文 | 机器最前线
2026年的具身智能赛道，融资风向正在发生变化。
IT桔子统计显示，2026年10月具身智能赛道发生33起融资、总额101.05亿元，而9月这个数字是60起、110.28亿元，下降趋势显著。
不过，资本追逐本体制造和大模型的热闹场面虽然有所降温，但还有一部分投资人开始选择把目光向上游迁移，灵巧手、触觉传感器、电子皮肤这些决定机器人“身体能力”的核心零部件，成为少数仍在逆势吸金的环节。
10月29日，山东柔性织物触觉传感企业矩侨工业宣布完成亿元A轮融资，由蚂蚁集团领投，老股东零以创投、威海港融及青岛和信跟投。这是蚂蚁在触觉赛道上的第二次出手，两个月前，它刚领投了视触觉路线的戴盟机器人。
可需要注意的是，触觉赛道技术路线高度分散，一家成立不到三年的公司拿到亿元级融资，究竟是行业拐点的信号，还是阶段性的资本卡位？
01 具身智能的融资逻辑变了
具身智能的融资逻辑似乎正在从“造一个聪明的身体”转向“让身体感知世界”。
毕竟，视觉可以靠互联网视频喂养，语言可以靠文本喂养，唯独触觉数据几乎无法从现成语料中获得。切向力、材质纹理这些物理量，仿真引擎很难高保真地生成，必须依靠真实操作一帧一帧采集。
有行业数据显示，在价格方面，触觉数据的价格比视觉数据多两倍左右，视觉等视频数据价格可能在小几百元每小时，而触觉数据价格在大几百元。
这种差异直接反映在融资数据上。2026年前七个月，国内人形机器人传感器领域发生了近20起融资，合计超过31亿元，仅7月单月就吞下约15.6亿元。某种程度上，这也意味着，接下来谁能提供高效率、低成本、可规模化的触觉数据采集入口，谁就卡住了现阶段具身智能的“数据隘口”。
不过现阶段国内触觉赛道并非只有一条统一的技术路线，以帕西尼为代表的霍尔磁传感方案，偏向灵巧手指尖的高精度多维力检测；以戴盟机器人为代表的视触觉路线，分辨率更高，触觉分辨率能做到每平方厘米数万个感知点，输出图像数据，天然适配端到端网络。
还有一种就是以矩侨工业为代表的织物路线。它的技术逻辑是把导电、压阻结构一路拆到纤维层面，用功能纤维直接织成传感网络，外力触碰时纤维内的导电通路变形，采集系统读信号、算法转成触觉数据。
这三条路线不是互相替代的关系，反而更像各自对应不同的任务。比如指尖和灵巧手要求高分辨率、多维力与快速响应，视触觉和霍尔磁传感方案现阶段更有优势；躯干、机械臂、足底等大曲面则需要大面积、可弯折、透气和成本可控的覆盖，织物路线似乎更容易铺开。
所以短期内很难有一条路线通吃，行业更可能走向按部位分层的混合方案。也正因如此，资本的下注逻辑不是赌某一项技术最终胜出，蚂蚁的两次出手恰好对应了这种分化，戴盟机器人代表视触觉的高精度方向，矩侨工业代表织物路线的大面积覆盖能力。这也解释了为什么在具身智能融资降温的背景下，一笔亿元级资金仍会落在“一块布”上。
02“反常识”的技术路径
资本同时下注两条路线，并不等于两条路线都已跑通。要判断这笔亿元级融资是行业拐点的信号，还是阶段性的资本卡位，需要回到矩侨工业本身。
这家公司成立于2023年底，创始人张景淇毕业于英国约克大学电子工程系，曾担任小米生态链公司CTO，拥有多项相关发明专利。核心团队也主要来自中科院及智能硬件大厂，具有深厚的技术积累以及智能硬件产品定义和量产经验。
公司的技术路径颇具“反常识”色彩。传统压阻薄膜的结构，是将柔性基底、感压层、导电层逐层叠加在一起，但矩侨工业选择以尼龙作为纤维内芯，把压阻材料、电极分别做成独立功能纤维。
这套方案的优势在于，其织出来的传感布料可裁、可折、可快速改版，能贴合机器人复杂曲面。这一创新成功推动织物形态电子皮肤真正走向规模化量产。2025年5月，公司对外披露卷对卷产线年产能20万片、良率95%；7月，威海临港区报道订单已排到10月，投产首年产值近千万元；9月，第1000套人形机器人织物电子皮肤下线，公司称产品上市不足半年，已有近百家海内外客户下订与复购，包括但不限于宇树、松延动力、零次方等。
