Recently, US investor Josef Chen released research notes after visiting Unitree Robotics' Hangzhou factory, sparking industry discussion. On-site, he observed that only one or two new automated production lines had been implemented in the actuator workshop, while a large number of processes, such as winding and assembly, were still completed by workers. This frontline observation, combined with data disclosed in Unitree's IPO prospectus, reveals the current manufacturing status of leading domestic robotics manufacturers.
Many people still remember Unitree's previously viral video: a G1 humanoid robot equipped with the UnifoLM‑X1‑0 model assembling actuator components on a production line. The video garnered over 28 million views, making "robots building robots" a hot topic. However, the prospectus statement was quite restrained: this scenario is merely a pilot project, and final assembly currently remains primarily manual.
01、Cold Cost Data Explains the Pace of Automation Advancement
By the end of2025, Unitree's book value of machinery and equipment was approximately USD 1.7 million, with annual output approaching 32,000 robots. This translates to an equipment investment of only about USD 50 per robot. In terms of cost structure, for 2025, materials accounted for 81.7% of Unitree's cost of goods sold (COGS), manufacturing overhead for 10.2%, and direct labor (including outsourced assembly personnel) for only 8.1%. The low proportion of labor directly reduces the urgency for automation transformation.
Calculations show that a new automated motor production line can reduce manual labor hours per motor by 57%, but this ultimately reflects only a 1.7% reduction in total unit cost. Humanoid robot products iterate rapidly, with frequent model updates, causing custom tooling and fixtures to become obsolete quickly. Therefore, Unitree's automation strategy is highly pragmatic: prioritizing the transformation of universal joint module production lines, where one line can be reused across multiple quadruped and humanoid products, thereby improving the return on equipment investment.
The prospectus also discloses that since 2023, thanks to scale expansion, process standardization, and improved worker proficiency, the manual labor hours required per robot have already been halved. Wang Xingxing explicitly stated at this year's World Robot Conference: Humanoid robots are not yet ready for large-scale mass production. Revenue structure supports this judgment: in the first nine months of 2025, three-quarters of Unitree's humanoid robot revenue came from scientific research and educational procurement; industrial scenario revenue was less than 5%, and more than half of orders classified as industrial use were actually exhibition display models.
02、The US-China Manufacturing Cost Gap Creates Completely Different Value for Automation
Josef Chen conducted a comparative calculation: The hourly wage in US factories is approximately 7.5 times that of China. If this production model were replicated in the US, the proportion of labor costs would skyrocket to 40%.
The same automated motor production line saves only 1.7% of total unit cost in Hangzhou, but in the US, the cost reduction could reach 9%, amplifying the commercial value of automation equipment by 5-6 times.
Policy factors further amplify the motivation for overseas manufacturers to build vertical integrated factories and advance automation. The US FCC introduced new regulations in July, requiring overseas humanoid robots and robotic dogs entering the US market to meet thresholds for local component cost proportions, with the threshold rising to 75% by 2029. For drones funded by the Pentagon, stator winding processes are prohibited from being completed within China starting August 2027.
Overseas manufacturers have already made vertical self-built factories a core strategy.
1X has built its own factory in California, developing and producing everything from wires to actuators internally;
Figure's BotQ factory capacity is ramping up rapidly, increasing from1 humanoid robot per day to 1 per hour;
Tesla's Fremont factory continues to build Optimus mass production lines, targeting an annual capacity of one million units;
Agility Digit produces locally in Oregon and recently listed on Nasdaq.
Of course, in terms of total volume, the shipment scale of these overseas players still lags significantly behind Unitree. Domestic player AgiBot previously disclosed that its cumulative finished product output exceeded 20,000 units, ranking among the top humanoid shippers in the first half of the year, while also laying out overseas production lines in Serbia to address global supply chain requirements.
03、New Industry Opportunity: Not in Finished Products, But in Actuators and Manufacturing Equipment
This research offers a disruptive judgment: The current investment bonus in the humanoid robot sector lies not with finished product manufacturers, but with motor bodies and specialized production line equipment for manufacturing motors.
According to industry calculations, one humanoid robot (excluding dexterous hands) requires 25-30 actuators. If Optimus achieves its annual target of one million units, the corresponding demand for motors will reach 25 million units, creating massive incremental opportunities for standardized motors and flexible stator winding equipment. Traditional automated production lines face a pain point: equipment investment can only be recouped once product design is finalized. However, AI tools are rewriting this economic model, significantly shortening production line debugging and robot programming cycles. One company disclosed that fault repair time was compressed from 21.5 days to 7.9 days, continuously improving the feasibility of flexible production lines.
The research also highlights supply chain risks: global permanent magnet production capacity is highly concentrated in China; however, MP Materials' Texas factory in the US is expected to deliver products by the end of 2026 and has already obtained General Motors qualification. Meanwhile, domestic card-style stator winding equipment technology has caught up with German and Italian competitors, priced at half that of overseas vendors, and has begun exporting to Europe.
Expert Commentary
Josef Chen's field research did not negate "robots building robots," but rather helped the industry return to rationality. The humanoid robot industry is still in its early stages, and automation is an input-output calculation that requires repeated scrutiny.
In China, under conditions of low labor costs and rapid product iteration, manual assembly is the most cost-effective solution at this stage; however, in the US, constrained by high labor costs and trade barriers, vertical self-built factories and automated production lines are survival necessities. Neither route is superior; they are simply two different answers to the industrial challenge.
In the long term, as finished product shipments continue to increase and product forms gradually converge, the standardization of joints and motors will improve, and the value of automated production lines will slowly emerge. In the next round of industrial competition, beyond finished products, large models, and dexterous hands, winding equipment, flexible assembly tooling, and actuator-specific manufacturing equipment are becoming a quiet but critical underlying trend.
For domestic supply chain manufacturers, it is essential to serve the current small-batch, fast-iteration flexible production needs of domestic finished product makers, while also preparing in advance to seize equipment export opportunities brought by overseas localized factory construction.
This article is compiled based on public network information, for reference only, and does not constitute investment advice.