This year, the hottest sub-direction in the embodied intelligence track should be embodied data, as recently reviewed by ITJuzi (you can check it out via the provided link); I believe the second hottest is the dexterous hand.
Dexterous hand startups are raising funds intensively—by early August, the dexterous hand track recorded 74 financing events, involving 47 companies, with a total disclosed amount of approximately 285.1 billion RMB. The total financing in the first seven months has already surpassed that of the entire year of 2025, indicating that the track's heat is still accelerating.
Changes at the industry level are also intriguing—more and more humanoid robot OEMs are choosing to develop dexterous hands as a core in-house module rather than procuring them externally. This raises a very interesting question: Is the competition for dexterous hands an independent supply chain track, or an ability that will be internalized by OEMs?
This report divides these 47 companies into two camps: the "Ontology Faction" (15 humanoid robot OEMs developing dexterous hands in-house) and the "Supplier Faction" (32 third-party suppliers independently selling dexterous hands and core components). It examines their financing data, technical routes, and competitive logic separately.
Our core findings are:
In-house development has become a de facto entry threshold for OEMs, but they still highly rely on external supply for underlying components;
The supplier track is showing distinctly different moats and competitive landscapes across three sub-directions, indicating a structural divergence is underway.
Chapter 1: The 2026 Dexterous Hand Financing Panorama: The "Ontology Faction" and "Supplier Faction" Form Their Own Systems
From January to early August 2026 (data as of August 3), the ITJuzi database recorded 74 financing events in the dexterous hand track, involving 47 independent companies.
Looking at a longer timeline, the financing heat in the dexterous hand track shows a steep upward curve:
In terms of the number of events, 2024 saw a more than twofold year-on-year increase, and 2025 nearly tripled compared to 2024.
In terms of amount, the growth is even more dramatic: The 285.1 billion RMB in the first eight months of 2026 has already exceeded the 169 billion RMB for the entire year of 2025.
Considering that nearly five months of data for 2026 are yet to be included, the total annual amount is highly likely to exceed 400 billion RMB. The dexterous hand track is heating up at an accelerating pace.
It is worth noting that while the number of events surged in 2025, the average amount per event decreased (169 billion / 99 events ≈ 170 million/event), indicating a large influx of early-stage projects that year. In contrast, in the first eight months of 2026, although the number of events is slightly lower than the pace of the same period in 2025, the average amount per event rebounded to 385 million RMB, suggesting the track is shifting from "quantity expansion" to "capital concentration"—money is converging towards leading projects.
It is not appropriate to view the financing of companies that make the main body (and incidentally make dexterous hands) and companies that primarily focus on making dexterous hands together.
We categorize them into two groups based on their business attributes (Note: The following includes only companies with financing records in 2026, not all companies in this track):
* The following only includes companies with financing records in 2026, not all companies in this track.
Ontology Manufacturers (15 companies): Their main business is humanoid robot complete machines, with dexterous hands as a self-developed core module—not sold separately to the outside, or even if they have the potential to supply externally, currently mainly used for their own products.
These 15 companies include: DynamicMotion AI, Stardust Era, Galaxy General, Independent Variable Robotics, StarDust Intelligence, Zhongqing Robotics, Magic Atom, Matrix SuperAI, MoonSpring Bionics, Qingbao Robotics, Zeroth Power Robotics, Zhejiang Humanoid, PineExtension Dynamics, Gesong Technology, Half-Awake Embodied.
In the first seven months of this year, they collectively completed 23 financing events, with disclosed amounts totaling approximately 180 to 200 billion RMB.
Third-party Suppliers (32 companies): Independently develop and sell dexterous hand complete machines, tactile sensors, micro motors, precision lead screws, and other core components. Their customers include humanoid robot companies, research institutions, industrial automation manufacturers, etc.
These 32 companies collectively completed 51 financing events in the first seven months of this year, with disclosed amounts totaling approximately 85.1 billion RMB.
Looking at these two sets of data together, there are three observations:
First, the "average amount per event" for ontology manufacturers far exceeds that of suppliers.
