He Xiaopeng: We will definitely sell more robots than cars in ten years.

Text by | Xiao Man
Edited by | Li Qin

On November 5th, after announcing the Physical AI strategy, He Xiaopeng didn't sleep all night. The internet went crazy over XPeng's IRON robot. Some people said that the robot displayed by XPeng was actually a person in a suit.

To clarify this, He Xiaopeng recorded a video this morning. He smiled as he read a comment, "The robot walks more smoothly than Tesla's. It's impossible. It must be a person." Then he walked behind the robot, unzipped it, showed the radiator on the back of the robot, and let the robot walk around again.

This technology demonstration has achieved better results than He Xiaopeng had expected. Many Chinese automakers claim to become AI companies, but XPeng Motors is the first to showcase its products and make people remember them.

In He Xiaopeng's view, in the future, large models will become the "operating system of the Physical AI world" and the foundation for all Physical AI applications.

At this Technology Day, XPeng also launched its first mass - produced large model for the physical world, the "Second - generation VLA", which eliminates the intermediate translation step of language. According to He Xiaopeng, the Second - generation VLA is both an action generation model and a physical world model for understanding and deduction.

XPeng is trying to enable machines to gradually have the ability to understand, interact with, and change the world. The exploration of these underlying technologies will be applied to smart cars, robots, and flying cars.

These are the most fashionable narratives in the current technology field. In recent years, Elon Musk has often emphasized that Tesla is the world's strongest Physical AI (artificial intelligence technology that can understand and interact with the real world) company. Tesla has heavily invested in Robotaxi and humanoid robots, and recently revealed that it will launch a prototype of a flying car.

XPeng's narrative framework is similar to Tesla's, but not exactly the same.

During the post - meeting communication, He Xiaopeng mentioned that XPeng's ecosystem is more open, which is different from Tesla. It will cooperate with Alibaba's AutoNavi Map to build a Robotaxi travel network. The advantage of this is that each branch of XPeng Motors can independently verify its business feasibility, and resources can be reorganized at any time to reduce the financial burden.

After a two - year adjustment period, XPeng Motors' delivery volume has returned to the growth track, exceeding 40,000 units for two consecutive months. This temporary financial security gives He Xiaopeng the confidence to "tell a more distant story".

"XPeng has never fully believed that simply making a car bigger, more beautiful, cheaper, and of higher quality can lead to victory. These are only necessary conditions, not sufficient ones. It is important to do a good job in hardware, which is the foundation, but software also needs to be well - developed."

Although the story of Physical AI is appealing, the consecutive quality recall incidents this year at XPeng, Xiaomi, and Li Auto serve as a warning to the leaders of these enterprises. The industrial logic and inherent attributes of automobiles have not changed. When it comes to making a reliable automotive product, one still needs to be extremely cautious.

Even for the Robotaxi, which is highly favored by the industry, its implementation still faces a series of regulatory and technical challenges, and no one has proven that this business model can be profitable and generate economies of scale. Robots are an even more distant dream.

XPeng has a similar technology layout to Tesla, which leaves room for imagination and possibilities. However, it still needs to continuously prove that it has sufficient self - financing ability to support its long - term investment in Physical AI like Tesla.

The following is an edited transcript of the communication between 36Kr Auto and He Xiaopeng, the chairman of XPeng Motors:

Question: Why does XPeng insist on making highly anthropomorphic robots? Highly anthropomorphic robots require extremely high investment costs. How does XPeng make trade - offs?

He Xiaopeng: I believe that in the future, high - level robots will come in different forms, but anthropomorphic robots have three major advantages -

First, for a robot to be intelligent, it cannot rely on rules but must be AI - driven. And the most data can be learned from the human world.

Second, most scenarios in the world, such as homes and factories, are designed, built, and operated for human use. The more anthropomorphic a robot is, the easier it is to adapt to the world.

Third, from a purchasing perspective, anthropomorphic robots are more likely to make people feel a sense of affinity, which means they may sell well. Higher sales volume can lead to economies of scale, and with scale, costs will be lower, forming a positive cycle.

Question: What is the reuse ratio of components between XPeng's new - generation robots and cars?

He Xiaopeng: I don't have an exact answer, but many processes are the same. For example, perception and domain controllers, and 70% of the AI software is the same. However, the joints and skin of robots are not found in automotive components.

Question: What proportion does XPeng hope the revenue from the Physical AI business will account for in the total automotive revenue?

He Xiaopeng: The global automotive market is worth trillions of dollars, with 90 million cars produced annually. I personally think that the robot market will be worth 20 trillion dollars, but this may take 10 to 20 years and won't happen overnight.

The development of the automotive industry often follows a relatively low - growth linear curve because it is closely related to strict safety regulations and policies. However, once robots pass the technological and product inflection point, they will experience extremely rapid growth.

I haven't thought about how many robots we can sell in a year ten years from now, but I believe it should definitely exceed the number of cars.

