An industry on par with new energy vehicles is on the rise.

As the biggest beneficiary of the AI era, NVIDIA CEO Jensen Huang's personal wealth has surpassed $100 billion, almost equal to the market capitalization of competitor Intel.

Recently, Huang traveled to Hong Kong University of Science and Technology to attend an honorary doctorate ceremony and said in a subsequent conversation that there are two major trends in the development of the AI industry: AI agents and robots.

Huang emphasized that while there may be many types of robots in the future, it is likely that only three types—autonomous vehicles, drones, and humanoid robots—will be able to achieve mass production.

Over a decade ago, companies had already developed robot waiters, robot chefs, etc. I have also experienced the service of food delivery robots. So why does Huang say that only three types of robots can be mass-produced?

The era of robots is approaching, but choosing the right path is essential for the future

Before the AI era, many global technology companies, university laboratories, and research institutions had already developed numerous robots. Huang's judgment that only autonomous vehicles, drones, and humanoid robots can achieve mass production may be due to the new definition of robots.

In the past, a robot was defined as a machine capable of semi-autonomous or fully autonomous operation. However, this definition was too broad, and almost any machine could fit into it. In the AI era, machines that can only perform simple tasks will be excluded from the new definition of robots.

(Image source: AI generated)

A new AI robot must first possess a sufficiently high level of intelligence and the ability to recognize complex scenarios, automatically identifying and making decisions based on the work environment.

The concept of embodied AI, which has gained popularity this year, aligns better with the new-era definition of robots. Embodied AI refers to integrating AI robots into physical entities, enabling them to possess dynamic capabilities such as perception, learning, and environmental interaction, and to perform certain physical operations.

Whether or not robot manufacturers mention the concept of embodied AI, their products are essentially developing in this direction. Currently, products classified as robots, such as cooking robots and food delivery robots, generally have low intelligence and poor perception capabilities and may be excluded from the new definition of robots.

Autonomous vehicles, drones, and humanoid robots are different. These products face many more use cases than cooking robots or food delivery robots. Taking autonomous vehicles as an example, intelligent driving and automotive robots share the same goal of navigating various road conditions, making decisions for different driving scenarios, and ensuring passenger comfort.

More importantly, the intelligence of automotive vehicles is a trend of the times. Nowadays, some cars priced at around $20,000 already come with intelligent driving features, and more vehicles will incorporate related intelligent functions in the future. The global sales of about 90 million automotive vehicles per year provide sufficient market space for the development of automotive robot companies.

(Image source: AI generated)

The same applies to drones, except that drones fly in the air and face relatively less complex scenarios. In terms of the market, the public's impression of drones may be limited to aerial photography, but in reality, drones have been widely used in pesticide spraying, power inspection, logistics, and other fields, with unlimited future prospects.

As for humanoid robots, they are more like a general term that may be further divided into household robots, security robots, and other categories in the future. Among the three directions mentioned by Huang, humanoid robots have the largest future prospects. Tesla CEO Elon Musk once posted on the X platform that he is very optimistic about humanoid robots, and their future sales will reach ten times that of automotive vehicles.

Just because these three types of robots have the potential for mass production does not mean that other robots cannot. Huang's viewpoint is actually directed at consumer groups, while the B-end scenario still provides ample development space for other robots, such as mining robots and port handling robots.

From the perspective of consumers, we are most concerned about the three types of robots mentioned by Huang, and companies have already entered these markets to explore development paths.

Tesla, Baidu, XPeng compete, but DJI is the king?

An automotive robot is, simply put, a vehicle that can intelligently perceive the state inside and outside the vehicle through sensors and adjust autonomously. At this stage, the functions of automotive robots that can be emphasized are mainly focused on intelligent driving. In fact, face recognition, seat memory, and voice interaction throughout the vehicle are all manifestations of automotive robot functions.

Currently, many new energy vehicles have incorporated features such as continuous voice interaction, multi-zone recognition, and AI graphic design. The ability to achieve high-level intelligent driving functions in urban areas has been brought to the market below $30,000 by XPeng's MONA M03. However, among mainstream new energy vehicle companies, only Jiyue Automobile explicitly brands itself as an automotive robot, leading consumers to be unclear about the difference between smart cars and automotive robots. Although there are certain differences in definition, they ultimately serve the same purpose.

In the field of drones, many companies worldwide are involved, but DJI dominates the industry. DJI drones have evolved through different eras, from manual control to AI obstacle avoidance and AI camera movement. Current DJI drones support multiple intelligent camera movements, further reducing the operational and photographic skills required of users.

