Intro: Automakers are meeting each other again on another track.
The automotive industry is undergoing a somewhat familiar transformation.
Previously, real estate companies pivoted to computing power, data centers, and AI; now it is the turn of automakers.
Xpeng is making robots and secured over USD 900 million in funding in August; SAIC Motor directly deployed humanoid robots into its battery production line; GAC Group established an independent robot company; Changan Automobile scheduled its robot mass production for 2028; and Chery has already delivered over 2,000 robots.
According to incomplete statistics, as of September 2026, 14 major domestic automakers have explicitly laid out plans for humanoid robots, ranging from Xpeng, Li Auto, and NIO, to Changan, SAIC, and Chery, and then to GAC, Geely, BAIC, and FAW.
Automakers are meeting each other again on another track.
The question is, why do automakers suddenly all want to "build humanoids"?
The answer is actually not complicated: it is becoming increasingly difficult to make money in the automotive business.
Data from the National Bureau of Statistics shows that from January to July 2026, the operating revenue of the automotive manufacturing industry above designated size reached CNY 6.08 trillion, a year-on-year increase of 2.7%. However, the total profit was only CNY 216.24 billion, a year-on-year decrease of 20.4%, with an operating revenue profit margin of about 3.6%. During the same period, the profits of industrial enterprises above designated size nationwide grew by 17.6%.
Data from the Ministry of Industry and Information Technology (MIIT) indicates that from January to July, national automobile sales reached 17.602 million units, a year-on-year decrease of 3.7%. Although new energy vehicles (NEVs) continued to grow, the traditional automotive industry as a whole has entered a stock competition.
In September this year, the People's Daily directly pointed out this change: in the first-half financial reports of automotive-related enterprises, some original equipment manufacturers (OEMs) experienced "revenue growth without profit growth." Meanwhile, profit performance in upstream segments such as batteries, chips, and intelligent driving solutions varied significantly.
The automotive industry is increasingly resembling a typical capital-intensive, low-margin business—sales volumes remain, but profits are gone.
Consequently, automakers have unanimously turned their attention to robots.
Automakers' Robot Myth
On the surface, automakers making robots is somewhat similar to real estate companies suddenly venturing into computing power.
A real estate developer and an automaker suddenly talking about AI and robots—it looks like they are both searching for a "second growth curve."
However, the connection between robots and automobiles is actually much closer than that between real estate and computing power.
What smart cars do, simply put, is: see the world, understand the world, make decisions, and execute actions.
Humanoid robots operate on the same principle.
Cars use cameras, LiDAR, and millimeter-wave radar to perceive the environment, using chips and models to determine what is ahead, and then deciding whether to accelerate, brake, or steer.
Robots, on the other hand, have a different body.
They use cameras and sensors to see, AI to understand the environment, and then decide to raise a hand, grasp, walk, or avoid obstacles.
Therefore, in March 2026, when reporting on the rush of automakers into humanoid robots, People's Daily Online summarized this relationship directly: the layout of humanoid robots by automakers essentially involves massive technological synergy.
This is also why automakers dare to charge ahead.
They already hold some of the hardest things for robot companies to make up for: supply chains, manufacturing capabilities, intelligent driving algorithms, computing power, real-world scenarios, and even data, such as Xpeng.
Unlike many companies, Xpeng does not just feed robot models with car data; it directly integrates the automotive AI system with robots.
The company previously explicitly stated that robot and intelligent driving R&D have synergistic effects, and planned to let robots use the same physical AI foundation as cars. In August this year, Xpeng's robot business completed its first round of financing of over USD 900 million, with a post-investment valuation of over USD 6.3 billion; in September, it officially launched its robot production line.
This is no longer "an automaker casually researching robots on the side," but treating robots as a genuine new business.
But there is a point easily overlooked by the market—car data does not equal robot data.
This is both the biggest advantage and the biggest misconception for automakers rushing into robotics.
Driving data primarily addresses road perception, driving decisions, and vehicle control.
What robots truly lack is another type of data: I see a cup, how do I reach out? After my hand touches the cup, how much force should I use? After grabbing it, how do I move it? What if I encounter an obstacle while moving? What if the object slips?
These are all "manipulation data."
In April, People's Daily Online cited relevant research reports pointing out that in 2025, the number of humanoid robot whole-machine enterprises in China exceeded 140, with shipments reaching 14,400 units, accounting for 84.7% of the global total; however, the industry is still in the stage of advancing from technology verification to industrialization.
Therefore, the truly advantageous automakers are not those with "many cars," but those that can integrate car data, robot data, models, and real-world scenarios. Among them, SAIC Motor is a very interesting example.
In March 2026, "Nengzai No. 1" officially entered the mass production line for Buick Zhijing E7 batteries, taking on tasks such as cell grabbing and loading. SAIC Motor stated that this is among the first batch of humanoid robot application cases in the Chinese automotive industry to truly enter mass production lines.
It is a completely different matter for a robot company to buy dozens of robots for its own training, compared to an automaker throwing robots directly into a car factory that produces every day.
The former is conducting tests, while the latter is building a data closed loop.
The Crossroads for Automakers
It is precisely the real business scenarios brought by factory production lines that complete the critical leap from prototype testing to a data closed loop.
But even with the unique soil for implementation at hand, it does not mean that all entering automakers can naturally achieve the commercialization of humanoid robots.