Ontology manufacturers: 23 events, ~200 billion RMB, average >8 billion per event; Suppliers: 51 events, 85 billion RMB, average <2 billion per event.
This indicates that ontology manufacturers are valued based on a "vehicle-level" logic; investors are betting on the entire humanoid robot track, with the dexterous hand being just one technical fulcrum. Suppliers are valued based on a "component-level" logic, which inherently has a lower ceiling.
Second, the "number of companies" in the dexterous hand supplier track far exceeds that of ontology manufacturers—essentially double.
This suggests that the barrier to entry for starting a dexterous hand as an independent product category is lowering. A core team of twenty to thirty people with tens of millions in funding can get started. However, it also means the competition intensity in the supplier track far exceeds that of the ontology track.
Third, the financing rhythms of the two camps are completely different.
Financing for ontology manufacturers is concentrated in Series B and later stages, with events beyond Series B accounting for over 56% of the total. Some have reached Series C and Pre-IPO. Suppliers, on the other hand, are highly concentrated between Angel and Series A rounds, with Angel to Pre-A rounds exceeding 45%. In other words, ontology manufacturers' financing has entered the "middle-to-late stage," while suppliers' financing is still in the "early-to-mid stage."
Chapter 2: The 15 Ontology Manufacturers, In-house Dexterous Hand Development Becomes a "Standard Move"
The approach of ontology manufacturers towards dexterous hand development has changed significantly. Currently, in-house development of dexterous hands has evolved from a "differentiated strategy" for individual leading companies to a "de facto entry threshold" for humanoid robot OEMs.
The basic information of the 15 ontology manufacturers that received financing in 2026 is shown in the table below. We selected three from among them—DynamicMotion AI, Stardust Era, and StarDust Intelligence—representing three different technical routes: industrial scenarios, full direct drive, and tendon drive, for a comparative in-depth scan.
Overall, the logic behind these 15 financed ontology manufacturers clearly developing dexterous hands in-house is evident: The dexterous hand is the core organ for the robot to "work." Outsourcing it means handing over the most crucial differentiating capability to others. In the early stages of the industry, where technical routes are far from converging, more and more emerging OEMs are unwilling to take this risk.
However, there is a noteworthy signal: The in-house development by ontology manufacturers is mainly focused on the dexterous hand body (mechanical structure + drive control), while underlying components such as tactile sensors, micro motors, and precision lead screws are still predominantly sourced externally by the vast majority of OEMs. The technical barriers and capital intensity of these sub-segments are difficult for OEMs to overcome in the short term. This is also why the "hidden champions" in the supplier track can still secure intensive financing this year.
DynamicMotion AI: The High-load Route for Industrial Scenarios
DynamicMotion AI has chosen a path defined by "scenario defines product." Its self-developed three-fingered dexterous hand is used with the full-size humanoid robot CL-2, focusing on high-load grasping in industrial scenarios—the three-finger scheme strikes a balance between flexibility and load capacity, leaning towards the latter, making it suitable for heavy-load tasks like handling and assembly on factory production lines.
This year, it completed a Series B round of $200 million and a Pre-IPO round of $200 million, totaling over 2.8 billion RMB, ranking among the top in financing volume among the 15 ontology manufacturers. The appearance of the Pre-IPO round indicates the company is already preparing for an IPO, with its financing pace clearly entering the "middle-to-late stage."
Stardust Era: Exploring the Technical Limits of Full Direct Drive
Stardust Era follows the full direct-drive route, self-developing the XHAND series dexterous hands. XHAND1 has 12 active degrees of freedom, can lift 25kg, and perform over 100 complex operations. The advantage of the direct-drive scheme is fast response and high control precision, but cost and size are also the most challenging among the three schemes—choosing this path means tackling hard engineering problems.
This year, it completed a strategic investment of 1 billion RMB + $200 million and a Series B round of 1 billion RMB, placing its financing volume in the first tier as well. Looking at the financing rhythm, Stardust Era is rapidly translating technological leadership into commercial barriers.
