Robots: A New Way Out for "Weixiaoli and Others" to Escape the Price War | Focus Analysis

Text | Tian Zhe

Editor | Su Jianxun

"Although it is not clear how to build robots, many automotive companies have established robot teams, and we must also keep up."

Automotive engineer Huang Ming recently received a task. His department must form a team to develop robots, but he has no detailed plan for the robot development cycle and specific uses. "It is generally in line with the industry progress."

Huang Ming doesn't have time to think carefully. He must quickly recruit experienced robot experts to join, because the pressure of talent competition comes not only from robot startups but also from internal competition - multiple departments of the company have started to develop robots including humanoid and robotic arms, "It's a bit like an internal race."

Since Tesla announced its self-developed humanoid robot in 2021, many automotive companies such as Xiaopeng, NIO, Chery, etc. have successively announced the dynamics of robot research and development and launched several robot products. Recently, Chery exhibited two robot products, a robot dog and a humanoid robot.

Chery Humanoid Robot

Another part of automotive companies is investing to lay out the robot territory in advance. Only in 2024, BAIC Group's BAIC Industrial Investment has already invested in robot startups such as Pasini Perception Technology and Galaxy General Robot. Among many robot startups, Zhiyuan Robot has received more favor from automotive companies, and its investors include well-known automotive companies such as BYD, BAIC, and SAIC.

A automotive company investor told "Intelligent Emergence" that the company has been paying attention to the robot industry since last year and has invested in several robot companies so far. When it comes to the reason, he said that the automotive factory scene is rich, and after the application of robots, it can accumulate rich data for the iteration of robots, but more importantly, The automotive industry is caught in a price war and is eager to find a new business growth point.

"Now if new energy vehicles do not reduce prices significantly, there may be no sales, but if the price is reduced too much, it will lead to losses for automotive companies, and at the same time, there will be no market voice. So we invest in robot companies, hoping that robots can become the company's second growth curve."

The initiator of the automotive price war is Tesla.

In January 2023, Tesla announced that the prices of its best-selling models Model 3 and Model Y would be reduced by 20,000 to 40,000 yuan to stimulate sales growth. The effect of the price reduction was immediate, and Tesla's sales increased by 12.6% month-on-month in the following month. Since then, various domestic automotive companies have followed suit and launched price reduction and preferential policies for new energy vehicles. The prices of some new energy vehicles launched by some automotive companies are even lower than those of fuel vehicles of the same level.

But the long-term price war has continuously compressed the gross profit margin of automotive companies. Tesla's annual report shows that its gross profit margin was 25.6% in 2022, while this figure was 18.2% in 2023, with a year-on-year decrease of 7.35%.

When "low price" is closely tied to sales volume, almost no player in the automotive industry can escape from the price war, and the more imaginative robot industry has naturally become a new battleground for the automotive industry.

Factory + Components, Opportunities for Automotive Players to Enter

A humanoid robot is roughly composed of three parts: the "brain" with a multimodal large model as the core, the "cerebellum" that controls the movement ability, and the body of the robot.

For automotive industry players who are good at precision manufacturing and intelligent driving, the "cerebellum" and the body are the best entry points for their layout of robots.

Tesla is the first automotive company to announce the research and development of a humanoid robot. In 2021, it launched its first humanoid robot Optimus, which is expected to be used in the Tesla factory first after small-scale production in early 2025. At the recent Tesla event, Optimus has the ability to dance and skillfully grasp objects.

The rapid iteration of Optimus' operational ability is due to its application in the factory. Public information shows that Tesla has equipped Optimus with a remote operation team of dozens of people. They must wear motion capture equipment and VR headsets to perform various designated tasks and collect AI training data.

Tesla Remote Control Robot

Remote operation is one of the ways for robot companies to collect real data, and its advantage is that after the collected data is used to train the robot, the success rate of the robot's related tasks will be improved.

An industry insider said that the automotive factory has diverse production links and clear requirements, which can collect a large amount of real data for humanoid robots. Therefore, the automotive factory has become the preferred scene for the landing of most humanoid robot products. For automotive companies with factories, the research and development of robots will have an inherent advantage in data collection.

On the other hand, the large-scale mass-produced automotive components applied to humanoid robots can reduce the development cost of the robot body.

