Last reviewed
May 19, 2026
Sources
12 citations
Review status
Source-backed
Revision
v1 · 2,531 words
Add missing citations, update stale details, or suggest a clearer explanation.
| Xynova | |
|---|---|
| General information | |
| Type | Private company |
| Industry | Robotics, embodied AI components |
| Founded | End of 2024 |
| Headquarters | Hangzhou, China |
| Key technology lineage | Team derived from research group of Xia Changliang, academician of the Chinese Academy of Engineering |
| Products | Flex 1, Flex 2, micro electric cylinders, hollow cup motors, planetary roller screws, joint modules |
| Notable investors | CATL Capital, Xiaomi Strategic Investment, CETC, Zhengxuan Investment, Orient Renaissance Capital, SEARI Capital, L2F Ray Entrepreneur Fund |
| Production target (2026) | 10,000 dexterous hands and 200,000 micro electric cylinders per year |
| Website | xynova.com.cn |
Xynova (sometimes rendered Xynova Future) is a Hangzhou-based Chinese robotics startup that develops high-degree-of-freedom dexterous hands and the underlying actuator stack used in humanoid robots and embodied AI systems. The company was founded at the end of 2024 and rose to prominence within its first year of operation through the launch of two dexterous hand products, the Flex 1 in August 2025 and the Flex 2 in May 2026, and through high-profile angel and Pre-A funding rounds backed by Contemporary Amperex Technology Co., Limited (CATL), Xiaomi, and the China Electronics Technology Group Corporation (CETC). [1][2][3]
Xynova is one of a small group of Chinese suppliers, including Inspire Robots, Linkerbot, Wuji Tech, and DexRobot, that compete at the high-DOF end of the robotic hand market. Its strategic distinction is that it manufactures essentially the entire bill of materials in-house, including brushless motors, motor controllers, planetary roller screws, reducers, the tendon transmission, and the control algorithms. The company describes itself as targeting both factory automation and household humanoid applications, and by early 2026 had reportedly secured orders for more than 10,000 hands from leading humanoid integrators. [1][4]
Xynova was incorporated in Hangzhou at the end of 2024. The core engineering team carries over twenty years of research and development experience in dexterous hand design and high-density electric drive systems, with the lineage traced publicly to the laboratory of Xia Changliang, an academician of the Chinese Academy of Engineering and current president of Tiangong University, whose group focuses on electrical engineering and intelligent control. The technical roots in academic electric drive research are reflected in the company's vertically integrated approach to motors, reducers, and screws rather than relying on off-the-shelf actuator suppliers. [1][2]
The company operates an integrated production line that covers machining, motor winding, assembly, and end-of-line testing for its hands and core components. Xynova describes its capability set with the formula "motors, electronic controls, reducers, screws, and algorithms," and uses that vertical integration to argue that it can hit cost, weight, and reliability targets that pure system integrators cannot. [1][4]
Xynova has completed two funding rounds since its founding, both within roughly fifteen months of incorporation.
Xynova closed an angel round of more than 100 million yuan (approximately 13.7 million US dollars) on December 26, 2025. The round was led by CATL Capital, the industrial investment platform of battery maker CATL, with co-investments from Xiaomi Strategic Investment, Zhengxuan Investment, Orient Renaissance Capital, SEARI Capital, and the L2F Ray Entrepreneur Fund. Lighthouse Capital served as exclusive financial advisor. [1][3][5]
Industry coverage at the time of the announcement noted that the size of the angel round was unusual for a company less than twelve months old, and reflected investor appetite for component-layer plays as humanoid robot shipments were forecast to scale through 2026. The capital was earmarked for production line construction, supply chain build-out, and the development of the second-generation Flex product. [1][3]
On March 20, 2026, Xynova announced the close of a Pre-A round of "hundreds of millions of yuan," with the lead investor described in Chinese-language coverage only as a "top-tier internet giant." Existing strategic investors, including Xiaomi Strategic Investment and funds affiliated with the China Electronics Technology Group Corporation, increased their stakes in the round. The published use of proceeds covers talent acquisition, scaling of mass production, commercial deployment with humanoid robot manufacturers, and refinement of the dexterous hand and component portfolio. [2][6]
| Round | Date | Reported size | Lead investor | Notable co-investors |
|---|---|---|---|---|
| Angel | December 26, 2025 | More than 100 million yuan (about 13.7 million US dollars) | CATL Capital | Xiaomi Strategic Investment, Zhengxuan Investment, Orient Renaissance Capital, SEARI Capital, L2F Ray Entrepreneur Fund |
| Pre-A | March 20, 2026 | Hundreds of millions of yuan | Undisclosed "top-tier internet giant" | Xiaomi Strategic Investment (continued), CETC-affiliated funds (continued) |
Xynova maintains two product lines: a finished dexterous hand family and a portfolio of underlying actuator and transmission components. The component portfolio is sold to other robot makers as well as used inside Xynova's own hands.