The supply chains, manufacturing, intelligent driving algorithms, and road data accumulated in the automotive track have only laid the foundation. To truly build robots, another set of capabilities such as body control, dexterous manipulation, and vertical scenario adaptation needs to be supplemented. The cards held by different automakers vary, and their development paths diverge accordingly.
The first category is the "software + AI + robot" type, like Xpeng.
Xpeng's advantage is not its automotive scale, but that it pushed itself towards being an AI company very early on.
In the first quarter of 2026, Xpeng defined itself as a physical AI company covering smart cars, autonomous driving, flying cars, humanoid robots, and AI chips; its robot mass production base covers an area of about 110,000 square meters, planning to achieve mass production by the end of 2026 and enter stores and commercial scenarios in 2027.
By September, the robot production line was officially launched.
In other words, Xpeng has completed the first closed loop of "model—robot—scenario—manufacturing."
The second category includes traditional automotive giants like SAIC Motor, GAC Group, and Changan Automobile.
Their biggest advantage is their factory and industrial organization capabilities.
SAIC Motor has already sent robots into mass production lines while making industrial investments around robot bodies and core components.
GAC Group went a step further. In February 2026, GAC officially established Huilun Technology, transforming the robot business from an internal group R&D project into an independent industrialization entity; in August, Huilun Technology completed a financing of over CNY 100 million, continuing to promote the scaled commercialization of whole humanoid robots and core components.
Changan Automobile explicitly proposed to achieve mass production of humanoid robots in 2028, deploying them in scenarios such as factories, stores, and homes.
Their advantages are not as sexy, but very practical: they have factories, supply chains, workers, customers, and after-sales systems.
Once robots are truly deployed on a large scale, these things will become important.
Besides the three categories mentioned above, Chery is particularly worth mentioning separately because it is not simply betting on the "general-purpose humanoid robot" route, but focusing on the commercialization of robot products.
In April 2026, Chery's MoJia Robot completed the signing of a 1,000-unit order and the centralized delivery of 100 units; in July, the global cumulative delivery of MoJia Robot exceeded 2,000 units, with products and services covering more than 60 countries and regions.
These 2,000 units cannot be simply understood as 2,000 humanoid robots, because MoJia's product matrix itself includes multiple categories such as humanoid robots, smart police robots, medical guide robots, and robot dogs.
But this is still very important because it proves one thing: robots do not necessarily have to wait until "general-purpose humanoids" mature to make money; cutting into vertical scenarios such as security, medical guidance, industry, and services first is also a viable path.
This actually poses two different questions for automakers: do they want to be a "robot company," or just run a "robot business"?
These two are worlds apart.
Knowing How to Build Cars Does Not Mean Knowing How to Build Robots
Some might ask: can any automaker build robots?
The answer is obviously no.
One of the most common illusions in the robotics industry is: having built cars, we can also build robots.
This does not hold true.
Automobiles can distribute complexity through platformization, component standardization, and mature supply chains.
The biggest challenge for humanoid robots at present is precisely the "last meter"—hands, feet, joints, dynamic balance, dexterous manipulation, and model generalization.
This is also why insiders in the automotive industry are very cautious.
The Dongfeng R&D team publicly pointed out that the brain of humanoid robots is still evolving rapidly, and the technical route has not truly converged; at the same time, the industry lacks massive, high-quality real-world scenario data, and issues regarding safety, reliability, whole-machine cost, and work efficiency still need to be resolved.
This means that several types of automakers will be in danger.
One type has only concepts and no robot R&D system. Releasing a robot, shooting a video, or attending a robot conference does not equal having a robot business.
Another type has only manufacturing capabilities and no AI capabilities. Such companies can "assemble" robots, but may not be able to make robots "learn."
There is also a type that has only car data but no manipulation data. Millions of cars can teach autonomous driving models how to drive, but cannot automatically teach robots how to screw bolts.
Therefore, in the future, the companies that might truly be eliminated in the robotics industry are not necessarily those without automotive manufacturing capabilities, but rather those that equate "automotive supply chain advantages" with "general-purpose robot capabilities."
Entering the game is just the first step; the real difficulty lies in the next ten years.
Migrating the automotive AI capabilities over, migrating the automotive supply chain over, turning factories into training grounds, and then selling the robots.
If any link in the middle fails, the second growth curve will turn into a second R&D black hole.
Xpeng is proving whether it can become a "robot + car" company; SAIC Motor is proving whether industrial capital, car factories, and robot companies can form a closed loop; GAC Group is proving whether robots can transform from group laboratories into independent companies; Chery is taking a more direct approach, seeking commercialization first from products and scenarios.
As for a large number of automakers such as Changan, Geely, NIO, Li Auto, and Dongfeng, the real test has already been laid on the table.
This race will not ultimately reward the automaker that "released a robot first," nor even necessarily the one with the "most human-like robot." It will likely reward a different capability: who can make robots actually go to work.
In the first few decades of the automotive industry, the competition was about engines, then batteries, then chips, software, and autonomous driving. Today, what automakers are beginning to compete for may be a new industrial entry point—putting together the robot's body, the AI's brain, and the automotive manufacturing system.
Therefore, this time, automakers making robots is truly not exactly the same as real estate companies transforming into computing power.
Real estate transforming into computing power is about wanting to leave their old world; automotive transforming into robots is somewhat like continuing to move forward along their old world.
Only at the very end will they discover that automobiles can teach robots how to be manufactured, but may not be able to teach robots how to live, and this is the true watershed after automakers collectively rush into robotics.