不过，织物路线的短板同样明显，其单点精度不及视触觉方案，决定了它更适合躯干、足底、机械臂等大面积覆盖场景，而非指尖高精度操作。所以公司也在向精细交互延伸，推出了触觉皮肤2.0，通过调整织造结构并结合模型训练，进一步拓展对受力方向、滑动趋势和切向受力的感知能力，但能否在精度上满足灵巧手的高端需求，还有待验证。
更深层的挑战在商业模式上，据悉，矩侨工业当前的产品矩阵横跨三条赛道，其中，机器人电子皮肤作为估值锚，确立了公司的技术标杆地位，不过现阶段业务高度依赖宇树等少数头部客户，订单持续性有待观察。
智慧康养作为现金流业务，2024年订单已超千万元，并落地泰康之家、国投健康、华润润家等知名机构；至于汽车智能座舱则是作为增长杠杆，支撑业绩。
因此，这笔亿元级融资更准确的含义，不是资本已为矩侨工业的织物路线投下“确认票”，而是在触觉技术路线尚未收敛、数据入口尚未定局的窗口期，提前卡位一个具备量产交付能力的差异化玩家。
它真正要检验的，是公司能否把先发的工艺优势转化为稳定交付、标准化模块和模型可用的触觉数据；若这一步走通，织物路线有机会成为机器人全身触觉的基础设施，若走不通，资本窗口也可能被后来者迅速压缩。
03 谁能抢占人形机器人感知制高点？
矩侨的融资之所以会引起行业的关注，背后其实折射出了机器人皮肤正在发生的变化。
过去，机器人穿衣服或包覆软皮，更多被理解为外观和防护问题。现在，这层外覆盖件正在同时承载安全、亲和、场景适配和触觉感知四重功能。
在商场、康养、家庭等近人场景，用户碰到的是坚硬金属还是柔性包覆，直接决定人机共处的体验；在工业场景，耐高温、防腐蚀、防粉尘、防静电是刚需……
这也使得赛道增速尤为显著，根据行业预测，2030年我国人形机器人电子皮肤市场需求将达152.5万平方米，对应整体市场规模约274亿元。传统纺织企业、专业传感器企业、材料企业正在加速入局，产业链上下游联动效应明显。
可热度背后，产业发展仍处于早期。标准化缺失是最突出的痛点之一，不同品牌、不同型号的机器人结构各异，电子皮肤大多需要一机一版定制开发，模具和版型无法批量复用，成本高、周期长。
高端电子皮肤模组受封装工艺、传感芯片、材料成本限制，全套搭载价格仍让主打性价比的整机厂难以承受。此外，柔性传感产品在长期高频弯折、复杂工业粉尘、温变工况下，容易出现信号漂移、灵敏度下降、使用寿命缩短等问题，产品稳定性和一致性有待提升。机器人皮肤要真正从选配变成刚需，还要跨过一致性、可靠性和成本这三道门槛。
总而言之，一台机器要进入人的空间，最先被感知的往往不是算法和参数，而是它碰到人时是硬是软、是冷是暖，触觉的价值也正在于此。
但这条赛道的成熟标志，不会是一笔融资或一款新品，而是皮肤变成不显眼却不可缺的普通零部件。到那时，讨论的重点不再是技术路线谁更先进，而是谁能在真实工况下稳定供货、控制成本、配合整机迭代。资本今天押注的，正是这种确定性。至于哪条路线最终胜出，或许要等机器人真正走进日常之后才有答案。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Man, the front line of the machine.
In 2026, on the smart track, the direction of financing was changing.
According to IT orange statistics, 33 cases of financing in October 2026, with a total of $10.105 billion, occurred on a smart track, compared to 60 cases and $11.128 billion in September, a significant downward trend.