| Product | Launch | Active DOF | Total DOF | Palm weight | Whole-hand grasp load | Architecture |
|---|---|---|---|---|---|---|
| Flex 1 | August 2025 | 20 | 25 (20 active + 5 passive) | 380 g | More than 30 kg | Pure tendon-driven |
| Flex 2 | May 13, 2026 | Not separately disclosed | 23 | 400 g | 12 kg single-hand grasp; 4 kg rated continuous | Hybrid (cable-driven plus direct-drive) |
The two hands sit at different points on the cost and capability curve. The Flex 1 prioritises raw load capacity and a high active DOF count, and is positioned for research, teleoperation, and integrators who want the maximum possible mechanical envelope. The Flex 2 reduces total DOF count and grasp load in exchange for a hybrid drive architecture, a dedicated "cerebellum"-style control layer, multimodal tactile and force sensing, and tighter precision targets (±0.1 mm repeatability and 0.05 N force control). The Flex 2 is the unit that Xynova has positioned as its primary humanoid integration product. [7][8][9]
Xynova manufactures and sells the building blocks of its hands as standalone components. These are used inside the Flex hand line and are also sold to other robot integrators.
| Component | Specification | Notes |
|---|---|---|
| Hollow cup motors | 8 mm diameter | Brushless, designed for finger-scale joints |
| Planetary roller screws | 7 mm wide | Used inside micro electric cylinders |
| Micro electric cylinders | 10 to 12 mm range, 100 to 300 N thrust | The actuators that drive Xynova's tendon and direct-drive joints |
| Integrated joint modules | 322 Nm per kg torque density | High torque density for humanoid robot wrist and forearm use |
The in-house planetary roller screw and hollow cup motor work is the company's most differentiated capability. These two parts are typically the supply bottleneck for Chinese hand makers, and Xynova's ability to produce them internally is the reason it can target a 200,000-unit annual run rate on micro electric cylinders. [1][6]
Xynova is constructing a production facility designed for an annual run rate of 10,000 high-DOF dexterous hands and 200,000 micro electric cylinders, with commissioning scheduled for the second quarter of 2026. The company has publicly stated that orders for more than 10,000 hands had been secured from leading humanoid robot manufacturers before the production line entered commissioning. The Pre-A round capital is directed in large part toward outfitting and ramping this facility. [4][6]
The target volumes place Xynova among the more aggressively scaled component suppliers in the Chinese dexterous hand market. By comparison, Inspire Robots shipped roughly 10,000 hands during 2025, and Linkerbot reported monthly shipments of more than 1,000 hands across its LinkerHand line during the same period. [4]
Xynova's published positioning emphasises four design pillars, each tied to a piece of in-house intellectual property. [1][7][9]
The company controls the full electromechanical chain from raw motor lamination through reducer, screw, controller, and tendon transmission. The published rationale is that high-DOF hands are particularly sensitive to the matching of motor torque, reducer ratio, and tendon stiffness, and that integrators who buy actuators off the shelf cannot hold those parameters to the tolerances required for sub-millimetre precision. The 322 Nm per kg joint module is presented as evidence that this approach yields specific torque numbers comparable to top-end Japanese and European actuator vendors at lower cost.
