In 2025, the sales volume of dexterous hands in China increased to approximately 19,200 units, with leading enterprises subsequently achieving annual deliveries at the 10,000-unit level.
The initial formation of mass production capabilities has shifted the competitive focus and brought the market into a new stage of development. Customers' attention has shifted from the continuously intensifying competition over performance parameters such as degrees of freedom (DOF), tactile sensing, and lightweight design, to operational capabilities and actual benefits under real working conditions.
There is now a preliminary answer to whether dexterous hands can be mass-produced, but whether they can be continuously and stably "put to use" in factories still needs to be tested.
To truly work on the factory floor, humanoid robots still need to bridge the "last centimeter."
From Parameter Competition to Scale Delivery: A Shift in Market Evaluation Dimensions
Over the past period, capital has continuously increased its bets on dexterous hands. According to incomplete statistics from Gaogong Humanoid Robot, as of July 30, 2026, 15 companies in the dexterous hand sector have completed 31 rounds of financing this year, with a cumulative amount exceeding CNY 8 billion, a year-on-year increase of up to 1,500%, surpassing the total number of financing events and amounts for the whole year of 2025. The relevant investors cover market-oriented institutions such as Sequoia China, Beijing Artificial Intelligence Industry Investment Fund, and Dongfang Precision Science, as well as government-backed funds and industrial capital, demonstrating the capital market's multiple recognition of the investment value, policy value, and industrial synergy value of dexterous hands.
In contrast, product performance has been continuously enhanced. Since 2024, the DOF of dexterous hands have evolved rapidly, with parameters once rising from 6 to 15, and now increasingly commonly exceeding 20. Tactile sensing capabilities, which were initially optional only for a few high-end products, are becoming standard for more and more products, with simultaneous increases in tactile dimensions and accuracy. While parameters are rising, the weight of the products is becoming lighter, with some high-performance products reducing their weight to around 400 grams.
From DOF and tactile sensing to product weight, rapidly refreshing parameters continuously raise the performance ceiling of dexterous hands, laying a solid hardware foundation for them to complete more complex and precise manipulation tasks.
Accompanying the influx of capital and performance improvements is the gradual maturation of production conditions, putting the shipment volume of dexterous hands on the fast track. According to data from the "2026 H1 China Dexterous Hand Industry Research Report" by Gaogong Industry Research Institute (GGII), the sales volume of the Chinese dexterous hand market was approximately 19,200 units in 2025, a year-on-year increase of 236.84%. It is expected to reach 70,200 units in 2026, a year-on-year increase of 265.63%.
The emergence of 10,000-unit-level deliveries means that dexterous hands have obtained the "ticket" to scale manufacturing. Fang Hainan, CMO of INSPIRE Robots, told reporters from China Electronics News: "From an industry perspective, the scale shipment at the 10,000-unit level confirms the engineering implementation capabilities and supply chain maturity of domestic dexterous hands, laying a core scale foundation for subsequent cost reduction, efficiency improvement, and large-scale scenario applications in the industry." From the procurement of core components and product assembly to quality inspection, dexterous hands are gradually bidding farewell to small-batch trial production dominated by manual debugging and moving towards standardized and scale manufacturing.
As mass production capabilities gradually mature, the focus of market evaluation is beginning to change. Dr. Xu Chunshan, founder and chairman of Psibot, told reporters from China Electronics News in an interview: "Currently, the industry has a tendency to 'emphasize parameters over capabilities,' and many product evaluations still focus on indicators such as DOF, weight, and payload. However, as products gradually enter the scenario validation stage, customers' selection criteria are undergoing profound changes. Stability under real working conditions, scenario adaptability, and full-stack delivery capabilities have become the core concerns of downstream customers."
From "competing on parameters" to "competing on delivery," the dexterous hand market is crossing a critical node from laboratory validation to scale application. However, the expansion of delivery scale only answers whether dexterous hands can be mass-produced, and has not yet answered whether they can truly be "put to use" after being produced.
Which customers exactly are the 10,000 dexterous hands flowing to, and how many of them have entered real operational scenarios to stably create production value, has become the next key question that needs to be asked.
