初创公司尽调
尽调报告 Robotics / Hardware Series A+ private 2026-10-15

Agilink

灵巧操作业务分拆公司

Agilink 早期商业化势头强劲,也受到投资者追捧,但独立验证仍然不足,$1B+ 的估值显得偏高。

覆盖范围与披露说明

本轮更新基于截至 2026-10-15 收集的公开证据。产品定位、市场边界、竞争和融资信号的覆盖最充分;收入、ASP 和独立客户集中度仍未公开。

封面要素

估值 01
$1B

公司概况

Agilink 是一家总部位于上海、从 Agibot 分拆出来的灵巧操作公司,向具身 AI、科研和工业自动化流程销售 OmniHand 灵巧手与 OmniPicker 夹爪。

官网
www.agibot.com
创立地点
Shanghai, China
总部
Shanghai, China
产品
灵巧机械手、夹爪及相关集成方案。
客户
具身 AI 团队、机器人 OEM 与集成商、科研实验室及工业自动化买方
商业模式
部件销售加销售驱动的集成支持
阶段
Series A+ private
融资情况
分拆后完成四轮融资;据报道估值超过 $1B

执行摘要

主要优势

  • 聚焦灵巧手和夹爪,集成流程清晰
  • 接入 Agibot 生态,加上 2026 年融资和估值上升,释放出强烈信号
  • 覆盖科研、OEM 和工业场景的商业化叙事正在展开

主要风险

  • 仍未披露独立口径的收入、ASP 和复购订单数据
  • 对母公司生态的依赖,使护城河的持久性难以与渠道拉动区分开
  • 关于规模、可靠性和客户结构的公开证据仍然有限

未决问题

  • 按 OmniHand 和夹爪产品系列拆分、且经过验证的收入、毛利率、ASP 和复购订单数据。
  • 客户名单、母公司与第三方客户的构成,以及 Agibot 内部项目与外部买家之间的出货量集中度。
  • Agilink 与 Agibot 之间的股权结构表、董事会构成、投资者权利,以及转让定价或 IP 共享安排。
  • 除 Longcheer 相关证据外,针对 OmniHand 3 及生产规模部署的独立可靠性、耐久性和安全性验证。

目录

Chapter 01

01公司概览

1.1 身份、定位与已披露经营规模

Agilink 当前的官方定位集中在灵巧操作系统,而不是通用人形机器人:公司称其为科研、工业自动化和服务机器人设计并制造高自由度灵巧手与自适应夹爪。官方渠道当前主推 OmniHand 3 系列和 OmniPicker 3,公开产品目录强调应用适配、安装、电源、通信和控制软件集成。规模方面,目前最醒目的数字仍是公司自称,而非独立基准验证:Agilink 主页宣称 2026 年第一季度市场份额第一、已发货 20,000+ 件产品,并积累 50,000+ 小时真实世界数据。这些说法说明商业化和部署野心确实存在,但还不能替代经验证的收入、客户集中度或复购披露。[CO001, CO002, CO006, CO007, CO016]

快照 KPI 表
指标数值 / 状态日期置信度缺口
公司成立 / 独立2023 年 8 月在 AgiBot 内部成立;2026 年 1 月成为独立公司2023-08 / 2026-01高独立经营历史很短。
总部 / 覆盖范围上海总部;官方关于页面显示北京、深圳、新加坡和海外地区当前中未披露员工或办公室数量拆分。
估值2026 年 5—6 月融资报道中超过 US$1 billion2026-05 to 2026-06中公开来源未披露条款、股价或稀释。
规模指标主页称已发货 20,000+ 件产品、积累 50,000+ 小时真实世界数据;Baidu Baike 称截至 2026 年 6 月 OmniHand 交付 8,000+ 台、夹爪发货 10,000+ 台2026-06 / 当前低指标使用不同分母,不应视为经审计收入代理指标。
团队规模官方关于页面称当前 60+ 人,计划在 2026 年第四季度扩大至 100+ 人当前中前瞻目标是公司自称,未经验证。
收入 / 客户本组证据未公开当前低需管理层 KPI 包,覆盖收入、ASP、复购和客户集中度。

公开快照混合了公司官方说法和第三方融资摘要;等同于 null 的状态反映证据缺口,而不是零值。

[CO003, CO004, CO007, CO008, CO009, CO010, CO016]

1.2 领导层、治理与母公司依赖

本组证据中最清晰的创始人信号来自 Xiong Kun;Baidu Baike 和 Sina 都将其与 Agilink 分拆及其此前在 Tencent Robotics X 和机器人工程领域的经历联系起来。这为灵巧末端执行器提供了可信的创始人技术—市场匹配,但也凸显关键人物依赖:公开记录对创始人的信息远多于对管理团队或董事会监督的披露。治理可见度仍不完整。Baidu Baike 将 Wang Chuang 列为法定代表人,而 Sina 2026 年 5 月文章称为 Deng Taihua,基础公司控制数据出现实时矛盾。官方关于页面还确认 Agilink 先在 AgiBot 内部成立、于 2026 年 1 月独立,因此仍需核查产品路线、销售结构、IP 和转移定价风险有多少仍绑定母公司生态。[CO003, CO005, CO011, CO012, CO013]

领导层与创始人表
人员职务背景创始人—市场匹配 / 职能覆盖关键人物依赖
Xiong Kun创始人 / 分拆负责人HKUST 机器人训练;参与创建 Tencent Robotics X;后来负责 AgiBot 灵巧手业务适合灵巧操作和具身机器人末端执行器的强技术匹配高:公开创始人信号远强于已披露的团队深度。
Wang Chuang / Deng Taihua公开引用的法定代表人(有冲突)Baidu Baike 列 Wang Chuang;Sina 2026 年 5 月文章列 Deng Taihua显示治理数据矛盾,尽调应在依赖登记/控制假设前解决中:矛盾影响公开治理图谱的可信度。

这是仅基于本组证据的部分领导层图谱,不是完整管理层或董事会名单。

[CO005, CO011, CO012, CO013]

1.3 融资速度、利益相关方与披露缺口

公开证据支持一条非常快的资本形成故事。关于页面称 Agilink 于 2026 年 1 月独立;Baidu Baike 和 Sina 描述其在数月内完成多轮融资,并在 2026 年 5 月或 6 月达到独角兽估值。 Baidu Baike 还指向由母公司控制的股权结构表:Shanghai Zhiyuan New Innovation Technology 持有约 67% 股份,腾讯和上汽关联主体成为股东。这对估值与控制权分析很重要:Agilink 似乎获得了强劲投资者拉力,但公开来源仍未披露董事会构成、投资者权利、清算优先权或与 AgiBot 的关联交易条款。本章可以确认利益相关方结构,却还不能厘清治理经济关系。[CO008, CO009, CO010, CO013, CO016]