According to "The Biography of Elon Musk", in order to make the wrist of Optimus have three degrees of freedom while maintaining a low manufacturing cost, the team applied the liftgate actuator of the Tesla car to Optimus, thereby reducing the cost by hundreds of dollars.

The humanoid robot industry is in the early stage of development, and automotive component suppliers may supply components for robots. Zhu Hui, the product director of Galaxy General Robot, told "Intelligent Emergence" that because the components are not standardized, the current humanoid robot products on the market are basically "Prototype 0", and the requirements of industry players for some non-core components are that they can be used.

Some automotive component suppliers have targeted the robot track. For example, well-known automotive first-tier suppliers such as Inovance Technology, Tuopu Group, and RoboSense have stated that they will develop core components of humanoid robots.

"Intelligent Emergence" has learned that some intelligent driving companies are also paying attention to the dynamics of humanoid robots and investigating the possibility of entering the humanoid robot industry.

The intelligent driving technology has been iterated for many years, and its technical maturity has been verified by the market. According to the statistics of the China Intelligent and Connected Vehicles Industry Innovation Alliance, in 2023, the sales of intelligent connected passenger vehicles with combined assisted driving functions were 9.953 million, and the market penetration rate reached 47.3%.

In recent years, the end-to-end large model that has been successfully verified by intelligent driving companies is showing a trend of being applied to humanoid robots. 

The end-to-end large model aims to integrate the traditional perception and control modules of intelligent driving into a unified model. Only by inputting the original data can the final trajectory or control instructions be output, thereby greatly improving the anthropomorphism and universality of the intelligent driving system. If the end-to-end large model is applied to robots, it can improve the universality and flexibility of robots in different operating tasks. 

The industry has begun to use the visual language large model for intelligent driving to solve the problems of scene understanding and planning.

In the view of Sui Wei, the algorithm technical director of Digua Robot, although the application environments of robots and automobiles are different, and the environment that robots face is more complex and diverse, the universality of the large model can also solve the problem of scene understanding of robots. Using the visual language large model for the scene understanding of robots can help improve the adaptability of robots in different environments.

The Robot Industry Has a High Threshold

The components of humanoid robots are not standardized, opening the door to automotive component companies, but not all players can enter smoothly. 

In Sui Wei's view, the differences between new energy vehicles and humanoid robots will cause the supply chains of the two to diverge in different directions. 

First, humanoid robots use more joint technology, and the sensors used in joints will not be strictly consistent with intelligent driving. Therefore, robot companies have more types of sensors to choose from. 

In terms of core components such as motors and actuators, many robot companies improve the upper limit of the robot's ability through self-research, which affects the relevant component market to a certain extent.

An robot practitioner told "Intelligent Emergence" that because components have a strong impact on the subsequent iteration of robots, robot companies are relatively cautious when purchasing components. In his view, when automotive component companies enter the robot field for the first time, the product maturity is difficult to guarantee, so the probability of cooperating with automotive supply chain companies is low.

Second, although intelligent driving technology can be reused for humanoid robots, the application scope is relatively limited. 

Zhu Hui said that because both new energy vehicles and wheeled robots travel through tires, some algorithms of intelligent driving can be applied to wheeled robots. But the difference between the two is that wheeled robots have leg joints and a relatively higher degree of freedom than automobiles. 

In addition, wheeled robots can be subdivided into types such as two-axis and four-axis, with large differences in degrees of freedom and overall balance.

In fact, for automotive companies that are good at manufacturing and integrating the supply chain, it is not easy to develop robots.

Elon Musk once said that by the end of this year, the dexterous hand of Optimus will double from 11 degrees of freedom to 22, but the recent Tesla event shows that the dexterous hand it displayed has not been significantly upgraded. It can be seen that even Tesla, which has rich manufacturing experience and invests a lot of costs, the research and development of humanoid robots is not smooth sailing.

The Dexterous Hand Displayed by Tesla

Nevertheless, Musk remains confident in the robot business. He expects that the humanoid robot business will exceed the sum of Tesla's other businesses and create a market value of tens of trillions of dollars for Tesla.

There is still a distance from the mass production and commercial application of humanoid robots, and everyone is on the same starting line. This is one of the reasons why automotive players are scrambling to enter. They all believe that they can get a share of the market cake. At least, the price war will not be repeated in the robot industry for the time being.