A leading humanoid robot joint company secured new funding in half a year, with Co-Win Ventures leading a multi-hundred-million-yuan investment | 36Kr Exclusive
Author | Huang Nan
Editor | Yuan Silai
36Kr learned that Zero Error Motion Control (Shenzhen) Co., Ltd. (hereinafter referred to as "Zero Error Motion Control") has recently completed a Series C+ financing of several hundred million yuan. This round was led by Tongchuang Weiye, followed by Guotai Junan Innovation Investment, with additional investment from existing shareholder Huakong Fund. The funds from this round will be used for production capacity expansion and global market layout. Yuanshi Capital served as the exclusive financial advisor for this round, while Yuanshi Capital and Yiwei Capital jointly acted as financial advisors for subsequent financing.
After the humanoid robot track experienced the boom of concepts and the frenzy of capital, the industry's key phrases are quietly changing, and the production capacity and reliability of the supply chain have become the real bottlenecks. As ontology manufacturers fall into mass production anxiety, the choices and judgments of upstream core component companies may better reflect the real temperature of the track.
Zero Error Motion Control is an enterprise that 36Kr has paid long-term attention to. Founded in 2016, the company has long focused on the R&D and manufacturing of highly reliable, high-precision, standardized key core components for robots, serving the global robotics and automation industry, and focusing on high-reliability application scenarios that originally relied on manual labor.
Official data shows that in 2025, the revenue of Zero Error Motion Control increased by more than 100% year-on-year, and it is expected to maintain a doubling growth rate in 2026. Among them, orders from the field of humanoid and embodied intelligent robots have initially increased, contributing about 65% of the growth rate.
Zero Error Motion Control Integrated Joint (Source / Enterprise)
"The demand of downstream customers has rapidly shifted from 'whether there are mature products' to 'whether products can be supplied in time'," Jia Xiqing, founder of Zero Error Motion Control, told 36Kr. "The orders of leading customers are almost open-ended — we will take as many products as you can produce. Therefore, not only are we striving to increase production capacity, but also our upstream motor and reducer suppliers are being pushed to expand production together."
Production capacity is only one side of the coin. At a more fundamental level, an increasingly sharp design paradox is emerging: under the current technical framework, it is almost impossible to simultaneously achieve the three major goals of anthropomorphic flexibility, human-comparable load capacity, and compact human-like size without breaking through physical limits. Zero Error Motion Control attributes this to the "Impossible Triangle of Humanoid Robot Design".
"This is not because people are unwilling to do it, but because of objective laws," Jia Xiqing explained. "If we require both joint degrees of freedom and load capacity to be comparable to humans, the volume will definitely be large, which will exceed the human body size; if the volume is limited and high load is pursued, it will be impossible to accommodate enough joints, and the flexibility will inevitably be compromised; if we try to balance flexibility and compact size, the load capacity of the robot will be greatly reduced in the end."
Zero Error Motion Control Joint Solution Diagram (Source / Enterprise)
Each side of the "Impossible Triangle" corresponds to a series of real-world compromises. In order to fit enough joints into the limited size of the robot, many manufacturers have to make the joints "protrude" or "bend", which eventually leads to the robot's kinematic model being different from that of humans. This also means that it is impossible to directly use a large number of human action videos for model training, and a separate training environment and dataset must be built for it.
Zero Error Motion Control's differentiated strategy is not to try to subvert the laws of physics, but to focus on mature paths that have been verified by large-scale consistent manufacturing without sacrificing core indicators. Instead of participating in the reconstruction of underlying materials, tooth profiles, and processes, it pushes the integration efficiency of existing solutions to the extreme.
In Jia Xiqing's view, the "customization" mentioned by enterprise customers at present has highly consistent demands. "It is nothing more than lighter, stronger, smaller, more precise, and cheaper. This is not individuality, but commonality," he added. "Customers just use the word 'customization' to express the universal performance demands that existing products have not yet met."
This judgment has directly affected Zero Error Motion Control's product strategy, transforming those seemingly individual but actually common demands into a multi-model, standardized product matrix. Its eRob series covers various forms such as straight Type I and corner Type T, with a minimum module diameter of only 70mm and a maximum allowable torque of 1180Nm; relying on the self-developed advantages of encoders and drives, under the same load specification, the axial length of the product is reduced by 44% compared with the industry's general solution, and the overall weight is reduced by more than 20%, which can cover a wide range of scenarios from logistics storage to heavy-duty industry.
Different from the traditional whole-machine assembly of 5-degree-of-freedom manipulators, which usually requires 30 to 40 independent parts, Zero Error Motion Control has reduced the number of parts of the same configuration to 7 after structural optimization, greatly reducing assembly difficulty and failure rate.
Zero Error Motion Control "Lego-style" Modular Design (Source / Enterprise)
This "Lego-style" modular design perfectly fits the design idea of humanoid robots that benchmark against the human anatomical structure. Standardized modules for key joints such as shoulders, elbows, and hips can be directly snap-fitted without additional processed parts. At the same time, the forearm space is fully reserved, which can be allocated to cable-driven dexterous hand manufacturers to arrange tendon drive components, compatible with the mainstream five-finger dexterous hand solutions on the market, realizing full-link adaptation from joints to end effectors.
The ultimate goal of modularization is to make joints a standard product. For downstream ontology manufacturers, the advantages of product standardization go far beyond structural simplicity: predictable delivery cycles, consistent quality control, and cost optimization under economies of scale are the more core industrial values. What Zero Error Motion Control provides is not an engineering sample that needs repeated verification, but an out-of-the-box power unit, which compresses the selection, design, procurement, and assembly of hundreds of mechatronic devices into simple module splicing.
