Price is no longer a barrier, but just a ticket? Ten thousand-yuan humanoid robots usher in a new competition of "real delivery"

The humanoid robot Xiao Bumi quickly became popular after appearing on the 2026 CCTV Spring Festival Gala and starring in a sketch with Cai Ming. Superimposed on its performance in the 2025 Robot Marathon, Songyan Power, founded by Jiang Zheyuan, who studied at Tsinghua University, is hoped by the industry to become the second Yushu Technology.

In March this year, Songyan Power completed Series B financing, led by Morning Road Capital, an industrial investment platform of CATL. So far, Songyan Power has completed 10 rounds of financing. But beyond the popularity, the real test lies on the delivery side.

At the beginning of this year, Jiang Zheyuan set a goal of selling at least 10,000 units of Xiaobumi in 2026. As of September this year, Xiaobumi has delivered just over 5,000 units. How much does Xiaobumi account for in B-side (enterprise side) and C-side (consumer side) sales? "I personally feel about half and half." Recently, Jiang Zheyuan, the founder of Songyan Dynamics, confessed to the reporter of "Daily Economic News" (hereinafter referred to as the reporter) at the "2026 BILIBILI NEXT Conference" at Station B.

Betting on price is how Songyan Dynamics differentiates itself from its peers. When Yushu Technology, Zhiyuan and others were competing for "how many units to roll off the production line" and "how many working hours can be delivered", Songyan Power used a price of 9,998 yuan to bring the bipedal humanoid robot into the 10,000 yuan threshold for the first time.

Faced with the industry turning to competition of "whose robot can do the job better?", can Songyan Power's 10,000-yuan machine support the business expectations of large-scale delivery? With the rapid evolution of general large-scale models, when players such as Zhiyuan, Independent Variable, and Galaxy General have already regarded the embodied brain as the core of their narrative, how can Songyan Dynamics, which started late, break through?

Low prices do not mean the market is fully mature

Ten thousand yuan-level humanoid robots are becoming a watershed in the commercialization of embodied intelligence.

Compared with the robots of Yushu Technology and Zhiyuan that have entered performance, housework, factory applications, shopping mall scenes, etc., Xiaobumi has avoided these tracks and scenes since its launch. At the beginning of 2026, Jiang Zheyuan once stated that in the short term, Xiao Bumi should first enter the family scene, especially families with children, to play the role of companionship, interaction and education. Later, he even called out the goal of selling at least 10,000 units of Xiaobumi in 2026.

However, low prices do not mean that the market is fully mature. Today, Songyan Power is still far away from this sales target.

Data shows that as of September 2026, Xiaobumi has delivered more than 5,000 units. Among them, B-side orders are close to 2,000 units. Every reporter noticed that as of November 3, in JD.com's "NOETIX JD.com Self-operated Flagship Store", the same model Xiaobumi Spring Festival Gala home robot priced at 9,998 yuan was on the top 20 smart robot gold list in November.

Jiang Zheyuan said: "I personally feel that Xiaobumi's B-side and C-side sales account for about half each. The shipment volume of the C-side will be larger, but the amount of the B-side will be larger." This means that although consumer-grade humanoid robots have a certain user appeal, their commercial value still depends more on institutional scenarios such as scientific research, education, and commercial exhibitions.

At present, price, mass production and intelligent capabilities are becoming common keywords in the embodied intelligence track, and the competition for 10,000-yuan phones has also quietly changed: price may no longer be a barrier, but just an admission ticket.

Every reporter noticed that in June this year, Yushu Technology's R1 humanoid robot lowered its price from 39,900 yuan to 29,900 yuan, and topped the JD.com sales list; in September this year, Zhiyuan Qiyuan Robot launched two personal robots, Q1 and T1, on the shelves at prices starting from 19,999 yuan. It has now ushered in the 5,000th personal robot to officially roll off the production line.

As more and more leading players aim at cost-effectiveness, the competition for 10,000-yuan machines is no longer just "who can sell the robot cheaper", but also "who can deliver capabilities that are truly worth buying at this price."

Robot companies cannot compete with companies that only make large models

If Wan Yuan Machine solves the problem of whether the robot can be sold, then the embodied brain determines whether the robot can move from performing preset actions to completing tasks autonomously.

Faced with the current rapid iteration of large models, "will large models replace the self-developed brains of robotics companies" has also become the focus of market attention.

In this regard, Jiang Zheyuan believes that the generative pre-training converter will not be a whole brain - the number of model parameters is too large and cannot be run on the device side, and the network delay problem is also difficult to solve. "The general large model can be used as a good planner, or a precursor to System2 (slow thinking), but the system access still needs to be handled by the robotics company itself. There is actually a lot of room for improvement here."

"For robotics companies, if they use their computing power and financial resources, they will definitely not be able to compete with those companies that only make large models. This is far from the case." Jiang Zheyuan further explained using Tesla's intelligent assisted driving system as an example. The reason why intelligent assisted driving systems are difficult to be replaced by general large models is that they hold continuously rolling and accumulated driving data. Large models have never seen this kind of data.

"Part of the barriers for future robotics companies comes from their own closed-loop data on home scenes and specific operating tasks. If these data can generate a flywheel, it will not be eaten by the large model. This is why we must build a brain." Jiang Zheyuan told reporters.

However, data closed loop is not a slogan, but requires enough robots, sufficiently rich tasks, and real scenarios that can continue to operate.

Since this year, leading players such as Zhiyuan, Autovariable, Galaxy General, etc. have long regarded the embodied brain as the core of their narratives. However, in comparison, Songyan Dynamics did not launch the embodied intelligence technology brand Scalabot until September this year, and it has lagged far behind the leading players in terms of time rhythm.

Faced with the question "Why do we build the body first and then the brain?", Jiang Zheyuan told reporters: "We are an enterprise, not a research laboratory. The definition of a high-tech enterprise, I think, is to develop technologies that are as high as I can reach by standing on tiptoes and jumping high. If you put on a telescope and look far away, you still can't see the possibility of large-scale implementation, then this is not what should be done at this stage."