Still in the Early Stage of Industrialization, with Return on Investment as the Core Consideration
The achievement of 10,000-unit-level orders for dexterous hands is closely related to the expansion of humanoid robot production, as well as the growth in demand for scientific research and education, data collection, and factory operations.
Xu Chunshan told reporters from China Electronics News: "The three major demands currently driving sales are mainly: first, the supporting demand for humanoid robot bodies; second, the demand for scientific research experiments and data collection; and third, the initial introduction into industrial scenarios, such as small-batch deployments in fields like automobile manufacturing, 3C electronics, precision assembly, and power line inspection." According to the "2026 China Dexterous Hand Industry Development White Paper" released by MIR in September this year, in the application scenarios of dexterous hands in 2025, scientific research and education and data collection accounted for a combined 74.4%, entertainment performances accounted for 8.3%, and industrial scenarios accounted for the lowest proportion, at only 3.5%.
Nearly three-quarters of applications in scientific research and education and data collection, compared to less than 4% in factory applications, expose the problem of structural insufficiency in the supply of dexterous hands. Xu Chunshan pointed out: "The supply of general-purpose, low-cost dexterous hands is relatively sufficient, but high-reliability products that can truly operate stably in industrial scenarios for the long term remain scarce. The industry does not lack 'hands that can perform demonstrations'; what it lacks are 'hands that can work stably.'"
Dexterous hands are still in the early stage of industrialization, transitioning from the laboratory to real scenarios, and the rigid demand in industrial scenarios has not yet been fully released. INSPIRE Robots delivered over 10,000 dexterous hands throughout 2025. According to GGII data, this delivery volume ranked first among Chinese dexterous hand enterprises that year. However, when it comes to specific scenario segments, the application progress on the industrial end is still relatively slow. Fang Hainan stated: "At this stage, INSPIRE Robots' customers are mainly robot body manufacturers, scientific research and education institutions, and algorithm data collection clients. Industrial scenario customers are in the POC (Proof of Concept) validation and penetration stage."
Meanwhile, affected by multiple factors such as scenario complexity, algorithm adaptation, and production line transformation, Fang Hainan pointed out: "The cycle from sample delivery and POC validation to formal production line implementation for dexterous hands varies greatly. The industry has not yet formed a standardized and fixed implementation cycle."
Technical parameters such as DOF and tactile points must ultimately be reflected in work performance in specific scenarios. The head of dexterous hand technology and products at PAXINI introduced to reporters from China Electronics News: "After truly entering practical applications, customers pay more attention to the comprehensive performance of products in real and complex scenarios. For example, when facing objects of different materials, shapes, and weights, whether stable grasping and precise force control can be achieved; whether performance can remain consistent during high-frequency repetitive operations; and at the same time, whether the product is easy to integrate, maintain, and develop secondarily. These questions actually point to the same core: a dexterous hand must not only 'be able to complete a task once,' but also be able to continuously and stably complete tasks in a real environment."
After crossing the threshold of operational stability, customers also need to consider a core indicator—Return on Investment (ROI). Fang Hainan frankly admitted: "Among the industrial validation clients engaged by INSPIRE Robots, the proportion of clients who achieve repeat purchases and small-batch procurement after completing POC validation is limited. The core reason is also the need to calculate ROI, so a willingness for scale procurement has not yet been formed."
For production-end clients, as a new type of "means of production," whether a dexterous hand is "cost-effective" will directly affect or even determine their procurement decisions. Enterprises need to start from multiple levels such as long-term and short-term benefits to compare the comprehensive costs and actual benefits between dexterous hands and skilled workers. Fang Hainan believes that for production scenarios, dexterous hands have not yet fully reached a mature state at the product level.
Moving Towards Scale Application, Engineering Closed-Loop Capability Remains Key
Currently, dexterous hands have not yet formed an industrialization closed loop covering performance, cost, reliability, hardware-software synergy, and standard validation, and the engineering maturity is still insufficient.
Performance, reliability, and cost constitute the "impossible triangle" that dexterous hands cannot avoid. Xiong Kun, CTO of Critical Point, told reporters from China Electronics News: "The essence that restricts the large-scale delivery of dexterous hands lies in the fact that dexterous hands are currently the most extreme systems in the field of precision mechanical electronic control and sensing, with 'the most severe volume constraints, the highest integration, and the most frequent dynamic environment interactions.'"