The difficulties in prototype R&D have been gradually overcome, but the core contradiction of large-scale implementation is concentrated on the supply chain end. In the manufacturing industry, the gap between "being able to make one prototype" and "being able to make products stably, in large quantities, and consistently" has not been eliminated.
By building self-owned 5-axis precision machining, testing, and assembly workshops, and conducting full inspection and working condition simulation for each joint module, Zero Error Motion Control has completed the full leap from product certification, large-scale consistent mass manufacturing to commercial-scale implementation.
At present, its products have covered industrial applications such as humanoid robots, industrial robots, surgical robots, collaborative robots, logistics, and transportation, with more than 2,000 downstream customers, and long-term customers cover leading enterprises in the global consumer electronics, robotics, automation, automotive and other fields.
The explosion of demand is forcing the upstream component industry to accelerate iteration. Humanoid robots have entered the stage of small-scale implementation, and the entire industrial chain is facing a unified proposition: to maintain reliability while expanding production capacity and reducing costs. There are no shortcuts on this road of industrialization.
The following is an excerpt from the interview between 36Kr and Jia Xiqing, founder of Zero Error Motion Control (slightly edited):
36Kr: There is an "Impossible Triangle" in humanoid robot design that flexibility, load, and human-like size cannot be balanced at the same time. Zero Error Motion Control relies on integration optimization to achieve size and weight reduction. What cascading industrial values can this design logic bring?
Jia Xiqing: Constrained by physical limits, most manufacturers have to make the structure convex or bent in order to fit a large number of components such as motors, reducers, encoders, and drives into the joints. As a result, the kinematic model of the whole machine has a huge deviation from the physiological structure of the human body, and it is impossible to directly use a large number of public human action videos for AI training, because the "skeleton" of the robot is different from that of humans. Whole-machine manufacturers have to additionally build dedicated motion capture equipment and datasets, which raises the threshold and cost of algorithm R&D.
In contrast, Zero Error Motion Control is doing the opposite. We first define the boundary of the human anatomical space — what is the maximum allowable envelope size for shoulder joints, elbow joints, hips, and knees, and then perform extreme integration within this "cage". Under the same load specification, the axial length of the joints is reduced and the self-weight is lowered. Moreover, the key joints of shoulders, elbows, and hips realize snap-fit assembly, and the forearm space is fully reserved, which can be compatible with various cable-driven and tendon-driven dexterous hand solutions on the market.
The biggest industrial value of this design is that once the external dimensions of the joints and the kinematic pair relationship are close to the native human structure, ontology manufacturers can directly use mature human motion databases, such as the public CMU MoCap and AMASS, for model training, fundamentally alleviating the long-standing problem of excessively high dataset customization costs in the industry.
36Kr: Combined with the current divergent situation of industry routes, how to understand the phenomenon of "pseudo-customization, real standardization"?
Jia Xiqing: At present, the technical routes of robot joints are far from converging, and they are even continuing to diverge. As a result, downstream ontology manufacturers can hardly find an "exactly right" ready-made product, and the interfaces, sizes, and performance parameters defined by each company are different. Therefore, customers can only come to us in the name of "customization", but their actual demands are nothing more than a few repeated points: lighter, stronger, smaller, more precise, and cheaper.
This is not individuality, but commonality. They just use the word "customization" to express the universal performance demands that existing products have not yet met. There is no real personalized development that needs to be developed from scratch.
Based on this judgment, we abandoned the old path of non-standard customized development, and precipitated these universal performance demands into a multi-model, standardized product matrix. Only in a very small number of specific working conditions, such as explosion-proof and IP67 waterproof, will we make differentiated configurations.
The value of this strategy is twofold. For customers, the selection, design, procurement, and assembly are compressed from hundreds of components to several modules, greatly shortening the R&D trial and error cycle; for us, the cost per unit is reduced through mass production, and both delivery cycles and quality control are predictable.
36Kr: Leading customers place large-volume locked orders for procurement. Where are the main difficulties in ramping up production capacity?
Jia Xiqing: The production capacity bottleneck of the entire industrial chain is a systematic problem.
On the one hand, the products we make are highly reliable, with a fundamental market covering scenarios that "cannot fail" such as medical surgery, industrial automation, and electric power operations. The implementation of new production lines, recruitment and skill training of operators, and the construction of a full-process quality inspection system all require a certain cycle of time. Blindly speeding up, without properly training new employees and improving the inspection mechanism, will easily lead to a decline in product consistency and problems in ex-factory yield. For companies like us that rely on reliability for survival, this is non-negotiable.
On the other hand, upstream motor suppliers and reducer outsourcing factories are also being pushed to expand production by the strong demand from downstream. They need to purchase equipment, renovate production lines, and train personnel, which also takes a cycle of at least several months.
At present, the demand in the humanoid robot industry has rapidly shifted from "verifying whether the sample works well" in 2024 and 2025 to "stable supply in large quantities with consistent quality". From materials, processing, components to final assembly, the entire supply chain is in a stage of passively catching up with demand. This is also the most distinct industrial feature of the "first year of mass production" in 2026: technical problems are gradually narrowing, but the supply chain's delivery capacity has become the core variable that affects the rhythm of the industry.
Whoever can ramp up production capacity while maintaining reliability will be able to seize this wave of dividends.