利益相关方或投资者图谱
利益相关方角色控制或经济重要性尽调要求
AgiBot / Shanghai Zhiyuan New Innovation Technology母公司生态及所称主要股东Baidu Baike 称其持有约 67% 股份,官方历史称 Agilink 在 AgiBot 内部成立核实股权结构表、关联方收入结构、IP 所有权和转移定价机制。
Xiong Kun创始人 / 技术发起人分拆故事的核心人才与市场信誉锚点确认留任、授权范围及创始人以下的继任深度。
Shanghai Qishan Investment(腾讯关联)投资者Baidu Baike 称腾讯关联资本在 2026 年融资期间成为股东澄清董事会席位、否决权、按比例认购权及战略商业联系。
SAIC Chuangyuan Venture Capital投资者Baidu Baike 称上汽关联资本成为股东,显示工业信誉检验该关系是否带来分销、制造或客户接触优势。
研究人员、开发者和机器人公司所称客户 / 伙伴基础主页称产品获得全球信任,商店围绕应用定制集成要求命名客户、母公司与第三方结构、复购数据及按终端市场的集中度。

这是锚定官方页面和融资摘要的部分公开利益相关方图谱;公开来源未披露完整权利、董事会或所有权瀑布细节。

[CO003, CO008, CO009, CO013, CO016]

1.4 里程碑与外部风险环境

对于一家如此年轻的独立公司,时间线异常密集:官方及第三方来源支持其从 2023 年内部孵化,经过 2025 年发布里程碑、2026 年 1 月独立、4 月推出新的 Omni 产品、5 月和 6 月出现独角兽估值信号,并在 2026 年 10 月通过一次灵巧操作演示保持曝光。不过,行业背景并非一面倒利好。Humanoids Daily 报道,北京警告中国人形机器人和具身机器人行业警惕“高度重复产品”和泡沫。早期动能由此进入一个奖励差异化执行、却可能惩罚同质化硬件和过度资本化供应链的市场。[CO003, CO008, CO009, CO014, CO015]

FO001: 公司里程碑时间线

Agilink 的公开里程碑显示,公司约三年内快速从内部项目成长为独角兽估值的拆分公司。

2023-08 [CO003]

2025-08 [CO003]

2025-12 [CO003]

2026-01 [CO003, CO010]

2026-02 [CO009]

2026-04 [CO003]

2026-05 [CO008]

2026-06 [CO009, CO010]

2026-10 [CO014]

[CO003, CO008, CO009, CO010, CO014]

1.5 展示项

Chapter 02

02市场分析

2.1 市场边界与买方图谱

本组证据指向的市场比泛化的人形机器人头条更窄。Agilink 官方定位围绕灵巧手和夹爪,AGIBOT 商店则按物体、动作和宿主机器人来选择产品,而不是按整套机器人选择。因此纳入的支出集中在末端执行器、集成和操作数据层,不包括证据无法单独拆出的更广泛人形机器人移动、整机平台和通用基础模型预算。买方和付款方初步集中在三组:采集操作数据的科研团队、把机械手装入人形和通用机器人平台的机器人 OEM 或集成商,以及在可靠性优先于拟人灵巧性的流程中使用夹爪的工业自动化团队。[CM001, CM002, CM003, CM004, CM005, CM009, CM012]

市场定义表
细分 / 类别纳入支出排除支出买方 / 付款方相关性
灵巧机器人手多指末端执行器、触觉感知、操作控制接口整机、移动栈、通用 AI 平台预算人形机器人 OEM、机器人集成商、研发实验室OmniHand 核心类别
工业机器人夹爪任务专用夹持硬件及相关集成工作末端执行器决策之外的工厂自动化一般资本开支工厂自动化团队、系统集成商将 Agilink 延伸到不只人形机器人的需求
集成与数据采集工作流安装、电源、通信、控制软件、远程操作或数据采集整个工作单元改造及无关软件订阅机器人工程团队和项目负责人采用取决于与宿主机器人和工作流的适配
排除的相邻领域:全栈人形机器人除机械手或夹爪部分外,核心视角中无纳入项移动、躯干、手臂、整机 BOM、通用自主栈机器人平台资本项目本组证据不支持可信的整机 TAM

纳入和排除的支出仅限于本组官方及第三方来源实际描述的范围。

[CM001, CM002, CM003, CM009, CM012, CM014]
客户细分 / 买方地图
细分市场买方使用者付款方工作流预算负责人采用触发因素
操作研究与数据采集高校实验室、AI 实验室、机器人研发团队研究人员与远程操控工程师研发预算采集操作数据并测试高自由度末端执行器研究负责人或实验室经理需要灵巧数据集、触觉反馈或多指操作实验
人形机器人与通用机器人集成机器人 OEM 厂商与平台集成商控制、硬件与系统工程师产品或平台项目预算为主机机器人选择并集成机械手工程或产品项目负责人真实机器人项目需要类人交互或复杂操作
工业搬运与应用自动化工厂、解决方案集成商和科学设备团队自动化与运营工程师资本开支或应用研发预算部署夹爪来搬运物体并执行重复任务工厂自动化经理或项目发起人在负载、节拍和接口要求上明确匹配

买方、使用者、付款方和预算负责人字段,是根据 Agilink 声明的应用场景以及 AGIBOT 商店的工作流和 集成表述推断得出。

[CM003, CM004, CM005, CM015]

2.2 规模化视角与需求驱动因素

证据支持多个规模化视角,但不足以形成单一干净的 TAM。IFR 2026 年发布资料称,中国约有 2m 台在运行的工业机器人,占全球年度工业机器人安装量的 54%,这是本组证据中最广泛且可信的需求背景。同一来源显示,中国供应商在国内安装量中的份额从 2020 年的 30% 升至 2024 年的 57%;这很重要,因为 Agilink 正把中国设计的末端执行器卖入政策支持的国产化周期。Humanoids Daily 还提供第二个视角:2026 年第一季度中国有超过 50 笔具身智能融资、资本规模约 RMB 20b;Humanoids Daily 和 Baidu 的 Agilink 发货数字则提供第三个当前吞吐量视角。合在一起,这些输入足以支撑需求叙事,但仍是装机基数、资本和单位发货量代理指标,而非有收入支撑的市场栈。[CM006, CM007, CM008, CM009, CM010, CM012, CM014]