However, while there has been some cooling in the capital chase and the hot scene of LLM, some investors have started to choose to move their eyes upstream, and the core parts that determine the robot's “body capacity” — smart hands, tactile sensors, electronic skin — have become a few links that are still in reverse.
On 29 October, the Shandong Softweed Sensor Sensory Industries announced the completion of the Billion A round of financing, with the ants group taking the lead, with the old shareholders at zero-to-one, the port of Wei and Qingdao and Tsing Dynasty. This is the second time an ant is on the tactile track, and two months ago, it just took the UFO on the visual touch path.
It should be noted, however, that the technological route of the tactile track is highly dispersed, with a company that has been in existence for less than three years receiving billions of dollar-level financing, whether it is a sign of a trade turning point or a stage capital card.
01 Smart finance logic changed.
Smart finance logic seems to be moving from “building a smart body” to “making the body feel the world”.
After all, visuals can be fed on Internet video, languages can be fed on text, and only tactile data can hardly be obtained from ready-made text. It is very difficult for a simulation engine to produce these physical amounts of stride, materials and texture, which must be collected within a frame of actual operation.
In terms of price, there are industry data showing that tactile data are about twice as high as visual data, and video data such as visual data can be priced at hundreds of dollars per hour, while tactile data are priced at hundreds of dollars.
This difference is directly reflected in the financing data. In the first seven months of 2026, nearly 20 cases of financing took place in the area of human robotic sensors in the country, totalling more than $3.1 billion, and approximately $1.56 billion was swallowed in July alone. To a certain extent, it also means that the next person who can provide an efficient, cost-effective, scalable access to the tactile data will be stuck in the “data portal” that is smart at this stage.
At this stage, however, the domestic tactile track is not only one single technical route, the Hole Magnetic Sensor Program, represented by Pasini, which favours high precision multi-dimensional testing of the sharp fingertips; the visual tactile route, represented by resolution, is higher, and the tactile resolution is capable of tens of thousands of sense points per square centimetre, output of image data, natural adaptation to end-end networks.
There is also a fabric route represented by the Rectangular Industries. Its technical logic is to tear the conductive, pressure-retarding structures all the way to the fibre level, to woven directly into a sensor network with functional fibres, to deform the conductive circuits in fibres when external forces touch, and to collect system reading signals, algorithms and transposes to tactile data.
These three routes are not supplanting each other, but are more like their respective mandates. For example, fingertips and smart hands require high resolution, multi-dimensional and rapid response, and visual touch and hormagnetic sensor programmes are more advantageous at this stage; large curvatures such as torso, mechanical arms, and bottoms require large, bendable, air-traffic and controlled coverage, and fabric routes appear to be easier to spread.
It is therefore difficult to find a route to eat in the short term, and the industry is more likely to move towards a tiered mix. This is why the logic of capital is not to bet on the ultimate victory of a technology, and the ants' two hands correspond to this dichotomy, with the unicorns representing the high-precision direction of touch, and the rectangular industry representing the mass cover of the fabric route. This explains why, in the context of the cooling down of smart finance, a billion dollar-level funding still falls on a “single sheet”.
02 Technical Path to Unusual
The fact that capital is on two routes at the same time does not mean that both routes are running. To judge whether this billion dollar-grade financing is a sign of an industry turning point or a staged capital card, it needs to be returned to the Rectangular Industry itself.
Founded at the end of 2023, its founder, Zhang Kyung-guk, graduated from the Department of Electronic Engineering at York University, United Kingdom, and has served as CTO, the Micro-Mi Eco-chain Company, with patents on numerous related inventions. The core team also comes mainly from the Central College of Sciences and the major smart hardware plant, with deep technological accumulation and experience in the definition and volume of smart hardware products.
The company's technological path is quite “unusual”. The traditional pressure-retarded membrane structure is a layer-by-soft collage of soft base, pressure layer, and conductor layer, but the Quasi-based industry has chosen to use nylon as the inner core of the fibre, with the pressure-retardant material and electrodes as separate functional fibres.
