The Trillion-Dollar Robot Market: Who Can Represent the Future, Unitree or Figure?

Unitree: Being the Leader Is Tough

Some investors point out that this year is the first year of commercialization of humanoid robots. Those without commercial mass - production capabilities and self - financing capabilities are no longer worth considering.

This statement is not an exaggeration. Currently, leading and mid - tier players are either preparing for listing early or have multiple large orders in hand. Capital is shifting from "betting on general technologies" to "deploying in real - world application scenarios."

Powerful players like Unitree can directly go through the Growth Enterprise Market, but it's not that easy for other players. They either choose the Hong Kong stock market or consider "back - door listing." Although the threshold of the Hong Kong stock market is relatively lower, it has also increased in recent years.

So since this year, we've seen many robot companies get orders. Ubtech officially stated that as of now, its Walker series of humanoid robots have received contracts worth nearly 500 million yuan, and some orders will be delivered within 2025. On October 16th, Zhipu Robotics held an online live press conference and officially launched its new - generation industrial - grade interactive embodied operation robot, Zhipu Elf G2. It is reported that currently, the Elf G2 has received orders worth hundreds of millions of yuan and has started the first batch of commercial deliveries.

Leju Robot announced in late July that it had received an order worth 82.95 million yuan; Zhongqing Robot signed a strategic cooperation agreement with Toronto Technology, agreeing to purchase no less than 2,000 units within three years; the total order volume of Songyan Power's N2 has exceeded 2,500 units. Galaxy General Solutions has started large - scale implementation and will apply about a thousand units this year. Basically, robot enterprises that can still get financing this year will have cooperation orders in the hundreds or thousands. The demand for commercialization is the top priority.

According to the statistical data from Tianyancha, Unitree has won bids for 25 public procurement projects so far this year, approaching the total number of 32 winning bids in the whole year of 2024. In terms of the number of winning bids, Unitree Technology takes the lead, ranking first with 7 direct winning bids. Moreover, in many projects won by integrators, Unitree Technology's robots are also integrated as standard equipment and become part of the solution.

There is no doubt that Unitree is still the industry leader, but there are more and more questions and doubts about it.

As an industry - leading enterprise, how can it not be leading in the "brain" and "dexterous hands" of robots? Can Unitree, which mainly relies on orders from state - owned enterprises and universities, represent the future of commercialization? Every time Elon Musk's Optimus robot appears, it either amazes the audience or brings enough surprises in the robot industry chain, expanding the boundaries and heights of the industry. What has Unitree brought?

When Unitree Technology entered the IPO process, the voice of "there is a bubble in the industry" emerged one after another in the industry. Some people even asked Wang Xingxing directly about his view on the valuation. His response was direct and firm: "Looking at the present from the future, there is no bubble." If compared with the overseas valuation level in the same period, this judgment is not groundless. Not long ago, after Figure completed its Series C financing, its valuation soared to $39 billion (nearly 270 billion yuan in RMB). In contrast, even the leading domestic robot manufacturers, even those in the first echelon, with a valuation in the tens of billions, seem rather "conservative."

The market hopes that Unitree can come up with something eye - catching to quickly catalyze the industry to break the outside world's doubts about the robot industry. Some industry insiders believe that Unitree lacks core barriers. At most, its motion algorithm is relatively stronger, and its products lack application scenarios. In February this year, there was even news that an investment institution was eager to sell Unitree Technology's equity at the previous round's valuation.

This view is not unfounded. Zhipu Robotics' annual R & D investment in large models is close to 400 - 500 million yuan, while Unitree's total R & D expenditure in the past three years is about 350 million yuan, of which 80% is spent on hardware, and the investment in algorithms is less than 20% of the total.

Although Unitree has launched its own UnifoLM - WMA - 0 world model - motion architecture this year, it is quite out of tune with the current mainstream VLA (Vision - Language - Action Perception Intelligence) architecture. Wang Xingxing even directly criticized that VLA will not be the way out. Currently, the data quality and quantity of the VLA model in real - world interactions are insufficient. Even with reinforcement learning (VLA + RL), it is still difficult to achieve efficient convergence, resulting in high training costs and difficult to achieve breakthroughs in general tasks.

