Humanoid robots are facing a situation where everyone is piling on when they're down. If they don't want to end up as just big toys, what should "people like Wang Xing" do next?

In recent days, the registration channel for the World Humanoid Robot Games, which will be held at the Bird's Nest and the National Speed Skating Oval, has been opened. Its high - profile nature has quickly sparked heated discussions, and many people are expressing their anticipation.

In 2024, humanoid robots were just learning to walk. Less than half of this year has passed, and several robot - related events have already gained widespread attention. In April, robot marathons held across the country attracted countless eyes. A few days ago, a humanoid robot fighting competition hosted by CCTV once again ignited the public's curiosity and enthusiasm for robots.

The popularity in the robot field has not waned since 16 Unitree robots wearing traditional Northeast - style flowered jackets "danced" on the Spring Festival Gala stage. Meanwhile, the sales market for humanoid robots is also booming.

"The company's orders are scheduled until May and June. The production line is operating at full capacity, and each robot needs to be delivered immediately after coming off the production line," Wang Song, the general manager of Leju (Suzhou) Robot Technology Co., Ltd., revealed in a recent interview with CCTV News. It is reported that Leju delivered 250 orders in the first quarter of this year, directly completing the half - year task ahead of schedule.

Wang Xingxing, the founder and CEO of Unitree Technology, whose name is strongly associated with cutting - edge robot technology, publicly stated at a recent external event that Unitree Technology has a full order book and is short of talent. "We are in great need of people in all positions, including clerical, procurement, sales, technology, R & D, and marketing."

This year, several humanoid robot companies have publicly announced that they are overwhelmed with orders. On major e - commerce platforms, robot products are not only selling well and even running out of stock at times, but also come in a wide variety. On the JD platform, there are currently multiple robot products for sale, including Songyan Power N2, Zhongqing PM01, Unitree G1 and Go2, Yun Shenchu Jueying Lite3, Weilan Alpha, Collibot Loona, and Ubtech alpha Ebot. Among them, 238 people have pre - ordered the Songyan Power N2, and the product is expected to be delivered 90 days after payment.

As more robots come into the public view, more user feedback has emerged, and there is more negative feedback. In the robot half - marathon, the "spectacular falls" of various robots strongly contrasted with the official promotional videos. On social platforms, many individual buyers have successively uploaded videos of robots running out of control and hitting the wall, having shortened battery life, and having no access to after - sales service.

After leaving the spotlight, can these "steel men" with mechanical arms stand firm in the future? This question is spreading among the public, and the "golden sponsors" who invest in robot companies also seem to be affected in their judgment.

Is the Name "Big Toy" Really Justified?

"Currently, robots are undoubtedly just 'big toys', and they may not even be that fun,"

Xiao Yiting, a partner at Chentao Capital, told us that the "falls" currently seen are exactly the same as the current situation of robots she is aware of. The public's perception of robots often comes from Spring Festival Gala performances or pre - rehearsed demonstration videos, which are actually beautified versions. In fact, many robot companies operate demonstrations or videos as a project - through a large amount of pre - preparation in the early stage and multiple recordings of error states, they can finally present an ideal effect, which is inherently different from the normal state of robots.

It is revealed that many robot companies even spend a whole month preparing for a single demonstration. This is similar to early - stage autonomous driving: running on the road to be demonstrated many times in advance, dealing with all situations, and then presenting it.

"In our view, the current actual level of robots is just like this," said Xiao Yiting. The "falls" in the events are normal, and they even found some bright spots in them. "Taking the robot marathon as an example, we think the performance of several participants was quite good."

Therefore, such events will not affect the valuation judgment of senior investors on relevant robot companies. "The public may have an overestimated and wrong perception of robot technology due to demonstration videos, but investors do not have such cognitive biases. At least among the investment community focusing on the field of embodied intelligence, our current perception of the development status of robot technology is always consistent with the level exposed by those falls. As I observe, most of the people who finally invest in robot companies are relatively professional."

However, the trend in the individual trading market has changed drastically recently. Not long ago, there were frequent sighs from individual buyers on social platforms about "being unable to rent out robots". The Unitree robots, which once sparked a boom in the rental market with a daily rental price of over ten thousand yuan, now not only have a lower daily rental price but also seem to have far fewer orders.

Xiao Yiting pointed out that this has little to do with the performance of Unitree Technology's products but is a manifestation of market rules. "Buying robots and then renting them out" was a highly profitable model some time ago. From the perspective of the attributes of demonstration - type products, renting is indeed more cost - effective than buying. But the standard business logic is always the same. When the whole world knows that a certain model is profitable and has a low threshold (for example, the cost of buying a Unitree robot is relatively controllable), the business will inevitably not be as good.

"When everyone buys robots to rent out, there are only so many customers. You either have to lower the price or you won't be able to rent them out."

Who is Spending Money on Robots?

