At the 2025 Tencent ConTech Conference and Tencent Technology Hi Tech Day, He Xiaopeng, the founder of Xpeng Motors, put forward a thought-provoking view: smart car companies are becoming explorers of the "physical AI world".
This is actually not the first time he has said this. Just a month ago, at the "Xpeng Technology Day", he announced: Xpeng is no longer just a car-making company, but will become a "embodied intelligence" company for the world - to put it simply, it is to let AI not only think, but also act in the real world, such as through robots.
In fact, Xiaopeng is not the only one who is taking this path. As of December 2025, about 18 car companies around the world have announced their entry into the field of humanoid robots, including Tesla, Xiaomi, GAC, BYD, Hyundai, Toyota and other major manufacturers.
Why are car companies making robots across borders? He Xiaopeng believes that there are two reasons:
Technology connection: building smart cars and robots both rely on core technologies such as AI, perception systems, and automatic control; The goal is the same: The ultimate goal of AI is not just the virtual world, but to really "move" in the physical world.
He used the "ant colony effect" as an analogy: just like individual ants are simple, but group collaboration can complete complex tasks - in the future, a large number of agents (cars, robots, equipment) will work together to completely change the pattern of the intelligent era in which we live.
Why are more and more automakers starting to make robots?
He Xiaopeng, founder of Xpeng Motors, believes that this is almost a "natural" thing - because building smart cars and robots uses the same set of technical foundations.
1. Technology is inherently connected
In the past decade, many car companies have developed a complete set of technology systems: from AI chips and perception systems to modeling and upper-level applications of the physical world. This set of capabilities can be used on robots without having to explore from scratch.
2. The hardware is highly similar
For example, Xpeng's latest robot uses three self-developed "Turing AI chips" with a computing power of up to 2250TOPS. The research and development experience of this high-performance chip comes directly from the technology they accumulated when doing autonomous driving. To put it simply: the "brain" for the car can also be put into the robot with a little adjustment.
3. Software algorithms can also be reused
Autonomous driving should identify road conditions, plan routes, and avoid obstacles in real time; Robots also have to move around and complete tasks at home or outdoors. The "thinking logic" of the two is very close, and the algorithms can learn from each other or even directly transfer them.
4. Manufacturing capacity is readily available
Car manufacturing pays attention to large-scale production, strict quality control and mature supply chain - these are what robot mass production needs most. Car companies use their experience in car manufacturing to robots to quickly achieve high-quality and low-cost mass production.
| The next flashpoint of AI is hidden in the "real world"
If car companies can make robots based on technical foundation, then what they really want to do is to let AI go out of the virtual world and into the physical world we live in.
He Xiaopeng believes that in the past 30 years, AI has mainly developed in the "digital world" of computers and mobile phones, relying on a large amount of text, pictures and other data. But today, the growth of this type of data has slowed.
And the real world - where we live and work every day - hides a larger and richer treasure of data, which is the key to the next breakthrough of AI.
Smart cars are the starting point, and robots are the extension. Smart cars are already an important "eye" and "brain" of AI in the physical world, but it mainly solves the problem of "how to drive". Robots can go even further: it can clean at home, serve in shopping malls, work in factories, and come into contact with various real-world scenarios.
Why is data in the physical world so important?
Today's AI models mostly rely on language and text, but human intelligence is not just about "speaking" - it also includes multi-sensory experiences such as seeing, listening, touching, and moving.
Robots can collect this "multimodal" data, such as:
The touch of the hand touching the object Spatial perception as you move around the room Feedback on actions when completing tasks
Robots are equal to the "data collectors" of the physical world. He Xiaopeng gave an analogy: robots are like the "tentacles" of AI reaching out into the real world. While they work, they transmit massive real-world information back to AI systems, helping models become smarter and more realistic.
When agents learn to "think for themselves", what will the world become?
He Xiaopeng, the founder of Xpeng Motors, proposed a very imaginative concept - the "ant colony effect". He believes that this will be the new rule in the future era of intelligence.
What is the "ant colony effect"? Imagine ants: they don't have a "commander-in-chief" but can collaborate efficiently; When an ant finds food, it will immediately notify nearby companions and move back together. This model of decentralization, autonomous decision-making, and mutual cooperation is the "ant colony effect".
In the AI era, smart cars, robots, aircraft, etc. will become "agents". When they can perceive their own environment, make judgments, and cooperate with each other, they will form a powerful and flexible system like an ant colony.
Why do car companies build robots? The answer is here.
In the past: the industrial era relied on "scale" (the more you built, the cheaper); The Internet era relies on the "network" (the more users, the greater the value).
Future: The intelligent era relies on the "ant colony effect" - a large number of agents work together to create greater value.
And smart car companies are already standing at the starting line: the car itself is the largest mobile agent, which can collect data, provide computing power, and cover various scenarios; new agents such as robots and flying cars can be expanded to more places such as homes, shopping malls, and factories; They can share data, exchange technologies, and complement scenarios to form a complete "intelligent ecosystem".
In the future, you are not facing a car or a robot, but the ambition of a "smart partner network" Xpeng: not to be a "car company", but to be an "explorer of the physical AI world" For companies like Xpeng, building robots and flying cars is not "not doing the right thing", but taking advantage of the trend:
The technology is inherently connected (hardware, software, and AI are all reused); Highly consistent goals (let AI land in the real world); The trend is unstoppable (the "ant colony effect" is coming).
The future competition is no longer "whose car is better", but "whose agent ecology is stronger".
When autonomous driving becomes popular, robots enter thousands of homes, and flying cars take to the skies, every aspect of our lives will be redefined. The smart car company that took the lead in the layout is transforming from a "travel service provider" to a core force in building a physical AI world.