TAM / SAM / SOM 或规模化视角表
发布方年份地理范围数值CAGR方法置信度局限
International Federation of Robotics(国际机器人联合会)2026中国 / 全球背景中国约有 200 万台在运行的工业机器人;全球年度安装量的 54% 部署在中国2026 年 IFR 发布资料引用的 World Robotics 2025 安装与存量统计中广泛自动化背景,不是灵巧手收入 TAM
International Federation of Robotics(国际机器人联合会)2026中国2024 年,本土供应商在国内工业机器人安装量中的占比达到 57%,高于 2020 年的 30%IFR 2026 年发布材料中的本土供应商份额趋势中供应商份额无法单独反映 Agilink 所处品类或价格档位
Humanoids Daily2026中国2026 年第一季度发生 50 多起融资事件,具身智能领域投入约 RMB 20 billion同比接近 60%2026 年 5 月一篇市场文章报道的行业融资总结中资本形成会拉动需求,但不等于终端市场收入
Humanoids Daily + Baidu Baike2026Agilink 公司视角已交付或发运 8,000 多只灵巧手和 10,000 个夹爪两个独立第三方来源重复提及公司的出货数据中部署吞吐量不等于市场份额或经常性收入
AGILINK 官方关于页面2026中国2026 年第一季度中国市场份额第一官方网站上的公司自述低指定来源池中没有出现独立基准、分母或方法论

这些是受证据约束的规模测算视角;指定来源池没有公开 ASP、收入或分部预算数据,无法精确搭建 TAM/SAM/SOM 桥接。

[CM006, CM007, CM008, CM010, CM011, CM014]
FM001: 从行业背景到当前部署的需求阶梯

指定证据支持一套从中国机器人行业大背景逐层下探到 Agilink 工作流就绪部署的需求视角,而不是数学上可相加的 TAM 到 SOM 堆叠。

各层宽度仅表示层级,不代表数值比例。

不同单位呈现的是受证据约束的市场视角,而不是单一的加总或换算市场模型。

中国工业机器人装机基数 [CM006]

中国占全球工业机器人年装机量的份额 [CM006]

本土供应商占中国 2024 年工业机器人装机量的份额 [CM007]

中国 2026 年第一季度具身智能融资 [CM008]

Agilink 灵巧手累计交付量 [CM010]

Agilink 夹爪累计出货量 [CM010]

[CM006, CM007, CM008, CM010, CM015]

2.3 采用约束与尚未解决的市场证明

最强的反证不是需求消失,而是广泛的人形机器人商业化仍比独角兽叙事更慢、更窄。IFR 明确表示,普遍的人形机器人工厂或家庭采用在近期和中期不会发生,并预计未来五至十年 AI 将与传统工业机器人结合采用。这把 Agilink 近期商业机会推向具体、可集成的工作流,而不是泛泛声称整个人形机器人市场已经打开。本组证据还留下两个与估值相关的缺口:没有独立分母验证 Agilink 所称 2026 年第一季度中国市场份额领先,也没有公开收入、ASP 或外部客户结构披露来把发货叙事转换为可辩护的 TAM、SAM 或 SOM 区间。[CM003, CM011, CM013, CM014, CM015]

2.4 展示项

Chapter 03

03竞争者

3.1 竞争集合不只是机械手对机械手

留存证据显示,Agilink 的竞争不止一个产品轴。其官方渠道把市场分为用于多指操作的灵巧手和用于简单物体搬运的夹爪,因此实际替代品往往不是另一只五指手,而是复杂度更低的夹爪或其他任务专用工具。官方指引还称,买方适配取决于安装、负载、电源、通信和软件接口;所以竞争部分是更大机器人项目中的部件选择流程,而不只是末端执行器之间的目录比较。[CP001, CP002, CP012]

功能 / 能力矩阵
购买标准AgilinkROBOTERAShadow Robot现有方案 / 更简单的夹爪
灵巧性上限旗舰产品线包含直驱 OmniHand Ultra-M,具备 20 个主动自由度和触觉感知突出直驱灵巧手架构;保留来源未披露确切自由度24 自由度机械手;研究标准定位围绕抓取任务设计,而非高自由度操作
商业部署证明出货和市场份额声称很强,但主要来自公司或关联方据称已进入 >10 个物流中心,并在二〇二六年第二季度实现千台级交付(Q2 2026K)已显示商业化状态,但保留来源更强调研究定位,而非机队规模在不需要完整灵巧手时的常见替代方案
集成 / 开发者栈已披露安装、供电、协议审查,以及 SDK、API、仿真和 ROS 2 资产已披露自研硬件栈和真实部署反馈;这里未保留 SDK 细节保留的目录条目未披露 SDK 或集成细节机械与控制更简单,可能降低集成负担
基准透明度以公司控制的市场份额和出货声称为主以新闻稿式商业化声称为主以目录式摘要为主替代方案清晰可见,但没有公开的同场直接性能数据

竞争对手单元格若缺乏支持,就保留为未知,或仅按保留来源所述的层次描述;本矩阵比较的是已披露信号, 不是经过验证的同口径性能。

[CP001, CP002, CP004, CP007, CP009, CP010, CP012, CP014, CP015]
FP001: 竞争定位图

对留存备选方案的序数映射显示:当买方看重灵巧性及与 AGIBOT 相关的部署路径时,Agilink 最具优势;ROBOTERA 在已披露部署证据上领先,Shadow Robot 则仍是最明确的研究领域老牌厂商。

X: 部署证据 / 市场进入路径强度(序数)
lowLabel: lab / narrow proof
highLabel: multi-site / scaled route
Y: 灵巧性广度 / 可编程性(序数)
lowLabel: simple gripping
highLabel: high-DOF tactile manipulation
条目部署证据 / 市场进入路径强度(序数) (1–5)灵巧性广度 / 可编程性(序数) (1–5)说明
1. Agilink OmniHand 4 5 2026 年快速商业化口径,加上直驱旗舰规格
2. AGIBOT 集成路径 5 4 与母公司相关的机器人已进入 Longcheer 和 SAIC 工厂
3. ROBOTERA 5 4 >10 个物流中心、千台级交付,以及直驱灵巧手叙事
4. Shadow Robot 2 5 24-DOF 研究标准老牌厂商,留存样本中的规模细节较少
5. Agilink OmniPicker / 夹爪路径 4 2 同一渠道覆盖更简单的物体搬运用例
6. 简单夹爪 / 现有工具 5 1 适用于简单抓取工作的广泛替代方案