According to currently publicly disclosed information, the prices of various humanoid robots currently range from 39,900 yuan to 650,000 yuan.

So, from the time when robots became well - known after the Spring Festival Gala to now, who are the people willing to spend money on robots?

Generally speaking, when a new technology emerges, it first enters these markets: first, the education field. This type of application does not emphasize functionality but focuses on providing educational value to students; second, demonstration scenarios. Enterprises will spend a lot of time on this, which can bring them some traffic; third, early adopters. These users buy out of love for high - tech, mainly to satisfy their emotional needs rather than considering the cost - benefit ratio of replacing human labor.

"In the early stage, the market for most such technologies is like this, and its sustainability often comes from a large number of early adopters."

In Xiao Yiting's view, the current robot market will remain relatively small, as it belongs to the category of luxury consumption and can only meet people's needs for demonstrations or experiencing new technologies. Although the market volume is small, it will continue to exist because robots can do more things every year, and there will always be new early adopters.

"Only when it can replace human labor and has an advantage in the cost - benefit ratio and becomes a household - usable product will a large - scale market be formed."

Xiao Yiting mentioned that the evolution from a small market to a large market is a common situation after the emergence of new technologies. Take autonomous driving as an example. In the early stage, scenic spots introduced autonomous shuttle buses, but safety officers still needed to be on board. This model was very suitable for tourists to experience; the autonomous cleaning vehicles deployed in residential areas were also popular because they seemed high - end. These all belong to the long - term development stage of the technology. It really takes a very long process to go from this early situation to truly achieving work replacement.

What Do Current Robots Lack?

In the recent robot marathon, apart from the performance of the robots, a scene has sparked national discussions: behind each robot, there was a sweating engineer. Some were urgently adjusting the robot with a remote control, some were carrying heavy battery boxes, and some were frantically reinstalling the robot's "head". Many people felt that "the technology is useless". What's the application value of a robot that needs human accompaniment throughout?

"The ultimate goal is to have robots do work, whether at home or in the factory. But it's not very realistic in the next few years to have a humanoid robot or an intelligent machine start working at home immediately. Before they can really do work, we hope to make more commercial expansions and attempts with the technological achievements we've already made," Wang Xingxing said when talking about various robot events at the Zhiyuan Conference in early June.

Xiao Yiting pointed out that events like the marathon are of a performance nature and may require human accompaniment. But in B - end work scenarios such as new retail, logistics sorting, and production line operations, if human accompaniment is still needed, it can't be said to reduce costs and increase efficiency - because the payback period for robots to replace human labor is calculated based on human wages and costs. So the application in such scenarios must not rely on human accompaniment.

"We believe that robots will be able to achieve the corresponding independent working ability within three to five years."

At the same time, regarding the current development status of robots, she also said bluntly that it is still in a very early stage, and there are many problems in practical applications. "If we make an analogy, current robots are a bit like autonomous driving in 2016 - 2017."

It is reported that many technical solutions for the hardware of robots have not been finalized, and aspects such as service life and battery life are still gradually developing. In addition, the instability of the physical structure also means that the algorithms cannot function 100% effectively after deployment.

In terms of software, the algorithm iteration is also in the early stage. On the algorithm side, companies have not yet achieved the so - called "embodied intelligence". The precision of robot operations and the hardware configuration cannot match the corresponding scenarios, or they can't fully replace human labor. "The core is the data problem. Due to the high difficulty of data collection, the problem of insufficient data volume for embodied intelligence has not been completely solved. Insufficient data can't train robots well. The popular demos we see are actually presented through a large amount of training."

Currently, the scale of robot data collection is gradually expanding. There are many ways to increase the data collection volume. Many enterprises use synthetic data in simulation environments. The continuous improvement of simulation technology can provide better support for the iteration of robot algorithms. "As the data collection volume continues to grow in the future, the robot algorithms will also become more and more complete."

Not long ago, Wang Zhongyuan, the director of the Zhiyuan Research Institute, also told us that the current data limitation makes it impossible to achieve the integration of the "big brain" and "small brain" of robots. In the next 5 - 10 years, the model of this integration may mature.

When talking about the current after - sales mechanism of robots, Xiao Yiting said that the core problem is that robots are not yet mature products and can't do as many things as people imagine. If users buy an immature product, they naturally can't expect it to achieve the usage effect of a mature product. For example, you can't expect a robot to cook, do laundry, and sweep the floor. However, if a robot has basic problems such as running out of power, being unable to charge, being unable to move, or crashing, the after - sales service usually provides solutions. If users ask the robot to perform functions not marked in the instruction manual, the after - sales service naturally can't meet the demand because this is not a product failure but a difference in the perception of the product's positioning.

In the future, there may be two paths for the after - sales mechanism of robots: similar to the after - sales service provided by the seller for sweeping robots, or similar to the after - sales service provided by the robot manufacturer like a car factory.

