuing past the early-stage feature contest, entering a second half focused on scaled implementation and real user usage. Xpeng's surpassing of 50% intelligent driving mileage proportion is like a starter's pistol for this second half.

Currently, the main focus of competition in the assisted driving industry has thoroughly moved from "proving technical feasibility" to "commercialized, scaled implementation."

In this new phase, car companies are not only competing on model advancement and computing power reserve, but also on reducing hardware costs, efficient and compliant data processing, and establishing irreversible trust in safety among broader user groups.

Looking back at industry changes, assisted driving has evolved from highway experimentation in 2021 to dominating over 50% of mileage today. Chinese car companies have reinvented the underlying logic of mobility in five years.

As the new round of accelerated development arrives, smart driving systems have become the soul of intelligent vehicles, and those companies failing to secure a ticket on this era's express trainwillface extremely severe survival challenges.

Risk warnings and disclaimerThe market has risks, and investment needs to be prudent. This article does not constitute personal investment advice and has not taken into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific situation. Invest at your own risk. ```