StarDust Intelligence: The Mass Production Pioneer of the Tendon-driven Route
StarDust Intelligence adopts the globally pioneered "tendon-driven" technical route, self-developing from the whole-body transmission to the dexterous hand. The idea of the tendon-driven scheme is to use tendons to replace rigid transmission. The advantage is that it is closer to the biomechanical structure of the human hand, offering better flexibility and anthropomorphism, but long-term reliability and precision control are engineering challenges.
Notably, StarDust Intelligence's S1 has achieved thousand-unit-level mass production delivery, with the T1 starting price at 89,900 RMB—among the three technical routes, it is the first to bring the price down to a range perceptible to consumers. It completed a Series B round of 1 billion RMB this year.
Chapter 3: Typology of Dexterous Hand Suppliers, The "Three Types of Players" in the Track
The current 32 companies engaged in dexterous hand-related supply (with financing in 2026) can be roughly divided into three categories. This classification is not academic but more based on business logic—the customers they serve, their moats, and the competitive landscapes they face are completely different.
Type 1: Dexterous Hand Complete Machine Manufacturers ("The Hand Sellers")
The core product of these companies is a complete dexterous hand—packaging everything from fingers and joints to drive control into a plug-and-play module, sold to humanoid robot companies, research institutions, or industrial automation integrators. Their moat lies in systems engineering capabilities: hardware design, mass production processes, cost control, and adaptation to the client's robot "brain."
Several unicorns have emerged in this category this year. AGILINK spun off from Zhiyuan Robot in less than half a year and reached a valuation of tens of billions. Lingxinqiaoshou, founded in 2023, now has a valuation of 15 billion RMB and is already preparing for a Hong Kong IPO. Xynova, founded a year and a half ago, saw its valuation exceed $1 billion.
A consensus is forming in this track: Dexterous hands will become one of the most core, highest-value independent modules in humanoid robots, similar to the engine in the era of fuel vehicles or the battery in the era of new energy electric vehicles.
But competition is visibly intensifying—according to ITJuzi statistics, nearly 20 companies claim to be working on dexterous hand complete machines—from independent manufacturers like In-time Robotics and Lingqiao Intelligent, to emerging players like Yuansheng Intelligent, Zhixing Robotics, Wanna Robotics, Boya Intelligent, and others focusing on specific scenarios like Zhongke Siliconji, Starlight Year, Huiling Technology. Each is competing on technical route, mass production speed, and customer lock-in.
AGILINK
Full Name: Shanghai AGILINK Innovation Intelligent Technology Co., Ltd. | Founded: January 2026 | Headquarters: Shanghai
Team: Founder Xiong Kun holds a Master's degree in Electronic Information Engineering and Robotics from the Hong Kong University of Science and Technology (HKUST). In 2018, he joined the newly established Tencent Robotics X Lab, participating in its creation from 0 to 1, responsible for robot body and software architecture. He later worked at IDEA Research Institute and Inovance Technology. In November 2024, he joined Zhiyuan Robot to lead the dexterous hand business. Co-founder Deng Taihua is the Chairman of Zhiyuan Robot, and Wang Chuang previously served as President of Zhiyuan's General Business. The team covers top-tier research, industry giants, and startups.
Financing History: Angel round in January 2026, Series A in February, Series A+ of several hundred million RMB in May, Series B of 1 billion RMB in June, valuation exceeding 10 billion RMB. It took only 5 months from registration to unicorn status.
Core Highlights: A star project incubated within the Zhiyuan system, starting from an extremely high point—profitable upon establishment. In less than half a year, delivered over 8,000 units of the OmniHand series dexterous hands and over 10,000 grippers. Currently the strongest in commercialization capability among pure dexterous hand companies. Xiong Kun's strategic judgment is: The competition for dexterous hands is not in the lab, but on the production line.