(Image source: AI generated)

Aerial photography is only a preliminary exploration of drone application scenarios. For example, the State Grid Suzhou Power Supply Company recently put into use a self-developed drone multi-nest remote take-off and landing autonomous inspection system. This equipment supports multi-nest networking, remote take-off and landing, and visual positioning, and will achieve unmanned inspection of 75,000 kilometers of power grid lines in the future.

Compared to AI technology, the development and production of drone-type robots place more emphasis on industrial strength, capable of mass production and deployment, which happens to be a strength of domestic enterprises. In the field of drones, domestic industries and basic AI capabilities have matured, and future AI functions can be developed according to usage scenarios.

Finally, there are humanoid robots, which have the largest future prospects and the highest research and development difficulty. Domestic and foreign automotive giants attach great importance to humanoid robots. For example, during this year's World Artificial Intelligence Conference, Tesla showcased the second-generation humanoid robot Optimus, which was released at the end of last year. This robot is lighter and moves faster, and has already been deployed in Tesla factories to sort batteries.

(Image source: AI generated)

Another new energy vehicle giant, XPeng, is also committed to robot technology research and development and named its fourth-generation robot "Iron" earlier this month. This robot has also been deployed in automotive factories. In the short term, it will be difficult for humanoid robots to be sold to the consumer market. Instead, they will be first deployed in factories to replace some workers and verify their capabilities, which is indeed the optimal choice.

Overall, automotive robots, drones, and humanoid robots, although differing in their current status, are all developing at an extremely fast pace. Automotive robots and drones are strengthening their functions on one hand and expanding their market presence on the other. Due to their uniqueness, humanoid robots are expected to take several years before entering the consumer market.

The era of robots has already begun, and it is up to countless upstream and downstream enterprises to make choices.

AI large model technology brings new possibilities to robots

The arrival of the AI era has brought the concept of embodied AI to a comprehensive implementation in the consumer market, giving the robot industry more possibilities. Based on the concept of embodied AI, new robots, through enhanced perception, learning, and environmental interaction capabilities, can think as much as possible like humans, thereby completing more and more complex dynamic tasks.

Undoubtedly, the biggest beneficiary of the robot era is still NVIDIA. After all, training robots requires computing power, and robots themselves also need high-performance chips. The China Electronics Society predicts that the domestic market for humanoid robots will reach 870 billion yuan by 2030. This is just for humanoid robots and the domestic market. The global robot industry as a whole is likely to replace the automotive industry as the largest industry, leaving ample opportunities for other upstream and downstream companies.

A large number of domestic upstream and downstream enterprises are at a crucial decision-making moment. With the rapid development of the robot industry, they must make choices as soon as possible to avoid missing the wave of the times.

Among the three robot routes pointed out by Huang, drones have the lowest threshold, but the industry leader DJI is too powerful, making it difficult for ordinary companies to compete. Fortunately, DJI mainly operates in the mid-to-high-end market, leaving the low-to-mid-end market for smaller manufacturers to survive. Many drones that we can buy now for tens to a couple of hundred dollars are made by these manufacturers.

Humanoid robots belong to an emerging industry with almost no presence in the consumer market and a lack of leading enterprises, making them more suitable for ambitious companies to venture into. However, humanoid robot technology is relatively immature and may require significant capital investment, which may be unaffordable for small and medium-sized enterprises. It is advisable for them to find investors before entering the market.

(Image source: AI generated)

Automotive robots are undoubtedly the most difficult to enter. It's not because of the technical difficulty; on the contrary, in China's mature supply chain system, it is possible to handcraft automotive robots at any cost. However, the automotive industry is dominated by giants, including traditional automakers like BYD, Changan, Geely, Great Wall, and Chery, as well as new forces like Hongmeng Zhixing, Lixiang, XPeng, NIO, and Zero Running, all of which have a certain influence. There is no market space left for new brands.

In my opinion, for companies planning to enter the robot industry, if they have sufficient financial resources, they can try humanoid robots, and if not, they can enter the drone market and take the low-to-mid-end route. The era of robots is coming, and the industry is full of variables. The current situation in the industry may change drastically due to a single technological breakthrough.

Like the new energy vehicle industry, growing into a giant in an emerging industry inevitably involves stepping on the corpses of countless peers. All entrants should advance towards their goals while also considering the consequences clearly. Going bankrupt after failure is not terrifying; what is terrifying is becoming the next Jia Yueting.

Source: Leitech