坐标轴是序数尺度,不是实测基准。分数反映保留下来的、由来源支持的部署路径与灵巧性信号,不代表推断出的转化率或市场份额。

[CP001, CP005, CP006, CP010, CP011, CP012, CP014, CP016]

3.2 直接同业、相邻供应商与老牌厂商分别争夺不同买方重点

在 2026 年留存材料中,Agilink 最明确的相邻压力来自 ROBOTERA;它把灵巧手架构与更广泛的机器人平台结合,并在物流领域展现商业化。Shadow Robot 构成另一种老牌威胁:本组证据较少显示其近期规模化部署,但它拥有长期科研信誉和一只 24 自由度机械手,仍是高端基准。Agilink 位于两者之间:比 Shadow Robot 更新、更快,但仍需证明早期发货和市场份额说法能否转化为相对于其他具身 AI 厂商的持久、独立验证份额。[CP003, CP004, CP005, CP006, CP007, CP012, CP013, CP014]

竞争对手画像表
竞争对手 / 选项类别规模 / 融资信号目标细分市场差异化信号保留证据中的局限
Agilink直接灵巧手厂商 / AGIBOT 分拆公司2026 年推出,五个月内完成四轮融资,估值声称超过 $1B,出货声称达到 8,000–15,000+具身智能平台、工业自动化、研究直驱旗舰产品加 AGIBOT 生态渠道该来源池中的大多数规模证据由公司或关联方控制
ROBOTERA邻近的人形机器人 / 物流机器人厂商,自研机械手在此前 RMB 1B 战略轮后又完成 >$200M 融资;据称二〇二六年第二季度实现千台级交付(Q2 2026K)以物流为主,并拓展至汽车、电子和服务直驱机械手加更完整的自研硬件栈和部署证明保留来源偏宣传性,并非中立基准
Shadow Robot研究领域标准灵巧手厂商1987 年成立;商业化运营研究用户24 自由度机械手及研究标准定位保留来源几乎没有近期定价或部署细节
简单夹爪 / 现有工具替代品类别不需要完整灵巧性的物体搬运任务更简单的抓取就能解决范围更窄的工作不是精细操作的同类替代品

快照由公司与关联方页面、一则联合发布的竞争对手公告和一家独立竞争对手目录构成;规模标签保留发布方 的原话以及不一致的出货分母,不作统一口径处理。

[CP001, CP003, CP005, CP006, CP007, CP012, CP013, CP014]

3.3 切换成本存在,但公开定价与赢单率可见度低

最明显的切换成本来自集成和开发者工作流,而不是标题价。Agilink 留存来源强调机械适配、通信协议和软件访问;Unite.ai 对已发布规格的整理提到 SDK、仿真资产和 ROS 2 参考节点,机械手接入后可能形成明显工程黏性。即便如此,留存来源没有披露 Agilink、ROBOTERA 或 Shadow Robot 的公开价格,买方和投资者无法从公开证据清楚比较功能与价格。这种不透明也让人难以判断 Agilink 赢在能力、折扣、母公司网络拉动,还是捆绑支持。[CP002, CP009, CP010, CP015]

定价 / 产品包装对比
选项公开定价证据保留来源中的产品模式未知项 / 注意事项含义
Agilink OmniHand保留来源未公开目录价完成任务和集成审查后,由销售主导适配讨论ASP、折扣、服务条款和最小订单量均未披露能力可能有差异化,但公开 ROI 测算较弱
Agilink OmniPicker / 夹爪路径保留来源未公开目录价同一商店提供更简单的物体搬运路径相对于 OmniHand 的价差未披露同一渠道内保留了更低复杂度的替代方案
ROBOTERA保留来源未公开价格有商业部署叙事,但没有报价的产品条款保留证据没有合同模式、维护或支持组合包信息没有公开价格锚点,但可能在整体解决方案价值上竞争
Shadow Robot保留来源未公开价格商业化研究标准定位保留证据没有部署组合包或支持成本信息既有厂商的信誉清晰可见,但公开 TCO 仍不透明

保留来源集对实际成交价格的覆盖较弱;因此,本表比较的是公开定价透明度和产品包装披露,而非标准化合同经济性。

[CP001, CP002, CP011, CP012, CP014, CP015, CP016]

3.4 护城河能否持久仍取决于证明其独立于母公司生态

留存证据表明,Agilink 当前优势部分来自渠道和生态速度。官方及关联来源称公司在 AGIBOT 内部成立;Baidu 记录 AGIBOT 的 Shanghai Zhiyuan 关联公司持有约三分之二股份;C Capital 将机械手业务与已在客户工厂运行的 AGIBOT 机器人联系起来。这能加快部署,但也意味着大量可见证明位于公司控制或财务关联网络内。ROBOTERA 已在销售自有直驱机械手和内部栈,并实现多地点部署;Shadow Robot 仍是可识别的科研基准。因此 Agilink 仍需更多独立的赢单/失单、可靠性和定价证据,证明其优势是持久能力,而不是快速推进、由母公司赋能的第一波。[CP003, CP008, CP011, CP013, CP014, CP016]

Chapter 04

04财务

4.1 收入模式与变现

Agilink 的公开商业形态仍像工业硬件供应商,而不是已披露的经常性软件业务。AGIBOT 商店只展示 OmniHand 灵巧手和 OmniPicker 夹爪;公司网站描述的是灵巧操作系统和末端执行器方案,而非订阅或按使用量计费的软件。销售路径也偏顾问式:买方先按应用选择,再与销售确认安装、电源、通信、软件、预算范围和时间表。收入可能从硬件订单加潜在捆绑集成支持开始,但公开材料没有披露标价、实际 ASP、合同期限、折扣或已确认服务收入。[CI002, CI003, CI009]

收入流表
收入流机制单位当前数值 / 状态质量尽调问题
OmniHand 灵巧手面向研究、人形机器人和工业操作项目销售硬件已发运数量 / 订单根据公司发布的报道,截至 2026 年 9 月已发运超过 15,000 只灵巧手公司声称的商业进展;没有公开收入转化数据拆分外部与内部占比、按 SKU 划分的 ASP,以及各机械手系列的毛利率
OmniPicker 夹爪面向物体搬运和抓取任务销售硬件已发运数量 / 订单根据百度摘要,截至 2026 年 6 月累计报告夹爪出货超过 10,000 个第三方对公司声称的摘要;没有本次运营期的收入披露提供当前出货量、ASP、复购率和客户集中度
集成 / 配置支持围绕安装、供电、控制和工作流匹配,开展售前应用范围界定与兼容性工作工程工时 / NRE / 打包服务公开销售流程可见,但变现方式未披露根据官方工作流推断;没有费用表或附加率量化服务收入占比、实施负担和售后支持经济性