Lingxinqiaoshou
Full Name: Lingxinqiaoshou (Beijing) Technology Co., Ltd. | Founded: July 2023 | Headquarters: Beijing
Team: Founder, CEO & CTO Zhou Yong, a prodigy class graduate from Huazhong University of Science and Technology (HUST), has over 15 years of combined experience in internet and robotics. He has had two complete entrepreneurial experiences (3D simulation and smart unmanned vehicles) and led national-level projects like the Hong Kong-Zhuhai-Macao Bridge. Co-founder Zuo Jiaping, from CloudMinds Robotics and Ninebot, is responsible for hardware mass production and supply chain. Chief AI Architect Su Yang leads cloud-based large model and data system construction. The R&D team exceeds 100 people, with over 90% of core components self-developed and self-produced.
Financing History: Seed round April 2025, Angel round August, Series A October, Series A+ November, Series A++ December 2025; Series B 1.5 billion RMB February 2026, Series B+ April 2026, valuation 15 billion RMB. Reports indicate preparations for a Hong Kong IPO have begun.
Core Highlights: Global market share exceeding 80% for high-DoF dexterous hands, monthly capacity exceeding 4,000 units, covering three major technical routes: linkage, direct drive, and tendon drive. Zhou Yong is an interesting figure—a HUST prodigy class graduate, worked on gaming communities (accumulating underlying 3D simulation experience), tackled unmanned vehicles (mastering hardware engineering), and finally applied all accumulated capabilities in the dexterous hand direction. He says his ultimate motivation for starting the business was a childhood obsession with "Doraemon."
Xynova
Full Name: Hangzhou Xynova Future Technology Co., Ltd. | Founded: December 2024 | Headquarters: Hangzhou
Team: Founder and CEO Xia Yuxuan holds a Master's degree from Columbia University. He previously worked at international top-tier private equity firms and investment banks such as CDH Investments, Morgan Stanley, and Warburg Pincus, with deep investment layouts in the robotics industry chain. This background of "investor turned entrepreneur" is quite rare in this track, implying he has a natural advantage in judging capital rhythm and industry inflection points. The team brings together talent from top institutions like Tsinghua, Zhejiang University, Shanghai Jiao Tong University, Columbia University, University of Pennsylvania, and industry leaders like DJI, Schaeffler, KUKA, Apple, and CATL.
Financing History: Angel round December 2025, Pre-A round March 2026, Series A of several hundred million RMB in May (investors include Xiaomi, Li Auto, CICC Capital, etc.), Series A+ in July 2026, valuation exceeding $1 billion officially becoming a unicorn.
Core Highlights: Creator of the world's first "tendon drive + direct drive" hybrid dexterous hand, Flex 2. Xia Yuxuan said something very memorable—he visited a charger factory and saw a female worker on the production line doing only one thing all day: peeling labels off products. This image became the origin of his entrepreneurship: If there were a sufficiently dexterous hand, such repetitive labor could be replaced. Xiaomi followed from Angel round to Series A+, and the intensive bets from industrial capital indicate the product has passed validation by automakers. The company is expected to reach an annual capacity of ten thousand units by the end of 2026.
Type 2: Tactile Perception Solution Providers ("Those Who Give Hands Feeling")
For a robot to truly work, having agile fingers is not enough. Not knowing how much force to use when picking up an egg, not knowing if a glass bottle is slipping—a dexterous hand without tactile feedback is just an expensive performance prop.
The tactile sensor direction was once dominated by overseas manufacturers (single unit price starting at 100,000 RMB), but several domestic companies are rapidly changing this landscape this year.
The common characteristics of these companies are: extremely high technical barriers and strong binding relationships formed with downstream clients. Once an OEM's dexterous hand adopts a certain tactile solution, the switching cost is very high—meaning first-mover advantage can accumulate in this track.
Pasin AI
Full Name: Pasin Artificial Intelligence Technology (Beijing) Co., Ltd. | Founded: June 2021 | Headquarters: Beijing
Team: Founder and CEO Xu Jincheng holds a Ph.D. from Waseda University, Japan, studying under the authority Professor Shigeki Sugano, whose mentor Dr. Ichiro Kato invented the world's first humanoid robot WABOT-1 in 1973. In terms of academic lineage, Xu Jincheng can be considered a third-generation successor of the "orthodox school" of humanoid robotics. Chief Strategy Officer Luo Xiaoheng is responsible for strategy and financing.