The advantage of the package is that the sensory fabric it fabrics can be tailored, folded and quickly adapted to fit the complex curves of robots. This innovation has been successful in driving fabric-form electronic skin to a real scale of mass production. In May 2025, the company ' s external disclosure paper had a capacity of 200,000 tablets per year and a good rate of 95 per cent for the Voltaline industry; in July, it had been scheduled to October, with the initial production of nearly tens of millions of yuan; and in September, it had ordered and repurchased nearly 100 domestic and external customers, including but not limited to the Okaki, pine-side power, zero-side, etc., for the 1,000th set of human-shaped robotic fabric electronic skins.
However, the short board of the fabric route is also clear, with a single point of precision less than the visual touching programme, which determines that it is more suitable for a wide area of coverage, such as the torso, foot bottom, mechanical arm, and so forth, rather than a pointy high-precision operation. The company is therefore also extending to fine, interactively, with the introduction of tactile skin 2.0 and further expanding the sensory ability to absorb, slide, and cut to stress by adjusting the fabric and combining model training, but the ability to meet the high-end needs of smart hands in precision is yet to be verified.
Deeper challenges lie in business models, where it is known that the current product matrix of the Rectangular Industries spans three tracks, where robotic electronic skin is used as a valuation anchor to establish the company ' s technical niche, although at this stage operations are highly dependent on a small number of front-line customers, such as Woo-tree, and the continuity of the order remains to be observed.
As a cash-flow business, the order in 2024 exceeded millions and landed in well-known institutions, such as Tacon House, National Investment for Health, and Hua Yunyun; as for car-smart cabins, it was a growth lever that underpinned performance.
As a result, this billion dollar-scale financing means more precisely not that capital has been “confirming” the fabric route of the rectangular industry, but rather that the early carding of a differentiated player with the capacity to deliver quantitatively during a window period in which the tactile technology route is not condensed and the data entry is still open.
What it really wants to test is whether the company can translate the pre-emptive process advantage into stable delivery, standardized modules and the tactile data that models can use; if this move goes, the fabric route has the opportunity to become a robotic sensory infrastructure, and if it fails, the capital window can be rapidly compressed by laterers.
03, who can take the human robot's sense?
The financing of the Rectangulars is of interest to the industry and reflects the changes that are taking place in robotic skin.
In the past, robots were dressed or covered in soft skin, which was more understood as appearance and protection issues. Now, the exterior cover is carrying a four-fold function of safety, kinship, scene suitability and sensory awareness at the same time.
In the near-human scenes such as malls, rest and well-being, users encounter hard metal or softness, which directly determines the experience of human coexistence; in the industrial scenes, high temperature resistance, corrosive resistance, dust resistance, electrostatic resistance, etc.
This has also led to a particularly significant increase in the number of race lanes, which, according to industry projections, will require 1,525,000 square metres of electronic skin for human robots in 2030, corresponding to an overall market size of approximately $27.4 billion. Traditional textile enterprises, professional sensors and material enterprises are accelerating their entry, and the upstream and downstream linkages of the industrial chain are evident.
Behind the heat, industrial development is still at an early stage. The lack of standardization is one of the most striking pains, with different brands, different types of robotic structures, most of which require one-size-size-fits-all customized development for electronic skin, non-quantitative replicability of molds and layouts, and high costs and long cycles.
High-end electro-skinners are subject to containment processes, sensor chips, and material costs, and the full load price continues to be difficult for the main cost-effective whole plant. In addition, flexible sensor products are subject to long-term high-frequency bends, complex industrial dust, temperature conditions and are prone to signal drift, reduced sensitivity and reduced useful life, with product stability and consistency to be enhanced. For robotic skin to truly move from choice to demand, there are three thresholds of consistency, reliability and cost.
In sum, when a machine enters human space, it is often first perceived not by algorithms and parameters, but by its hard, cold or warmness when it comes to humans, and by the value of the touch.
But this track is not a mature sign of a single financing or a new product, but a common part of the skin that becomes invisible but indispensable. By that time, the focus of the discussion would no longer be on who was more advanced in the technical route, but who could stabilize the supply, control the costs and cooperate with the whole process under real conditions. It is this certainty that capital is betted today. As to which route eventually wins, perhaps the answer is not until the robot actually walks into the day.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.