质量字段区分出货证据与收入披露缺口;“未知”表示指定公开来源池没有发布该数值。

[CI002, CI003, CI007, CI008, CI009, CI014]
定价 / 变现表
产品价格 / 单位 / 合同目录价与实际成交价折扣 / 未知项来源
AGIBOT OmniHand 2025(带触觉)未公开列价商店展示型号选择,但没有目录价或合同期限实际 ASP、折扣、最小订单量和付款条款未知AGIBOT Store
OmniPicker 3未公开列价商店按物体搬运使用场景描述产品匹配,而非公布价格经销商条款、服务附加项和保修经济性未知AGIBOT Store
OmniHand 3 系列 / 旗舰变体未公开列价发布信息已公开,但没有公开的变现细节开发访问条款、渠道经济性和合同期限未知AGILINK 关于页面 + AGIBOT Store

官方来源确认了产品和发布节奏,但没有确认实际成交价格或收入确认;空白的经济性信息是尽调问题,不是零值。

[CI006, CI009, CI014]
FI001: 收入模式桥接图

公开证据显示,这是一条以硬件为主、由销售协助推进的路径,从应用范围界定走向终端执行器交付。

节点 / 连接节点 / 起点终点说明
节点 1 应用需求简报 [lead-intake]买方提供任务要求、地点、预算范围和项目时间表。
节点 2 产品选型 [product-selection]销售团队引导买方选择 OmniHand 灵巧手或 OmniPicker 夹爪。
节点 3 集成检查 [technical-fit]确认机械安装、负载能力、电源、通信和控制软件的适配性。
节点 4 硬件订单 [hardware-order]目前看,收入似乎始于终端执行器硬件订单,而非已披露的软件订阅。
节点 5 发货 / 部署 [deployment]发货里程碑显示出规模,但公开来源没有区分内部需求与第三方需求。
节点 6 支持 / 更换 [support-loop]持续支持和更换需求具有合理性,但未披露公开的经常性收入指标。
[CI002, CI003, CI007, CI009, CI014]

4.2 牵引力与单位经济代理指标

公开牵引力信号强于公开经济数据。公司发布的 2026 年 9 月报道称 Agilink 已发货超过 15,000 只灵巧手;Baidu Baike 2026 年 6 月摘要称累计 OmniHand 交付超过 8,000 台,夹爪发货超过 10,000 台。官方关于页面还显示运营快速扩张:Agilink 称当前团队有 60+ 人,计划在 2026 年第四季度扩大至 100+ 人;OmniHand 和 OmniHand Pro 于 2025 年 8 月推出,2025 年 12 月交付第 1,000 只机械手,2026 年推出多个 Omni 3 产品。这些信号支持真实生产活动,但由于公司没有拆分实际定价、外部客户结构或服务成本,发货量仍无法转化为收入、毛利润或客户质量。[CI004, CI005, CI006, CI007, CI008, CI014]

单位经济性表
指标数值 / 空值置信度重要性尽调问题
按 SKU / 渠道划分的实际 ASP低没有实际定价,就无法把出货量转化为收入质量按 OmniHand 和 OmniPicker SKU、直销与合作伙伴渠道,提供已入账 ASP
按产品系列划分的毛利率低硬件投资判断取决于 BOM、良率、保修和服务负担分享按产品系列拆分的毛利率桥接,包括材料成本和保修假设
销售周期 / 方案工程负荷以兼容性范围界定为核心的顾问式销售;时长未披露中决定 CAC、回本周期以及当前 GTM 动作能否规模化报告漏斗转化、试点到下单的时间,以及每单技术支持工时
装机基数构成低母公司内部需求与第三方需求的差异,会改变集中度和收入质量拆分 Agibot 内部项目与外部客户的出货量和收入
现金余额 / 现金跑道低融资势头不能证明当前流动性或对融资的依赖程度提供现金、烧钱、债务额度、资本开支承诺及月度现金跑道桥接

空值表示指定来源池没有披露公开指标;置信度反映该行依据的是直接披露还是流程推断。

[CI003, CI004, CI009, CI010, CI012, CI014]

4.3 资本充足性与融资依赖

Agilink 显然具备融资能力,但当前资本充足性无法公开评估。Crunchbase News 报道,公司在 2026 年 1 月分拆后完成四轮融资并达到 $1B 估值,新资本将用于模型开发、数据和硬件。Agilink 关于页面还称,2026 年 6 月一轮融资将投后估值推至 $1B 以上。这一融资动能很重要,因为公司正在增员并发货更复杂的硬件,通常会提前拉动营运资金、库存承诺、保修准备金和支持成本。Baidu Baike 的持股摘要还称 Agibot 母公司仍持有约 67% 股份,因此即使 Agilink 向外部机器人市场拓展,集团内依赖仍是现实考量。[CI004, CI010, CI011, CI012]

4.4 财务结论与尽调阻塞点

财务结论因此是混合的:商业化似乎真实存在,但收入质量和资本充足性尚不透明,无法形成干净的投资判断。发货和融资里程碑显示动能,但所审阅的公开来源均未披露收入、毛利率、现金消耗、现金跑道、债务义务、客户集中度,或 Agibot 内部需求与第三方买方的比例。Humanoids Daily 对北京警告行业“泡沫”和重复人形产品的报道构成可信的不利背景,因为它恰好在 Agilink 扩大生产时提高了差异化和资本纪律门槛。在公司提供 SKU 级 ASP、产品系列利润率、流动性和母公司交易细节之前,本章支持“继续研究”,而非对利润路径或融资韧性形成确定判断。[CI007, CI010, CI012, CI013, CI014]

4.5 展示项

Chapter 05

05产品与技术

5.1 产品范围清晰,但供给仍由目录驱动

Agilink 官方渠道始终把公司定义为灵巧操作企业,而不是广泛的机器人平台:关于页面将公司定位为机械手和夹爪制造商,AGIBOT 商店则要求买方围绕目标应用中的物体、动作和宿主机器人选择末端执行器。这种定位有价值,因为它把产品锚定在客户工作流,而不是笼统的具身 AI 叙事上。不过公开目录仍很窄。主灵巧手集合页当前可见产品是两个 OmniHand 变体加 OmniPicker 3,说明它更像聚焦的末端执行器业务,而不是拥有大量生产 SKU 的完整模块化栈。Baidu Baike 的场景摘要则把同一故事从目录扩展到 go-to-market:数据采集、机器人集成以及工业或科研部署。[CE001, CE002, CE003, CE008]