Financing History: Angel round November 2021 (MiraclePlus), Pre-A round December 2022, Series A/A+ rounds 2024, two A+ rounds 2025, to Series B 1 billion RMB March 2026, strategic investment 1 billion RMB August 2026, valuation 15 billion RMB. A total of 10 rounds over about five years, with a steady pace.
Core Highlights: The absolute leader in the domestic tactile sensor track. Its self-developed ITPU multi-dimensional tactile sensing technology is globally leading, with some module prices entering the three-digit range—compared to overseas同类products at the 100,000 RMB level, costs have been reduced by two orders of magnitude. From sensors to dexterous hands to humanoid robot complete machines, it follows a path "from the inside out." BYD and JD.com have transitioned from customers to shareholders, a strong signal in the manufacturing track. Xu Jincheng's dedication to tactile sensation is almost religious—he says, "It's not the brain commanding the hand, but the hand shaping the brain."
Dimeng Robotics
Full Name: Dimeng (Shenzhen) Robotics Technology Co., Ltd. | Founded: December 2021 | Headquarters: Shenzhen
Team: Founded by Professor Wang Yu, Founding Dean of the Robotics Institute at HKUST, and his student Dr. Duan Jianghua. Professor Wang Yu has decades of experience in robot manipulation and is a recognized authority in the field. The team has deep collaborative origins with MIT's Professor Edward Adelson (inventor of the visuotactile technology GelSight), improving upon MIT's technical scheme from the source.
Financing History: Angel round September 2023, Angel round November 2024 (Note:原文有两个天使轮，可能是不同轮次或信息有重复), Angel+ round August 2025, strategic investment 100 million RMB December 2025, Series A nearly 100 million RMB June 2026 (Inovance Industry Investment + China Telecom), valuation 3 billion RMB.
Core Highlights: The first domestic enterprise to achieve independent mass production of visuotactile sensors, with 40,000 sensing units per square centimeter, far exceeding human fingertip density. In April 2026, it open-sourced 10,000 hours of tactile data on its platform, with downloads exceeding one million times in the first month, topping the physical world embodied dataset rankings. What sets this company apart is—it's not just selling sensor hardware, but building "tactile infrastructure." The open-sourcing of data indicates Professor Wang Yu is thinking about the industry ecosystem, not just winning individual battles.
Qianjue Robotics
Full Name: Qianjue Robotics Technology (Shanghai) Co., Ltd. | Founded: May 2024 | Headquarters: Shanghai
Team: Founder Ma Daolin holds a Bachelor's and Ph.D. in Mechanics from Peking University, was a postdoctoral researcher at MIT's Mcube Lab, and currently serves as Associate Professor and Ph.D. supervisor at Shanghai Jiao Tong University's School of Naval Architecture, Ocean & Civil Engineering. In 2021, he won the sole Best Paper Award at the top robotics conference ICRA, proposing a new path for robot manipulation: "Relying on tactile sensation to construct spatial perception." Co-founder Liu Jin holds a Ph.D. in Mechanical Engineering from SJTU, responsible for company operations; Zhao Haonan is responsible for underlying sensor hardware R&D. Core team members are from top institutions like MIT, PKU, Tsinghua, and SJTU.
Financing History: Angel round October 2024, Angel+ round May 2025, Pre-A round October 2025, Series A round July 2026. Four rounds in two years, serving over 300 clients.
Core Highlights: Creator of the world's first miniaturized multimodal tactile perception system, building a complete technical closed loop: "tactile sensor, data acquisition, simulation, VTLA model, scenario solutions." Ma Daolin represents the typical path from "top conference paper to industrial implementation," increasingly recognized in the hard-tech track. Qianjue's VTLA model fuses tactile and visual information at the model level, rather than simply splicing sensors together.