In the long run, the ceiling of quadruped robots is quite limited. In terms of humanoid robots, although Unitree is significantly leading in hardware performance, cost control, and mass - production capabilities, it is relatively conservative in AI model R & D. And the competition in AI models is changing rapidly with huge investments. How can such a Unitree take on the responsibility of the industry leader?

From a positive perspective, Unitree can be said to be "pragmatic" at present, but on the other hand, it can also be said to be "too conservative." Unitree seems to have prepared the path for listing early and started considering the unit price and profit margin issues during the product R & D stage. It can't be said that this is wrong. If the industry bubble bursts next, Unitree will be far - sighted. If there is a R & D explosion and it enters a new stage, with its current investment, Unitree may not be able to keep up.

Unitree's Rival Figure, Valued at 270 Billion Yuan

Figure 03 Amazed the Audience

On October 9th local time, the American humanoid robot company Figure AI officially launched its third - generation humanoid robot, Figure 03. Figure, which just completed a $1 billion Series C financing last month, has become the humanoid robot enterprise with the highest valuation globally with a post - investment valuation of $39 billion (about 270 billion yuan in RMB). Compared with the first two generations of humanoid robots, Figure 03 has significant differences in appearance, structural design, energy management, etc.

In terms of appearance, the biggest difference between Figure 03 and the first two generations of robots is that Figure 03 is dressed. According to the video released by Figure, the new - generation product is covered with gray fabric, hiding the internal cables and mechanical structure. In the video, Figure 03 showed about five different "outfits," making it more "human - like" than the first two generations of robots.

Figure Company emphasized: "The company's focus on the household market does not in the least weaken the potential of Figure 03 in the commercial market. By solving the variability and intractability of the home environment, Figure is developing a general - purpose product capable of handling various tasks."

Figure 03 is also officially marked as "designed for large - scale application," and its BotQ production line has the capacity to produce 100,000 units per year. It seems that the only thing left is mass - production. But will it really be that smooth?

Another Case of Premature Publicity

Bernt Bornich, the founder of the Norwegian robot company 1X Technologies, reposted a post by Brett Adcock (CEO of Figure) and said: "Imitation is still the sincerest form of flattery, and taste is hard to fake." In the content reposted by Bernt Bornich, the fabric covered the mechanical appearance of the robot, which is somewhat similar to the idea of the robot Neo launched by 1X Technologies in August last year, implying that Figure copied its product.

Copying can be regarded as "a coincidence of minds" and is not a big problem. The bigger problem is that Figure's robots are too mysterious, and it's difficult for the outside world to judge their real progress.

After all, for such a long time, Figure has never had a real product on offline physical display. All the information people know can only come from the demo videos released by the official. Facing the doubts from netizens, Figure's CEO directly responded: "There is no remote control for any action in the video."

A reporter witnessed Figure 03 successfully putting items into the dishwasher and cleaning up the debris on the table. However, it still can't fold T - shirts, which is also a common problem for current humanoid robots. Although the demo is wonderful, there is still a long way to go before it becomes a "fully autonomous household robot."

When being interviewed by TIME, Adcock himself admitted: "We hope this robot can autonomously complete most tasks in your home and work around the clock. We haven't achieved this yet. We think it can be achieved in 2026, but it requires great efforts."

It's not the first time that Figure has been questioned for being "unworthy of its reputation." In early April this year, Fortune magazine "questioned its cooperation with BMW," accusing it of serious exaggeration and even false publicity. As of March this year, Figure's humanoid robots only carried out simple parts - handling training during non - production hours, mainly responsible for "picking up parts from the logistics container with both hands and placing them on a fixture."

In response, Brett Adcock expressed his anger and said that he would sue Fortune magazine and the reporter for the unfavorable impact on Figure.

Subsequently, Fortune's report didn't have much impact on Figure because the problem of "publicity first, testing while doing" is common in the industry, and even Elon Musk's Optimus has such a situation.

