Unitree Returns to the 300 Billion Mark, Wang Xingxing Has Survived the Darkest Moment

Wang Xingxing is back in the game.

Two months ago, Unitree Robotics saw its stock price halved, with endless doubts pouring in. As of November 11, its stock price has doubled in less than a month, and its market value has rebounded to 300 billion yuan.

Even JPMorgan Chase, which had previously held a bearish view on it, raised its rating by two notches from "underweight" to "overweight".

For the same company, why did the market's attitude make such a sharp U-turn in just two months?

【The Wind Has Shifted】

The key to this round of rebound for Unitree is that the market has begun to believe that the robotics sector will see an earlier boom.

On October 22, Wang Xingxing judged that physical AI has shown initial signs of breakthrough, and the window of opportunity may come in the next one to two years, or two to three years. Subsequently, JPMorgan Chase upgraded Unitree's rating, and raised its 2028 humanoid robot shipment forecast from about 41,600 units to 111,600 units.

Behind this lies a long-standing threshold for the robotics industry: it is no longer rare for robots to run, jump and put on performances, but there has never been a clear answer to how far they are from large-scale practical work.

How many types of environments a robot can adapt to directly determines how large a market it can tap into.

Advances in models are lowering this threshold.

On November 4, Apple's research team announced the research results of HALO. They carried out 360 tests on 6 types of individual skills for the Unitree G1 in 6 unfamiliar environments, and the average success rate of individual tasks reached 91%. These environments had never been involved in training, nor had they been scanned and mapped in advance.

The significance of this result is that with the continuous advancement of models, the same set of hardware will have the opportunity to adapt to more scenarios and undertake more types of work.

Training methods are also evolving. In the past, data relied on real robots and manual collection item by item, but now it can be supplemented through simulation. HALO used about 1,000 virtual environments to generate training data. The scale of training can be expanded more relying on computing power, which makes it possible to accelerate model iteration.

The improvement of capability on one hand, and the acceleration of training on the other, jointly influence the timeline for robots to enter practical applications.

Of course, there is still a gap between research results and stable operation, and the forecasts given by institutions are just predictions. But market expectations often run ahead of actual orders: once robots can work stably, part of the labor demand in factories and warehouses may be converted into robot orders.

What the market is betting on is that the step for robots to evolve from demonstration to large-scale practical work may come earlier than expected.

For Unitree, this means that its existing hardware may usher in broader demand. The closer the industry breakthrough is, the more intense the competition over cost, product and mass production capacity will be.

【Unitree's First-Mover Advantage】

To implement large models, a physical carrier is indispensable.

Once the industry speeds up, the party that can deliver robots to customers in batches at lower costs will have a better chance of securing orders. But when facing the same opportunity, why can Unitree occupy a more favorable position?

First, take a look at the price. The starting price of the basic version of Unitree G1 has been reduced from 99,000 yuan at launch to 85,000 yuan.

While the price has dropped, the profit margin remains considerable. In 2025, Unitree lowered the selling price of some of its humanoid robot products, yet its overall gross profit margin still reached 63.18%.

Behind this is the long-accumulated cost advantage. Key components such as motors, reducers and encoders are independently developed by Unitree, and some parts of quadruped robots and humanoid robots are interchangeable, which not only saves the cost of repeated development, but also reduces costs through larger procurement volumes.

To popularize robots, more customers need to be able to afford them. The fact that Unitree can lower the selling price while maintaining a high gross profit margin means it still has room to further expand its market share.

Once price competition intensifies, the cost advantage will become Unitree's bargaining chip to secure orders.

Mass production is another card in Unitree's hand.

In 2025, Unitree shipped 5,511 humanoid robots, more than ten times the figure of the previous year. In October this year, Wang Xingxing disclosed that the cumulative output of G1 has reached about 10,000 to 20,000 units.

The value of these products lies not only in the sales that have been realized. After producing thousands of robots, Unitree has accumulated experience in how to organize production, ensure quality and deliver goods on time. With more orders, it will be able to increase production capacity faster.

Breakthroughs in models may rapidly expand the use cases of robots, but mass production and delivery capabilities cannot be built overnight. The earlier the industry boom comes, the easier it will be for enterprises that have already achieved scale to seize the first opportunity.

Unitree can benefit from more than just breakthroughs in its own models.

The Apple team's verification of HALO on the G1 is a case in point. Algorithm teams can use mature robots to avoid the investment of designing and manufacturing hardware from scratch. By launching open-source models and providing development tools, Unitree is also attracting more teams to develop applications around its products.

Once more models run successfully on Unitree's hardware, other parties' technological breakthroughs may also be converted into new orders for Unitree.

【Wang Xingxing's Confidence】

Wang Xingxing has spent years figuring out how to make robots with better performance at lower costs.

When he was still a student, he made parts by himself, and repeatedly adjusted the matching between drivers and motors. Later, when he founded Unitree, these technical explorations became the methods for the company to reduce product costs and expand the market.

In 2016, Wang Xingxing received 2 million yuan in investment and started his business with a team of three to four people. Two years later, the financing process did not go smoothly, and at the most difficult time, there was only about 100,000 yuan left in the company's account.

But at that time, Unitree had already produced prototypes, received orders, and customers had paid deposits. Wang Xingxing was determined to deliver the products. He recalled later that his thought at that time was: "I need to get this thing done."

In the second half of 2018, the products were finally delivered to customers. The quadruped robots brought sales revenue, and gave Unitree the capital to continue R&D and improve its products.

By 2023, Unitree launched its humanoid robot H1; in 2024, it launched the G1. Although the product form has changed, the previous investment in components, motion control and manufacturing has found new applications.

From making parts by himself, to R&D of complete machines, and then to batch delivery, the efforts Wang Xingxing put into technology in his early years have gradually turned into Unitree's advantages in cost control and mass production.

But he is well aware that the leading position will not stay with him forever. Wang Xingxing once said bluntly that if there is no iteration for one or two years, the company will become mediocre. What new achievements the company can come up with next will determine how far it can go.

In the first three quarters of this year, Unitree's R&D investment reached 263 million yuan, a year-on-year increase of 191.48%, significantly higher than the 55.71% revenue growth rate in the same period. The company continues to invest in embodied large models, computing power and new products.

Past accumulation has given Unitree a first-mover advantage, and continuous R&D is the only way to maintain this edge.

With the advancement of models, robotics enterprises have begun to compete for the market after large-scale application.

Morgan Stanley predicts that by 2050, the global robotics industry covering humanoid robots, autonomous driving, drones and other fields may reach 25 trillion US dollars in hardware sales.

The more work robots can undertake, the larger the market they can enter. The current leading sales volume is just the starting point of this competition.

In the past, Wang Xingxing needed to prove that robots can become a viable business. Today's round of industrial breakthroughs is giving him a chance to compete for a seat among the global tech giants.