Type 3: Core Drive and Transmission Component Suppliers ("The Parts Makers for Hands")
If the previous two types of companies are the protagonists on stage, these companies are the "hidden champions" behind the scenes. They don't sell complete dexterous hands, but the core components within them—motors, reducers, lead screws—might come from them.
In the long run, these companies are the most stable because regardless of which technical route ultimately prevails for dexterous hands, motors, lead screws, and reducers will be essential. They benefit from the dividends of "import substitution" and "technological cost reduction": on one hand, breaking foreign monopolies; on the other, driving costs down to one-tenth or even lower through process innovation.
Wuji Technology
Full Name: Shanghai Wuji Technology Co., Ltd. | Founded: March 2019 | Headquarters: Shanghai
Team: Founder Pan Yunzhe graduated from the University of Illinois Urbana-Champaign (UIUC) in 2018 with a double major in Computer Science and Chemistry. His entrepreneurial motivation was simple—dissatisfaction with the现状of motors on the market being "both heavy and weak." He spent three years refining the F-series motor from scratch.
Financing History: Angel round August 2023, Pre-A round November 2023, Series A November 2025, Series A+ February 2026, multiple intensive Series A+ rounds from March to July 2026, with a豪华investor lineup.
Core Highlights: The F-series motor achieves 8 Nm peak torque at 200g weight, a power density that is碾压in its class and globally top-tier. The intensive Series A+ financing this year indicates strong capital confidence, surging downstream orders, and the need for continuous capacity expansion.
Zhijian Zhiqing
Full Name: Suzhou Zhijian Zhiqing Technology Co., Ltd. | Founded: August 2024 | Headquarters: Suzhou
Team: Founder and CEO Zhang Yang holds a Bachelor's in Electrical Engineering and Automation (motor方向) from Central South University, with 12 years of motor industry experience. He previously worked at Midea Group, Weifu High-tech, and Innovusion, leading the development of the world's first 1550nm automotive LiDAR scanning system with annual sales exceeding 150,000 units, possessing full-stack capabilities from product definition to mass production. CTO Han Jie has 10 years of motor R&D experience, previously leading mass production of multiple main drive motors and precision motors at Geely and Innovusion. The founding team also includes top scholars from CMU providing technical foresight.
Financing History: Angel round August 2025, Pre-A round June 2026. The two rounds恰好completed the cycle from technical validation to commercial validation.
Core Highlights: Creator of the world's first mass-production process for PCB axial flux stator windings—applying semiconductor process thinking to motor manufacturing. The 48-layer PCB stator is only 4.2mm thick, completely颠覆traditional hand-winding methods for micro motor manufacturing. Product outer diameter starts from 4mm, making them the world's smallest commercially produced hollow cup motors. Mass production yield is接近100%, with costs over 30% lower than traditional solutions. If this technical route can continue to iterate, the rules of the game in the micro motor industry could be rewritten.
Nuoshi Robotics
Full Name: Shenzhen Nuoshi Robotics Co., Ltd. | Founded: July 2023 | Headquarters: Shenzhen
Team: Founder Xu Yang holds a Master's in Mechanical Manufacturing from Tongji University, previously served as R&D Department Head at the Asia-Pacific R&D Center of the French automotive parts giant Valeo. Co-founder Wang Xiaobin previously served as Deputy General Manager of a Valeo subsidiary, a serial entrepreneur who has taken teams from zero to Pre-IPO stages. Core Technical Advisor Professor Xu Genlin is a retired professor from Shanghai University, with nearly 50 years of深耕in mechanical manufacturing processes—he is Xu Yang's father and the technical foundation of the company. This company is essentially a story of "two generations接力": the父辈accumulated half a century of lead screw manufacturing know-how, and the子辈industrializes it with modern corporate organization.
Financing History: Angel round December 2024, Angel+ round April 2025, Series A of several hundred million RMB March 2026.
