<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/2972b729-95c0-4586-93be-703059f158f4.jpeg?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/2972b729-95c0-4586-93be-703059f158f4.jpeg"/> On the evening of March 16, He Xiaopeng rarely revealed the details.
Xiaopeng's second-generation VLA is defined by He Xiaopeng as &#34;the first version designed for L4 capabilities,&#34; with the core logic being to solve problems entirely with AI rather than rules. During the Two Sessions, representatives have been intensively calling for the acceleration of the revision of the &#34;Road Traffic Safety Law.&#34;
To provide sufficient computing power and data for the second-generation VLA, Xiaopeng invests a pure monthly amount of 300 million, which has been continuously burned for over ten months.
After the merger of the autonomous driving and intelligent cockpit business lines, the General Intelligence Center led by Liu Xianming has a clear task: to turn this money into user trust. The traffic of test drive customers attracted by intelligent driving at terminal stores is indeed recovering.
However, at this moment, a video of &#34;four children lying across the road&#34; sparked heated discussions online, pushing the boundaries of intelligent driving capabilities into the spotlight. The system detected the anomaly and slowed down, but ultimately it was human intervention that completely stopped the vehicle.
The focus of the autonomous driving competition has changed: it is no longer about crushing parameters, but about clarifying to the public where the safety bottom line of machines lies and when humans should take over.
<h2>Response and Boundaries of Intelligent Driving</h2>
During the live broadcast, a recently circulated video was discussed: several children lying in the middle of the road, and a Xiaopeng car with the intelligent driving function recognized the anomaly and slowed down, followed by the driver intervening to brake.
Liu Xianming retrieved and interpreted the backend data during the live broadcast. The data showed that the vehicle did indeed recognize the obstacle and triggered a slowdown, but he also pointed out that the deceleration at that time was insufficient to bring the vehicle to a complete stop.
He Xiaopeng added that the driver was an internal employee of Xiaopeng testing the system, who noticed the vehicle's abnormal deceleration and took action to brake.
The system detected the problem, but ultimately, a human still has to make the decision. This is the current real state of intelligent driving: it can provide perception and warnings, but extreme scenarios still require human judgment.
Different technical routes have varying strategies for responding to this—some solutions focus on algorithm generalization capabilities, while others emphasize sensor redundancy, but all face the challenge of long-tail scenarios.
To make ordinary people willing to drive, the core is to establish trust. The above case is precisely about building trust.
During the live broadcast, He Xiaopeng once again proposed the concept of &#34;national intelligent driving,&#34; defining it as &#34;intelligent driving that even moms love to drive,&#34; emphasizing that the intelligent driving system should return to the essence of safety and ease of use. He mentioned that after inviting employees without technical backgrounds to experience it, acceptance significantly improved.
The data is more intuitive: the second-generation VLA has reduced hard braking by 99% and rapid acceleration by 98%. No more sudden braking or rapid acceleration, significantly reducing passenger discomfort—this is a key step for intelligent driving to transition from &#34;usable&#34; to &#34;willing to use.&#34;
After moving away from high-precision maps, the second-generation VLA can cover more unstructured roads. However, Liu Xianming admitted that the current version occasionally does not follow navigation completely.
He explained this as a necessary stage in the transition from rule-driven to reasoning-driven—where the system is trying to understand the environment and make autonomous decisions, but requires users to provide more adaptation time in certain scenarios
He Xiaopeng also drew a line: in extreme weather and scenarios where even human drivers find it difficult to cope, it is not recommended to use intelligent driving.
This is also the consensus in the industry: the system is an auxiliary tool, not a substitute.
The competition in intelligent driving is shifting from parameters to trust.
In the past, the competition was about the breadth of functional coverage; now it is about who can make users truly dare to use and be willing to use.
<h2>Investing 300 million monthly, returns are coming</h2>
“I’m spending 300 million every month to bet on this, for more than ten consecutive months, and at that time, I was also quite anxious.” He Xiaopeng admitted rarely during a live broadcast. Where is this “bet” placed? Liu Xianming's answer is: from chips, compilers to software architecture and data closed-loop, full-stack self-research. “Mainly, you have to dare to bet.”
Why dare to bet like this? He Xiaopeng threw out a highly controversial judgment—China's intelligent driving should leap directly from L2 to L4. If it stays at the L3 level, it is easy to lose in global competition.
The core of this leapfrog logic lies in the division of responsibility. Liu Xianming explained that L4 requires the system to hard-solve all problems and cannot pass difficult issues to users. He Xiaopeng revealed that the second-generation VLA is the first version designed for L4 capabilities, with the core logic being to solve problems entirely with AI, rather than rules.
The market is also providing positive feedback.
He Xiaopeng revealed that since March 11, when XPeng's 732 stores nationwide opened test drives for the second-generation VLA, the test drive rate at stores has doubled, with many users coming specifically to experience this system. More direct returns are reflected in the order structure: the sales proportion of the Ultra version models equipped with this system has significantly increased.
In terms of push rhythm, gradual push began on March 19, prioritizing the XPeng P7 Ultra, followed by the XPeng G7 and X9 Ultra. Users of these three models will receive updates within this month. In April, more models will be gradually pushed.
Regarding the version differences that users are concerned about, He Xiaopeng provided a clear distinction: the Ultra version is designed for L4 level capabilities, supporting full-scenario passage; the Max version mainly covers high-frequency scenarios such as highways and urban main roads.
Beyond the immediate commercial returns, XPeng's goal is global. By 2025, the XPeng autonomous driving team has set a flag to compete with Tesla.
Recently, the media has been intensively comparing the second-generation VLA with Tesla FSD V13, and He Xiaopeng provided his judgment.
“From V13, we have a clear advantage. But I think it’s because XPeng is in China, and the data is in China, making us more familiar with Chinese road conditions.” He emphasized that the second-generation VLA performs better in “human-vehicle games,” such as with delivery personnel, pedestrians, and narrow roads. “This is not only a characteristic of China; there are many narrow roads in Europe and Southeast Asia as well. As we enter more countries, XPeng may have more advantages.”
Liu Xianming, however, is more cautious: “Actually, we don’t know how Tesla does it; it’s more like crossing the river by feeling the stones, stumbling through many pits, and wasting a lot of money. But we believe that the ultimate solution may converge through different paths.”
He Xiaopeng believes that both China and the United States are in the first echelon of intelligent driving, but the complexity of China's roads is ten times that of the United States—there are not only highways and urban roads but also rural paths outside third and fourth-tier cities, where one might encounter cows, sheep, and chickens while driving.
&#34;Autonomous driving is a comprehensive competition of software and hardware, engineering capabilities, and scale capabilities. Currently, both China and the U.S. are in the first tier. However, China's roads are more complex, and only by first overcoming the challenges to enhance the generalization ability of AI large models can the second-generation VLA truly achieve global implementation,&#34; said He Xiaopeng.
Any technology must undergo user verification. The industry's current judgment is that the turning point for autonomous driving has arrived, but the actual situation depends on whether users who have experienced it are willing to recommend it to more people.
The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take 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 circumstances. Investment based on this is at one's own risk