"Everyone Will Own One or Two Humanoid Robots in the Future"

"Humanoid robots are definitely the ultimate solution in the robot field," Xiao Yiting pointed out. The core reason is that the infrastructure in the world we live in is designed for humans. From the height of kitchen countertops and the location of sinks to the steps and stairs in buildings, countless houses and public places around the world have been built according to human physical characteristics. If non - humanoid robots are to adapt to these environments, large - scale infrastructure transformation is required, and the cost of such transformation is much higher than designing and mass - producing humanoid robots. Therefore, the humanoid design of robots is essentially the optimal solution for them to adapt to the existing world at the lowest cost.

As for when humanoid robots will become a reality, her judgment is that the final realization of humanoid robots may take 10 to 15 years. Sam Altman, the CEO of OpenAI, holds a more optimistic view. In a recent interview, he said that in 5 to 10 years, we will have great humanoid robots that will walk on the streets and do various things.

For long - term capital, such a time span is definitely worth investing in. "Take autonomous driving as an example. When the industry emerged in 2015 - 2016, many people thought that autonomous driving would be realized within a few years, but we judged at that time that it would take 15 years. From the current situation, this time estimate is quite accurate. By analogy, it usually takes about 15 years for a new technology to really be implemented after it emerges."

More specifically, Xiao Yiting believes that humanoid robots can be implemented in some specific scenarios within 3 to 5 years and are expected to be applied in more general scenarios after 10 to 15 years.

"The development direction is very clear." According to her introduction, robots may first be implemented in scenarios with simple actions, specialization, and controllable scope, such as pharmacies in the new retail field, logistics sorting, and the last - mile delivery. Currently, the implementation of the Galaxy General Wheeled Upper - Limb Robot is progressing relatively fast. The company has carried out pilot work in the Beijing area, deployed robots in pharmacy scenarios, and plans to further expand the deployment scope and put them into use in more stores by the end of the year.

Next, robots will gradually start to undertake some work on the production line that replaces human labor, such as production line work with relatively low precision requirements. Take Ubtech as an example. It has reached a cooperation with leading domestic and foreign automobile enterprises to deploy humanoid robots in automobile factories, but it has not been implemented yet.

In addition, inspection, the military field, and some pre - arranged performances are also key directions for early implementation and have the basis for implementation. Currently, the main use of robot dogs is for inspection. Some leading companies have already deployed them in specific scenarios, but they have not yet entered the profit - making stage.

At the same time, Xiao Yiting firmly believes that humanoid robots have huge market potential, which is a consensus in the investment circle and there is no disagreement.

"In the future, everyone may own one or two robots, one for work and one for housework, which means the robot market will be extremely large." When talking about the future purchase model, she speculated that large - scale robot manufacturers may be similar to current automobile factories, and people may also buy a robot by loan or full payment. And there's no need to worry about maintenance issues, as large enterprises will provide guarantees.

How Many Companies Will Survive?

"From a ten - to fifteen - year time perspective, we are very confident in the technological development of robots, but in the short term (such as one year), the progress will not be as rapid as people imagine," Xiao Yiting said.

In the investment circle, there is a well - known Gartner curve, which reveals that after the emergence of any revolutionary new technology, it often first experiences a bubble - that is, an "over - optimistic period": in the first one or two years, people often overestimate the implementation speed of the technology, followed by a sharp decline in valuation after it peaks, and then it enters a slow - rising stage and gradually recovers within the next 10 to 15 years.

Xiao Yiting said that the definition of the "over - valued period" essentially depends on whether the technology can ultimately succeed. Even if you enter the market at the peak, if the technology can be successfully implemented, you may not necessarily suffer losses. From the perspective of the technology development cycle, it usually goes through an over - valued period, a disillusionment period, and finally enters a stable growth period - this is almost a common rule in fields such as the Internet, autonomous driving, and robots. The reason is that people often overestimate the progress of technology in one year but underestimate its potential in ten years.

"We need to maintain a neutral attitude: neither blindly chasing after the breakthroughs in one year nor easily denying the possibilities in ten years. Some people now claim that 'robots won't work for ten years' just because they see the so - called falls, while others firmly believe that 'robots will mature within one year'. Both of these views are too extreme."

Xiao Yiting believes that it may still take ten to fifteen years for general - purpose robots to mature.

It is reported that her current investment logic mainly focuses on specific niche scenarios. For example, logistics sorting in the new retail field may be the first to see the implementation of robots. Robots in such scenarios mostly perform single and fixed actions, and the technology generalization level is low. They will be implemented earlier than household service robots. Just like in the early stage of autonomous driving, it was first applied in specific scenarios such as mines, ports, and last - mile delivery, and then cross - scenario algorithm companies and enterprises in niche supply - chain fields gradually emerged. The robot industry may also follow a similar development path.

"In investment selection, we won't be limited to specific scenarios but will weigh the two directions of 'a large - scale market but slow implementation' and 'a small - scale market but fast implementation'."