Placing dozens of micro joints, drive, transmission, sensing, and control units in a "palm-sized" space, while also taking into account heat dissipation, wiring, and structural strength, poses extremely high requirements for enterprises' system integration and engineering capabilities.
Meanwhile, the improvement in performance such as DOF and tactile sensing will also increase process complexity, thereby posing higher requirements for long-term operational stability and cost control. Xu Chunshan pointed out: "The market price of high-DOF tactile dexterous hands ranges from tens of thousands to over a hundred thousand CNY, but their actual lifespan under continuous industrial operation is limited. Customers face the dilemma of 'cannot afford it, and it is not durable either.'"
The market is still in its infancy, and the hardware and software have not yet been "well integrated." On the one hand, the bottleneck of "having hands but no brain" remains prominent. Xu Chunshan told reporters: "Currently, embodied intelligence large models still face significant challenges in driving high-DOF dexterous hands to complete complex and precise operations, and the open-source real data for dexterous manipulation is also relatively limited."
On the other hand, mature hardware products that can support model R&D and be widely adopted by researchers are still absent. Chen Zhe, founding partner of Alphaist Partners, stated in a previous dialogue that there is also an ecological niche in the dexterous hand market benchmarking against "Unitree G1," meaning there should be a product that can become the preferred solution for all dexterous hand researchers in China and the US, thereby forming a de facto industry benchmark.
However, this "throne" is currently still vacant. To this end, some dexterous hand enterprises regard full-stack in-house R&D from hardware to models as a breakthrough path. But from the perspective of industry synergy, vertical integration within enterprises is still difficult to eliminate the differences between different hand types, interfaces, and data systems. The cross-platform reuse of models and data still relies on the construction of industry standards and an open ecosystem.
Meanwhile, as dexterous hands enter real working environments, the issue of scenario adaptation is also placed before enterprises. Xu Chunshan pointed out: "Dexterous hands are not standard off-the-shelf products; more often, they require deep adaptation for specific scenarios." Xiong Kun told reporters from China Electronics News that, based on the clients currently engaged by Critical Point, the demand focus of clients in different fields varies: "Clients leaning towards academic research may pay more attention to performance parameter indicators such as DOF and sensing capabilities of dexterous hands, as well as the friendliness of secondary development interfaces; while industrial and commercial clients value reliability, lifespan, cost-effectiveness, MTBF (Mean Time Between Failures), and maintainability more."
A dexterous hand that pushes performance to the extreme does not equal one that can maximize efficiency in actual work. When purchasing, customers often need to make corresponding trade-offs among indicators such as payload, accuracy, speed, reliability, and cost, and enterprises also need to possess mature solution customization and full-stack delivery capabilities.
In addition, when truly engaging with demands, dexterous hand enterprises also face the dilemma of lacking a "common language" with customers. Xu Chunshan observed: "Regarding the conditions for scale replication, the most critical constraining factor at present is that the industry has not yet formed a unified standard and evaluation system for dexterous manipulation capabilities, and different scenarios lack a quantifiable measuring scale for 'how much work can be done.'" This situation not only increases the cost for customers to compare horizontally, select products, and conduct on-site validation, but also makes it difficult for suppliers to clearly demonstrate the real capabilities of their products in specific working conditions, invisibly increasing the communication costs for both supply and demand sides.
For humanoid robots to truly "get their hands on" work, they still need to bridge this closest yet farthest "last centimeter." The shift from parameter competition to capability comparison, from product delivery to full-stack services, and from general solutions to scenario definition will be the three core shifts in future industry competition; and the supplementation of relevant evaluation standards will provide important support for industry development in regulating market competition and promoting supply-demand matching.
Regarding the future development direction of the industry, the head of dexterous hand technology and products at PAXINI judged: "In the next 2 to 3 years, industry competition will inevitably evolve towards a dynamic balance of 'algorithm, cost, and performance.' Products that can truly build barriers are not necessarily the ones with the most aggressive parameters, but must be the ones that can best achieve scale implementation."
Authors | Lu Mengqi, Hu Xiuqing (Intern) Editors | Qiu Jiangyong Art Design | Ma Liya Supervisor | Lian Xiaodong