XPeng's bold bet on intelligent driving in the second half

Author | Wang Xiaojuan

Editor | Zhou Zhiyu

"During a business trip to the United States this year, I found that American companies are all focusing on end-to-end, large-scale models, while we are only focusing on weekly sales rankings."

He Xiaopeng, the usually low-key chairman of XPeng Motors, has also joined the recent "criticism" of domestic car companies focusing on weekly rankings.

At the "AI Autonomous Driving Technology Release Conference" on July 30th, He Xiaopeng openly stated that just this year, XPeng has invested 3.5 billion yuan in autonomous driving and has also hired 4,000 people.

Over the past three years, XPeng's R&D expenses were 4.114 billion yuan, 5.215 billion yuan, and 5.277 billion yuan respectively, accounting for over 17% of XPeng's total revenue. In the first quarter, XPeng's R&D investment was only 1.35 billion yuan, indicating an accelerated investment in autonomous driving.

Behind the bold bet on autonomous driving is XPeng's desire to make its autonomous driving label clearer.

At the conference, XPeng revealed that in the future, XPeng's autonomous driving will be available globally, without restrictions on cities, routes, or road conditions in specific scenarios.

Since major automakers started focusing on autonomous driving at the Shanghai Auto Show last year, the autonomous driving field has seen rapid development. Some car companies have even achieved "nationwide availability" of autonomous driving.

However, XPeng believes that phrases like "nationwide availability" with qualifiers are not enough. This time, with the upgraded XNGP, XPeng has achieved full availability.

In addition to nationwide availability, the updated features include camera perception for parking, and 100-meter trajectory reverse parking.

The "100-meter trajectory reverse parking" feature automatically returns along the original driving trajectory in extreme scenarios such as dead ends and narrow roads where U-turns are not possible, achieving "one-click return to the original position."

XPeng also discovered that 60% of its users are Apple users. Therefore, in this OTA update, the system is now integrated with the Apple ecosystem, adding functions such as remote voice control, gesture control, and versatile NFC touch control.

These capabilities were made possible by XPeng being the first to achieve end-to-end large-scale model mass production and installation. Currently, XPeng is the only one globally and the first in the country to achieve end-to-end large-scale model installation by an automaker.

After achieving end-to-end installation, XPeng's OTA update speed has also increased. Currently, they can achieve "version iteration every 2 days and experience upgrade every 2 weeks."

At the beginning of the year, XPeng set KPIs ahead of time during a sharing session. At that time, XPeng said they wanted to expand autonomous driving to more long-tail scenarios. This time, XPeng has already covered scenarios such as ETC and aims to achieve "door-to-door" completion within the year.

Externally, the immediate priority for XPeng is still increasing sales. However, He Xiaopeng is not anxious about sales.

He believes that the automotive industry is a marathon, and this year, he has already felt internal changes. Sharpening the knife will not delay the work of cutting wood. In recent months, XPeng has been sharpening the knife, and the increase in sales will not take too long.

Furthermore, XPeng is also planning for new sales growth points. With so many years of accumulation in the autonomous driving field, He Xiaopeng must have plans. He clearly stated that XPeng will also enter the Robtaxi field in the future In He Xiaopeng's view, following computer networks, the internet, mobile internet, new energy vehicles, and cloud services, AI will start leading new trends and technological waves after 2023, bringing about four new directions: chips, large models, autonomous driving, and robots.

Among these, Xiaopeng aims to focus on autonomous driving, planning to enter the industry in cooperation with operators in the future. He revealed in an interview that he recently met with the CEOs of Didi and Uber.

However, even in the autonomous driving industry, competition remains fierce. Xiaopeng's bold bet on intelligent driving also requires validation from the market.

Below is a transcript of the dialogue between He Xiaopeng, Chairman of Xiaopeng Motors, and Li Liyun, Head of Xiaopeng Intelligent Driving, as reported by media outlets such as Wall Street News (edited):

Question: People have high expectations for Xiaopeng in terms of intelligent driving, but the sales volume has not met expectations. How do you view the sales volume? Why has it not met external expectations?

He Xiaopeng: I think there are several points: first, our AI technology is not yet good enough; second, the pressure on pure electric vehicles has been high in the past two years; third, the competition in our segment is the most intense, with two or three companies above 300,000 units and two or three above 200,000 units. We are in the segment of 200,000 to 300,000 units, where there are the most players, and new entrants will want to compete with us.

Since two years ago, our changes have become more and more evident, and I am increasingly confident in the long-term marathon. You won't have to wait long; everyone will see that Xiaopeng is well-prepared. If sales are like chopping wood, our recent actions are all about sharpening the axe. If you ask me next year at this time, you might say, "Xiaopeng, why did the first phase of your axe sharpening work so well?"

Building a business requires a long-term perspective.

Question: Will Xiaopeng enter the autonomous driving taxi industry? Or will it provide technical support to related companies?

He Xiaopeng: We will get involved in Robotaxi, but we are not willing to take over operations. We hope that in the future, Robotaxi will not only serve China but also the world.

In the past month, I met with the CEOs of Didi and Uber, both of whom are doing well in the field of autonomous driving. I believe there will be many such operators globally in the future, and I hope to cooperate with them.

Building cars is difficult, operating them is even more challenging, it's not about "adding difficulty," it's about "multiplying difficulty." There should be more partners working together to do this well.

Question: How much does intelligent driving contribute to Xiaopeng's sales volume? Can intelligent driving truly become a moat for a car company?

He Xiaopeng: The current doubts about intelligent driving are very similar to the attitudes towards new energy vehicles five years ago. In 2019, many people dared not predict the future of new energy vehicles.

I believe that the real game-changer will be AI. In the past 12 months, the proportion of high-end vehicles with intelligent driving features in Xiaopeng's sales volume has been increasing. We believe that AI is a crucial moat and valuable data. Even for cars in the 200,000 price range or even cheaper, they can leverage this data well. The difference between having 10,000, 1 million, and 100 million data points is significant I believe that it will not take 5 years, only 3 or 4 years, for everyone to see that autonomous driving will be a huge change. I am very confident.

Question: What is XPeng's next plan after XNGP can be driven nationwide?

He Xiaopeng: The next step is to make it drivable nationwide, taking 18 months to achieve something similar to Waymo. In the fourth quarter, we want to achieve "door-to-door" service. Currently, it is only for some people, but our goal is for all XPeng owners to be able to use it. I think this is quite challenging because there are too many different scenarios, each of which not only needs to be done but also needs to be more reliable, stable, and accurate, which is very difficult.

Question: What are the practical differences that end-to-end large models bring to users after being installed in the car?

He Xiaopeng: My goal, as I mentioned to Li Yun, is that by the second half of next year, with the current car (Max version), it can be OTA upgraded to the level of Waymo abroad.

This means that on average, a person driving 70 kilometers a day will only need to take over the control once or twice a week. This experience will be revolutionary and completely different from the current driving experience.

I think this is no longer distant, it can be achieved within 18 months or even shorter. Although there will be pressure, just imagine, when we were young, we saw people driving cars, but three, four, or five years later, there will be fewer and fewer people driving, holding the steering wheel, and people will be able to do other things inside the car.

After autonomous driving is activated, others may not know that you are using autonomous driving, but in the future, they will know. At some point this year, all new models will have this feature because I am using autonomous driving, so I am different from others.

Question: Why did you persuade Lei Jun to make cars?

He Xiaopeng: I told him at that time, in the end, cars will still depend on autonomous driving. From a different perspective, we are happier with more friendly competitors