产品模块 / 资产矩阵
模块 / 资产 / 产品线主要用户状态 / 成熟度差异化信号尽调缺口
AGIBOT OmniHand 2025(带触觉)研究 / 集成团队公开目录条目当前商店产品线中强调触觉的灵巧手需要按 SKU 明确性能、负载和支持的控制栈
AGIBOT OmniHand 2025研究 / 集成团队公开目录条目当前商店产品线中的基础灵巧手选项需要明确其与 OmniHand 2025(w.Tactile) 及 OmniHand 3 系列的差异
OmniPicker 3工业 / 物体搬运买方公开目录条目;时间线显示 2026 年 4 月发布将简单抓取与多指机械手使用场景分开需要客户部署、负载上限和循环寿命数据
OmniHand 3 系列(Lite / Ultra-T / Ultra-M)前沿机器人 / 高端操作项目官方时间线显示于 2026 年 4 月和 6 月陆续发布围绕更新的 Omni 3 架构快速迭代发布需要一份对齐后的 SKU 表,涵盖自由度、驱动、感知和目标工作负载

快照整合当前目录与官方时间线页面;各行总结的是公开定位,并非经过验证的量产性能。

[CE001, CE002, CE004, CE013]
工作流 / 使用场景表
用户任务当前工作流公司方案可量化收益局限
类人交互 / 操作研究研究人员为多指运动和演示选择灵巧手OmniHand 各变体定位于交互与操作研究公开目录展示专门的灵巧手产品线,而不是泛化的夹爪宣传指定来源池中没有与同类研究机械手对比的公开基准
物体搬运买方根据物体尺寸、质量、表面和运动需求匹配末端执行器OmniPicker 3 定位于抓取任务清晰区分夹爪与机械手使用场景,可以减少不必要的复杂度采购没有公开吞吐量、故障率或维护数据
柔性材料精密任务操作员需要对可变形材料进行力、张力和接触控制OmniHand 3 Ultra 演示涵盖丝线分离、穿线、套圈和缝纫独立报道显示其任务范围超出静态抓取与放置演示证据尚不能证明产线节拍下的可重复性或故障恢复能力
机器人集成 / 数据采集项目具身智能团队需要采集数据,并将机械手集成到更大的机器人栈中Baidu Baike 描述了算法 / 数据采集和机器人集成场景据此,Agilink 不仅能提供末端硬件,也可能覆盖上游数据生成需求指定来源池没有记录公开 SDK、API 或工具接口

收益仅限于保留页面明确支持的内容;缺乏公开运营指标的项目仍列为局限,不把它们推断成正面因素。

[CE003, CE005, CE008, CE012]

5.2 有公开证据的架构集中在末端执行器机械、触觉感知和集成依赖

本组证据中最强的产品证明是柔性材料演示。两篇独立报道称 OmniHand 3 Ultra 完成丝线分离、穿针、绷圈和缝纫,并都描述其为 21 个主动自由度、直驱、带触觉感知的机械手。The Rundown 还补充了一条训练路径:先远程操作采集数据,再用强化学习优化,比泛泛的“AI 驱动”定位更具体。与此同时,部署负担仍由集成商承担:商店要求买方在采购前验证安装、电源、通信和控制软件兼容性。这很重要,因为 Humanoids Daily 将 OmniHand 3 Ultra-T 放在 AGIBOT 更大的“One Body, Three Intelligences”栈内,公开记录因此显示出显著生态依赖,而不是一套单独记录的 Agilink 软件/控制平台。[CE005, CE006, CE007, CE009, CE010, CE011, CE014]

技术 / 运营架构表
层 / 流程 / 组件作用依赖风险
末端执行器机械结构提供与任务匹配的灵巧手或夹爪形态官方产品目录和确切 SKU 选择公开证据没有把 OmniHand 3 系列规格整合成一份权威 SKU 矩阵
触觉感知 + 驱动支持薄材料处理与操作反馈取决于具体型号的硬件设计(直驱或腱驱)变体级差异可能实质影响精度、耐久性和成本
训练闭环从远程操控演示推进到学习式执行演示报道描述了远程操控数据采集和强化学习优化没有公开成功率、样本效率或故障恢复指标
主机机器人集成将机械手 / 夹爪连接到买方的机器人平台机械安装、负载或供电范围、通信和控制软件集成负担似乎很大程度落在客户或系统集成商身上

这是有证据支撑的运营地图,不是完整的内部架构披露;尚未解决的软件 / 控制细节仍属于证据缺口。

[CE006, CE007, CE009, CE011, CE014]
FE001: 产品架构图

公开证据呈现出分层的产品逻辑:从任务适配和硬件 SKU,一直延伸到训练闭环与对 AGIBOT 生态的依赖。

客户工作流适配
根据物体、动作和主机机器人选择终端执行器。
条目
  • 交互与操作
  • 物体处理
  • 操作研究
公开产品系列
灵巧手产品集合页目前可见的产品阵容。
模块
  • AGIBOT OmniHand 2025(带触觉)
  • AGIBOT OmniHand 2025
  • OmniPicker 3
终端执行器硬件
公开信息提供了型号级硬件信号,但尚未汇总成一份统一规格表。
模块
  • OmniHand 3 Ultra:21 个主动自由度
  • 直驱 + 触觉感知
  • OmniHand 3 Ultra-T:22+3 个腱驱自由度
训练闭环
公开演示对学习方法的描述,比对生产 KPI 的描述更具体。
条目
  • 远程操作数据采集
  • 强化学习优化
  • 独立扭转与抓取序列
主机机器人集成
兼容性检查是明确的售前要求。
输出
  • 机械安装
  • 负载 / 电源
  • 通信协议
  • 控制软件
母生态依赖
APC 覆盖情况将 OmniHand 的发布与 AGIBOT 更广泛的软件及平台栈连接起来。
模块
  • GO-2 (VILLA)
  • Genie Sim 3.0
  • WITA Omni
  • AIMA / Link-U OS
[CE009, CE010, CE011, CE014]