Xynova reports that its tendon transmission has been validated past one million open-close cycles at rated load. Tendon durability is a recurring weakness of competing hands; the Wuji Hand article notes that traditional tendon-driven hands often fail after roughly 10,000 grasping cycles, and the Xiaomi CyberOne hand programme cites a 150,000-cycle target as a 15-fold improvement over typical designs. Xynova's published one-million-cycle figure sits significantly above both. [1][10]
The Flex 2 moves away from the pure tendon transmission of the Flex 1 and combines cable-driven joints with direct-drive joints. The hybrid layout is intended to keep the hand light (400 g) and biomimetic while removing some of the cable stretch and friction hysteresis that limit pure-tendon precision. The result, as quoted on Xynova's product page, is ±0.1 mm positional repeatability and 0.05 N force control on a back-drivable mechanism that supports both force and position control modes. [7][8]
The Flex 2 integrates a multimodal sensing layer covering force, tactile, and visual cues, with the wrist-mounted camera positioned at the wrist joint rather than the centre of the palm. Xynova describes a dedicated "cerebellum"-style control layer that handles slip detection, adaptive grasping, and compliant reflexes. The control software stack includes a parameter-optimization algorithm with physical constraints, advertised as enabling adaptive joint control across varying payloads and object shapes. [7][9]
Xynova's investor list, and Xiaomi's continued participation across rounds, place the company at the centre of a wider competition over who supplies the hands of Chinese humanoid robots. Xiaomi is itself building dexterous hands for the CyberOne platform, including an April 2026 bionic hand that introduced full-palm tactile sensing and an evaporative "sweating" cooling channel, and in late 2025 hired Lu Zeyu, a former member of the Tesla Optimus dexterous hand team, to lead in-house hand R&D. [10]
Xiaomi's strategic investment in Xynova therefore covers the component layer below its own integrated hand, and is consistent with the company's broader pattern of putting capital into adjacent suppliers across the "Human x Car x Home" ecosystem. As of May 2026, it is publicly confirmed that Xiaomi Strategic Investment is on the Xynova cap table through both the angel and Pre-A rounds. It is not publicly confirmed that the hand fitted to any Xiaomi humanoid robot uses Xynova technology, and Xynova has not been listed as a hand supplier to CyberOne in Xiaomi's own communications. [3][7][10]
Xynova competes in a small but increasingly crowded high-DOF dexterous hand market. As of the first half of 2026, Chinese-language industry coverage typically groups the leading hand suppliers as follows. [4][11]
| Supplier | Architecture | Active DOF range | Notes |
|---|---|---|---|
| Inspire Robots | Linkage drive | 6 to 12 | Largest by shipped units; about 10,000 hands shipped in 2025 |
| Linkerbot | Tendon (L30) plus other architectures | 6 to 22 | Dominant by revenue in high-DOF; monthly shipments above 1,000 hands |
| Xynova | Tendon (Flex 1), hybrid (Flex 2) | 20 to 23 | Highest published tendon durability; vertically integrated |
| Wuji Tech | In-hand direct drive | 20 (Wuji Hand) | Direct-drive variant; Genesis AI partner |
| DexRobot | Tendon | Unpublished | Smaller-scale Chinese peer |
| Tesla Optimus hand program | Tendon, 25 forearm actuators | 22 (Optimus Gen 3) | Captive, not sold externally |
| Unitree Dex5 | Mixed tendon plus linkage | 20 | Captive; sold with Unitree platforms |
The wider Chinese dexterous hand market shipped more than 30,000 units in 2025, with humanoid-grade hands accounting for an estimated 15,000 units, up from about 2,000 the year before. Forecasts published in early 2026 expect the global market to exceed 1.4 million units and roughly 3 billion US dollars in revenue by 2030. [4]
An industry comment from AAC Technologies characterises the design space as an "impossible trinity" of degrees of freedom, size, and force output, where any two can be maximised but not all three at once. Xynova's two-product strategy can be read as a direct response to this constraint: the Flex 1 optimises DOF and force output at the expense of finesse, while the Flex 2 trades raw force for compactness, sensing, and precision. [4]
Coverage of the Flex 2 launch in May 2026 generally framed Xynova as a serious component-layer entrant rather than another humanoid integrator. The Aihola technical writeup described the Flex 2 as "a deliberate pivot toward sensing and control," arguing that the loss of grasp load relative to the Flex 1 was a chosen trade-off rather than a regression. The Threads coverage from the AI Continuum highlighted the millisecond-scale response time and the combination of slip detection with the 23-DOF biomimetic configuration. [7][8][9]
The RoboHub thread that announced the Flex 2 in English emphasised the broader strategic point: that Xiaomi was putting money "into the component layer that decides whether humanoids can actually handle objects, tools, and daily tasks," and that this is the unglamorous part of robotics that decides what a robot can actually do with its hands. Industry analysts have generally agreed that vertical integration at the component level is becoming a defensible position as the humanoid market scales. [9][12]