Currently, there are reports that Tesla's Optimus project has also encountered setbacks. The original mass - production target of 5,000 units in 2025 has been rumored to be postponed. It is reported that the main reason for the suspension is that the hardware problems have not been solved, including overheating of joint motors, insufficient load - bearing capacity of dexterous hands, and insufficient battery life.

In addition to hardware, it is even more difficult to establish a viable business model for humanoid robots in industrial scenarios. Because the working plan of a humanoid robot customized for Company A may not be suitable for Company B, resulting in a lack of standardized production and operation for humanoid robot products, making the input - output ratio of enterprises relatively low. Moreover, there are also problems such as a large "data island effect" (devices work independently and data between processes do not circulate) and difficulty in data leaving the factory.

It's true that the household scenario is a trillion - dollar market, but the broader and non - standard requirements are enough to discourage many startup robot enterprises. The sweeping robot industry is an industry with a history of several decades, and there are still problems such as limited perception ability and occasional errors in item recognition. Household robots with the VLA model will be much better, but they still can't understand task instructions 100%, and their task generalization ability is poor. That's why Adcock thinks that Figure may be able to complete most tasks by 2026.

Is a Real Order Really That Important?

Currently, well - known robot enterprises at home and abroad all have the problem of specific customers, either universities or enterprises with investment and cooperation relationships.

In addition to Unitree, robot enterprises such as Accelerated Evolution and Xinghaitu also frequently appear in universities and research institutions. For example, the robots of Accelerated Evolution have become the "standard equipment" for university teams in RoboCup, and the robots of Xinghaitu have become the "guests" of Fei - Fei Li's team.

In October, there was a video of the Optimus humanoid robot attending the premiere of the Disney movie "Tron: Ares." In the video, Optimus performed kung fu and had a "duel" with movie star Jared Leto. Earlier, Optimus also danced various funny dances.

Ubtech's Walker S humanoid robot mainly conducts tasks such as door lock inspection and seat belt detection at NIO's Hefei base. International enterprises such as Amazon and DHL are also testing humanoid robots for warehouse picking. Zhipu Robotics has jointly established factories with listed companies such as Fuling Precision Industry, China Mobile Information, and Lens Technology and invested in enterprises in the industrial chain. These application scenarios are still mainly in the pilot or small - scale deployment stage, covering limited industries and task types, far from being "ubiquitous."

This kind of demand has always been criticized as a "pseudo - demand," not a real industrial demand, but a demand nurtured and subsidized.

This kind of problem will first be presented to Unitree. Facing these questions, Wang Xingxing did not avoid them. Instead, he actively defined the value of scientific research demand. Wang Xingxing said that "the educational robots and humanoid robot products represented by Unitree Technology have entered nearly a thousand universities around the world, and more than 3,000 excellent technical papers have been produced."

Objectively speaking, every basic scientific and technological change is inseparable from the long - term capital investment with official background. The rise of the Japanese robot industry is inseparable from the early support of the government and giant enterprises. In the 1970s, the Japanese government encouraged the R & D and application of robots through special funds, tax incentives, etc. Giants such as Kawasaki Heavy Industries and Yaskawa Electric Corporation all deployed in this field, and finally Japan became a global power in industrial robots at that time. The development of the US semiconductor industry is also the same. Early orders from the Department of Defense provided stable market demand and opportunities for technological iteration for semiconductor enterprises, helping them through the high - risk stage with large R & D investment and uncertain commercial applications.

Objectively speaking, as of now, the humanoid robot industry still stays in vague expressions such as "pilot," "POC verification," and "in the process of development." The humanoid robots that can operate stably and create considerable commercial value may still not have appeared.

But it is still worthy of praise for robots to move out of the laboratory and into specific application scenarios. Because without a large number of customers using them, it is difficult to generate real data and various feedbacks, and product iteration often deviates from reality. It may be difficult to form a real commercial flywheel cycle.

As pointed out in a research report released by McGaugh Securities in April, it is precisely because the domestic robot industry is hot enough to attract a