5.3 发布节奏和发货说法支持动能,但尚未达到可据此投资的成熟度

Agilink 官方时间线显示 2025—2026 年节奏异常快:2025 年 8 月推出 OmniHand 和 OmniHand Pro,2026 年 4 月推出 Omni 3 系列,2026 年 6 月在 ICRA 发布旗舰 OmniHand 3 Ultra-M。同页的组织建设信号也支持这一点,包括 60+ 人团队向 2026 年第四季度 100+ 人扩张。公开销量说法同样快速上升:6 月 OmniHand 交付超过 8,000 台,9 月 30 日灵巧手超过 15,000 只,主页快照显示产品总发货超过 20,000 件。这些数字对商业化方向有利,但不能直接比较,因为披露范围从 OmniHand 交付变为灵巧手,再变为全部产品,且没有区分 AGIBOT 内部项目与第三方客户单位。[CE004, CE012, CE013]

5.4 技术尽调应聚焦 SKU 对账、可靠性与信任证明

当前证据集最大的产品风险不是缺少活动,而是丰富演示和发布叙事与薄弱运营证明之间的差距。公开材料没有把 OmniHand 3 系列整理成清晰的 SKU 架构图:独立演示报道将 OmniHand 3 Ultra 描述为 21 自由度直驱机械手,而 APC 报道将 OmniHand 3 Ultra-T 描述为 22+3 自由度腱驱系统。这可能代表不同产品而非矛盾,但买方仍需就驱动、传感和支持的工作负载进行精确的 SKU 级尽调。同样重要的是,本组证据没有建立生产可靠性指标、面向客户的支持 SLA 或外部信任/合规文件。对投资判断而言,这些缺口比再看一次英雄演示更重要,因为它们决定 Agilink 卖的是可复用子系统,还是仍高度依赖工程师的集成项目。[CE009, CE011, CE014]

5.5 展示项

Chapter 06

06客户

6.1 客户覆盖广,但付款方质量更多是推断而非证明

本组证据最能支持的是客户覆盖面,而不是命名客户深度。AGIBOT 商店围绕“交互与操作”“物体处理”和“操作研究”定位 OmniHand 与 OmniPicker;AGILINK 关于页面则把公司描述为服务科研和工业的灵巧操作供应商。这一组合支持一张覆盖机器人科研团队、人形或通用机器人 OEM 以及需要末端执行器而非整机平台的工业集成商的买方图谱。但它没有按细分披露付款方结构,也没有显示多少销量来自关联 AGIBOT 项目、多少来自外部客户,因此安装基数的商业质量仍有部分是推断。[CU001, CU002, CU003, CU006, CU007]

客户细分表
细分市场买方 / 使用者 / 付款方使用场景规模信号收入 / 战略价值缺口
具身 AI 算法与数据采集团队买方:机器人项目负责人;使用者:操作研究人员;付款方:实验室或 OEM 项目预算数据采集与灵巧操作开发被列为公司三大应用场景之一为早期导入和评估需求埋下种子未披露具名客户或付费量拆分
人形与通用机器人集成商买方:机器人 OEM 或集成商;使用者:集成工程师;付款方:平台厂商将灵巧手集成到人形与通用机器人官方材料将 AGILINK 定位为核心部件供应商可能是早期平台需求的主要来源外部客户与关联方的占比尚未验证
工业物体搬运部署买方:自动化团队或集成商;使用者:运维工程师;付款方:制造商或仓储运营方抓取和物体搬运任务AGIBOT 商城重点展示 OmniPicker 及应用范围从研究型买方拓展出去的路径未列出生产部署客户
AGIBOT 母生态买方:关联平台团队;使用者:人形机器人工程师;付款方:母公司或关联实体AGIBOT 人形机器人平台的核心部件指定样本中最明确的部署场景可能加速出货量和市场可信度若第三方收入低于出货量所暗示的水平,将形成集中度风险

各行仅汇总指定样本中出现的公开细分市场;不能据此证明各细分市场的收入占比、合同规模或留存。

[CU001, CU002, CU003, CU006, CU007]
FU001: 客户旅程图

公开证据暗示,这是一条从任务范围界定走向重复采购量的应用工程旅程;首次部署之后的验证缺口最大。

旅程阶段综合了指定来源池中的产品选型、出货和渠道证据;它不是一条披露了转化率的漏斗。

[CU002, CU005, CU008, CU012, CU013]

6.2 采用证明真实存在,但命名第三方生产部署仍很少

采用轨迹强于纯概念阶段的机器人供应商,但证明面不均衡。AGILINK 称项目在 2025 年 8 月推出首批产品,并于 2025 年 12 月达到第 1,000 只机械手的生产里程碑;Baidu Baike 称截至 2026 年 6 月累计 OmniHand 交付超过 8,000 台、夹爪发货超过 10,000 台。这些里程碑意味着有意义的实物销量,但同一来源集仍未显示有结果证据的命名第三方客户、合同期限、复购或按地域的部署数量。实际上,本组证据更能支持“存在商业牵引力”,而不是“存在持久且多元化的外部客户基础”。[CU004, CU005, CU006, CU012, CU013]

客户增长 / 采用轨迹表
指标数值日期来源置信度含义缺失分母
首次产品发布OmniHand 和 OmniHand Pro 已发布2025-08AGILINK 关于页面中商业化供货早于 2026 年拆分没有发布到下单的转化率或客户数
生产里程碑第 1,000 只灵巧手下线2025-12AGILINK 关于页面中表明公司独立前已经具备真实的制造产能没有内部需求与外部需求的拆分
灵巧手累计交付量>8,000 OmniHand 台2026-06Baidu Baike中对早期部件供应商而言,说明出货量已经不小没有装机基数、活跃使用或复购率
夹爪累计出货量>10,000 个夹爪2026-06Baidu Baike中表明更简单的末端执行器产品线可能把覆盖面扩展到人形机器人之外没有细分市场构成、地域或客户集中度

轨迹表将公司发布的里程碑与一份第三方出货汇总放在一起;没有任何一行披露留存、活跃部署或收入质量。

[CU004, CU005, CU012]
具名客户证据表
客户 / 证据面细分市场部署 / 用例生产还是试点结果局限
AGIBOT 人形机器人平台关联平台客户灵巧手被描述为 AGIBOT 人形平台的核心部件根据公司里程碑口径,接近生产部署指定样本中最具体的具名部署场景关联方证据弱于第三方客户验证
与 CP Robotics 相关的垂直应用(通过 AgiBot)渠道 / 应用开发伙伴生命科学、新零售、新消费以及健康与保健领域的商业应用Agilink 尚在扩张,尚无证据证明已进入生产阶段显示出进入垂直应用的潜在路径证据只落在与母公司的关系上,尚非 Agilink 的具名收入
未具名的研究与工业买方研究实验室与工业集成商操作研究、交互演示和物体搬运尚未确认处于试点还是生产阶段支持较宽的细分市场适配度没有具名第三方客户、合同或结果指标