Question: Many Robotaxi companies haven't made a profit yet. How does XPeng Motors ensure profitability when promoting its Robotaxi plan? How is its business model different from other Robotaxi companies? And how do you view the competition with them?

He Xiaopeng: XPeng may be a different kind of Robotaxi company. We have mass - produced cars with pre - installed systems. Our thinking is not technology - first. Instead, we consider whether what we do has commercial value, user value, whether it can reach a technological inflection point, and whether the government and society will accept it.

XPeng will also launch Robo intelligent driving version models for consumer sales, which can significantly share the BOM cost (a key cost component in the product design, procurement, manufacturing, and assembly processes of automotive manufacturers) and R & D expenses. In addition, XPeng's Robotaxi and XPeng cars can share BOM and R & D expenses, giving us a natural advantage of tens of percent or even several times over other companies in these two aspects.

Moreover, we don't need high - definition maps, street - scanning, or lidar. We think more like a person in the physical world, so our technology has broader applicability and generalization, and there is no need for deployment costs.

I firmly believe that in the future, four - wheeled vehicles will be a combination of shared and private use. I don't think all cars will become Robotaxis. In this case, XPeng chooses to provide a "toolbox" for Robotaxi, including cars, software, and SDK interfaces. We open up our capabilities, and partners from different countries and regions can buy our Robotaxis for operation.

Question: Why did XPeng choose to cooperate with AutoNavi for a global ecological partnership for its Robotaxi first?

He Xiaopeng: First, AutoNavi was previously under my management; it's my former employer. Second, AutoNavi is a very large travel ecosystem platform in China. So, it operates, and we provide the toolbox. I think this aligns with the strategic positioning of both companies.

Question: XPeng mentioned at the press conference that it will launch three Robotaxi models next year. What are the differences among these three models?

He Xiaopeng: They are priced differently. We have different Robotaxi models at different price points to meet the needs of different users. There are also 7 - seat, 6 - seat, and 5 - seat versions. For now, this is all I can disclose. These three models support different Robotaxi requirements.

Question: Why didn't the press conference mention L3 and directly aimed at L4? Is the progress of L3 products stuck due to laws and regulations?

He Xiaopeng: I believe that in the future, there will only be L2 and L4, no L3, because L3 is neither L2 nor L4.

Question: Has XPeng's Second - generation VLA (Vision - Language - Action) really completely removed the "L", or has it been converted into other forms of tokens?

He Xiaopeng: Ours is V + L. We didn't convert it into human language and format but into a new language for the physical world. It's not a language that humans can see and recognize, but it's very efficient and more comprehensive. Second, it can be decoded in the middle, and we can completely deduce the intermediate process. For example, we can figure out why it should have turned left but didn't. We've achieved this in the evaluation of the physical world model.

Question: At the XPeng Motors Technology Day press conference, Physical AI was applied to different carriers. What other ideas do you have for the application scenarios and carriers of Physical AI in the future? Also, as the Sino - US technological competition intensifies, which side do you think has more development advantages in the field of Physical AI?

He Xiaopeng: Companies that develop earlier will have a greater advantage. Today, in Physical AI, the First - mover advantage is very important, which was not seen in the past physical world or on Earth. In the future, data will be the most important guarantee. We've seen that the progress of many large - model companies in the digital world has slowed down in recent months. I believe the root cause is not the algorithm, the model, the computing power, or the electricity but the data.

In the field of Physical AI, I think everyone has a chance. The core is who can do it well first and who has good engineering capabilities. Good engineering capabilities mean good implementation, and after implementation, good user experience and better service are needed to form a huge cycle.

This is also why XPeng has never fully believed that simply making a car bigger, more beautiful, cheaper, and of higher quality can lead to victory. These are only necessary conditions, not sufficient ones. It is important to do a good job in hardware, which is the foundation, but software also needs to be well - developed. That's why I talked about Physical AI for nearly two hours at the Technology Day. How to organically combine the physical world and the digital world to meet customer experience, I believe, is just the beginning of the emergence of Physical AI.

Question: Tesla is also exploring flying cars recently. Tesla and XPeng may be the two companies with the highest business overlap in the world. So, what is the biggest difference between XPeng and Tesla?

He Xiaopeng: Both Tesla and XPeng have found a way to cross - domain innovation by changing dimensions. For example, when applying VLA to robots, the original robots used reinforcement learning for the foot joints. It was the optimal solution for the previous technology, but it's not good enough for the new technology.

I don't know exactly how Tesla is developing flying cars, but I think both are crossing boundaries by changing dimensions. Tesla conducts self - research on machines but promotes integrated innovation across domains.

Let me share with you the VLT (Vision - Language - Task/Thinking) of robots. Once the robot's brain matures, we'll apply it to the next - generation cars. You'll see that countless new logics will converge at a certain point, which will be really interesting.

Gu Hongdi: I think there are two points worth noting when comparing Tesla with us. One is the similarity: neither of us focuses on a very narrow niche but aims for mass - production scale. The difference is that XPeng's ecosystem is very open, which is different from Tesla.