Core Highlights: The world's smallest planetary roller screw—1.5mm diameter, 5.5mm nut, thinner than a toothpick, yet can stably output 10kg load force, lifespan exceeding 3 million cycles, unit price locked below 100 RMB (industry sample price above 10,000 RMB). Through self-developed high-precision metal forming processes, production efficiency is increased 10-fold. Currently, humanoid robot and automotive线控底盘business each account for half, giving it more容错空间in industry rhythm with this dual-engine驱动structure.
Chapter 4: Divergence, Convergence, and Irreversibility in the Dexterous Hand Track
The dexterous hand track is undergoing a structural divergence.
The Hybrid Model of "Ontology In-house Development + Supplier External Procurement" Will Become the Industry Standard
Financing data from 2026 shows that all 15 ontology manufacturers have clearly committed to in-house dexterous hand development, collectively securing约200 billion RMB in financing—this could easily be interpreted as negative for suppliers. But the reality is not a simple "binary opposition."
The key is that the in-house development by ontology manufacturers is concentrated on the dexterous hand body, such as mechanical structure and drive control. However, for underlying components like tactile sensors (Pasin AI, Dimeng Robotics, Qianjue Robotics), micro motors (Wuji Technology, Zhijian Zhiqing), and precision lead screws (Nuoshi Robotics), the vast majority of ontology manufacturers still rely on external supply.
The technical barriers and mass production know-how for these components are not automatically acquired just because you are an OEM; in fact, companies that make complete machines often are not good at making parts—this is an iron law of manufacturing.
The industrial structure of dexterous hands is unlikely to be completely internalized by OEMs like power batteries, but more likely类似to automotive electronics: Tier1 (dexterous hand complete modules) and Tier2 (core components) compete in layers, each with its own生态位.
Ontology Manufacturers Receive "Track Valuation", Suppliers Receive "Product Valuation"—The Two Are Not Comparable
Among the 74 financing events in 2026, ontology manufacturers had 23 events for ~200 billion RMB, suppliers had 51 events for ~85.1 billion RMB—the former's average per event is over 4 times that of the latter. But this does not mean suppliers have lower investment value.
It's primarily because their valuation logics differ: The former corresponds to the expectations of the entire embodied intelligence track, betting on the direction; the latter corresponds to the market share and revenue growth of a specific product, betting on certainty.
In the next two years, as ontology manufacturers successively enter the上市or Pre-IPO stage, their valuations will face the "big test" of业绩检验; while suppliers that achieve profitability and scale may end up more stable than many OEMs.
The Supplier Track Will Experience a Brutal Shakeout, with the Dexterous Hand Complete Machine Category Being the First to Bear the Brunt
Among the 32 suppliers, nearly 20 claim to be working on dexterous hand complete machines—this number itself说明问题. Dexterous hand complete machines are a category with极强的scale效应: The larger the capacity, the higher the yield, the lower the unit price. Small and medium players without sufficient shipment volume support can be拖垮by模具摊销and production line depreciation alone.
AGILINK and Lingxinqiaoshou are already碾压competitors with capacity and price. Xynova has极高的order certainty backed by Xiaomi. By the end of 2027, the number of effective players in the dexterous hand complete machine category may大幅缩减.
The "Hidden Champions" in Tactile Perception and Core Components Will Enjoy the Longest红利期
Unlike dexterous hand complete machines, in the two sub-directions of tactile sensors and micro motors/lead screws, each leading company has built a 3 to 5-year first-mover barrier in a highly vertical technical point.
Pasin AI has over 5 years of积累in tactile sensing; Dimeng's open-sourced data assets have network effects; Wuji Technology's motor power density is difficult to模仿in the short term; Nuoshi's micro lead screw process has nearly 50 years of family know-how传承.
The common characteristics of these companies are: The track is small enough that big players are unwilling to enter; the technology is deep enough that new players find it hard to catch up quickly; the customers are集中enough that once bound, replacement is difficult. They may not become百亿级platforms, but the certainty of becoming "small and beautiful" companies in the十亿to数十亿级range is极高.
This article is from the WeChat public account "IT桔子" (ID: itjuzi521), author: Wu Meimei.