MOTO

XPEV

DRIV

He Xiaopeng revealed during a live broadcast that XPeng invests 300 million yuan each month in the research and development of the second-generation VLA, aiming to solve autonomous driving issues through AI. Although test drive traffic has rebounded, a recent incident involving a child lying on the road has raised public concerns about intelligent driving safety. Liu Xianming pointed out that the system can recognize obstacles and slow down, but manual takeover is still required. He Xiaopeng emphasized that building user trust is key, and intelligent driving should return to safety and ease of use

- On March 16, He Xiaopeng provided insights on the second-generation VLA, emphasizing AI's role in achieving L4 capabilities.  
- The company is investing 300 million monthly to build trust with users as it navigates challenges like balancing automated driving performance and human intervention.  
- As demand for test drives increases, there is a shift in focus from technology parameters to user trust, with ambitions to compete globally against Tesla, leveraging advantages in understanding complex Chinese road conditions.

Wallstreetcn

SmartETFs Smart Transportation & Tech ETF

XPeng

Global X Autonomous & Electric V

ChinaAMC CSI Hong Kong Connect Automobile Industry Thematic ETF

Hua Xia CSI New Energy Vehicles ETF

Max Auto Industry 3X Leveraged ETN

Guotai CSI 800 Automobile and Parts ETF

E Fund CSI Auto Components Thematic ETF

Hwabao WP CSI Intelligent Electric Vehicle ETF

First Trust S-Network Future Vehicles & Tech ETF

XPENG-W

He Xiaopeng burns 300 million every month. Can it really make ordinary people dare to use intelligent driving?