这份清单是指定样本中的部分公开客户证据,不是完整的客户名册。

[CU003, CU006, CU008, CU013]

6.3 扩张依赖伙伴渠道,持久性与集中度仍待确认

独立证据指向机会,也指向采购摩擦和集中度风险。Yicai 将母公司 AgiBot 与 CP Robotics 的关系联系到生命科学、零售和健康领域的垂直应用开发,这暗示 Agilink 的商业扩张可能仍经由更大的 AgiBot 生态,而不是可重复的独立需求。Humanoids Daily 对 Morgan Stanley 中国人形机器人调查的报道也提醒保持谨慎:兴趣很高,但满意度低,已经评估或启动试点的只占少数。同一尽调挑战也出现在检索过程中:无关的软件和线缆公司以“Agilink”名义出现,使外部客户验证比标题上的发货故事更嘈杂。本组证据未披露 NRR、GRR、流失、续约或头部客户结构,因此持久性和集中度仍是重大缺口。[CU008, CU009, CU010, CU011, CU012, CU013]

6.4 展示项

Chapter 07

07风险

7.1 剩余风险排序

最强的留存证据指向三项风险:母公司/控制权集中、部署可靠性和估值纪律。如果无法解决,它们都可能击穿投资逻辑。官方历史称 Agilink 在 Agibot 内部成立;Baidu 报道 Agibot 关联股东仍控制约三分之二股权,因此公开材料尚未证明治理独立。产品证据也更接近集成指引、伙伴工作流和展示演示,而不是经过审计的现场可靠性。最后,Crunchbase 在分拆数月内采用独角兽叙事,但本次审阅的官方公开材料仍未披露足以支撑该价格的外部客户收入质量。因此,否决标准应聚焦控制权、现场故障数据和外部客户经济性,而不是只看发布节奏。[CR003, CR006, CR009, CR016]

FR001: 风险热力图

根据证据,从可能性、影响和缓释成熟度三个维度呈现主要剩余风险。

热力图分数总结的是保留证据,而非经审计的事件频率;获得客户构成、现场可靠性和治理文件后,应更新这些分数。

[CR003, CR006, CR009, CR012, CR015, CR016]

7.2 治理与监管验证风险

公开治理证据有方向性价值,但并不完整。Agilink 自己的历史称项目在 Agibot 内部启动,直到 2026 年 1 月才成为独立公司;Baidu 报道 Agibot 关联股东在外部投资者加入后仍持有约 67% 股份。这一组合使董事会控制、IP 转让、关联交易和转移定价存在实质剩余风险。此外,本章留存公开资料没有出现明确的许可证、认证、诉讼结果或执法清白证明。这并不证明存在问题,但也不足以排除法律或监管敞口;下一步应由法律顾问和管理层直接提供尽调材料,而不是从营销页面推断。[CR001, CR002, CR003, CR014, CR015]

监管 / 法律风险登记表
风险司法管辖区 / 范围留存来源中的证据可能性严重程度缓释成熟度剩余风险敞口尽调路径
许可 / 认证清晰度中国及出口市场没有任何留存来源披露 OmniHand 部署的公开认证或许可文件包。中高低在直接核实认证、测试报告和出口边界前,风险维持高位。索取产品认证、安全测试报告和出口管制律师备忘录。
拆分治理 / IP 转移公司结构与关联方治理官方历史将该业务与 Agibot 关联起来,Baidu 报道称一名与 Agibot 相关的股东持有多数股权。高高中在董事会权利、IP 转让和转让定价形成文件前,风险维持高位。审阅董事会构成、IP 转让链、关联方协议和转让定价政策。
诉讼 / 执法可见度公司及主要市场留存的公开证据没有识别出具名诉讼或执法结果,因此尚未完成排查。低中低中,因为没有留存案件不等于不存在案件。委托当地律师核查中国公司的工商、诉讼及监管登记。

仅为公开来源登记表;评级反映留存证据质量和披露完整度,不构成法律排查意见;“没有留存来源”是指当前指定样本未找到该文件。

[CR003, CR015]

7.3 产品集成与商业化执行风险

产品栈技术野心很大,但留存来源描述了明显的集成复杂度。AGIBOT 商店称兼容性取决于宿主机器人的安装、负载、电源、通信和软件接口;MANUS 说明 OmniHand 控制可能需要重定向、逆运动学和关节耦合转换,而不是直接的关节对关节映射。这些事实提高了客户实现稳定部署的门槛,也让可靠性证据比会议演示更重要。本组看起来最独立的高曝光文章仍是 Agilink 赞助的 IEEE Spectrum 文章,重点是 ICRA 的气球扭转;IFR 更广泛的市场背景则称现实世界人形能力仍集中在演示或试点项目。剩余问题不是演示是否令人印象深刻,而是 Agilink 能否把它们转化为可重复的开机率、服务和续约结果。[CR004, CR005, CR006, CR007, CR008, CR009]

运营 / 质量 / 安全风险登记表
失效模式留存来源中的证据可能性严重程度缓释成熟度剩余风险敞口未解决缺口
主机机器人集成不匹配AGIBOT 表示,适配取决于安装、负载、电源、通信和软件接口。高高中对系统集成能力不足的买方而言,风险较高。没有按机器人平台披露留存的部署成功率或安装时间基准。
控制栈 / 重定向不稳定MANUS 描述了机械耦合关节,以及重定向和逆运动学层。中高中在演示场景之外测得稳定的现场表现前,风险维持高位。没有留存的 MTBF、故障率或退货率披露。
从演示到生产的转化IEEE 赞助报道聚焦 ICRA 气球演示,而 IFR 表示许多人形机器人能力仍处于试点阶段。中高低可重复的商业部署证据仍然薄弱,风险较高。没有留存的付费生产客户群及其正常运行时间或续约证据。
快速扩张下的质量体系压力官方里程碑显示公司近期完成拆分、宣称产量扩大,并计划招聘超过 100 人。中中低在核实流程成熟度和现场支持覆盖前,风险维持中位。没有留存的质量认证、现场服务或召回流程披露。

这份登记表混合了直接产品文档、伙伴工作流证据和行业背景;它不意味着已知事故,只说明留存的生产可靠性公开证据仍然有限。

[CR006, CR007, CR008, CR009, CR012]

7.4 融资与依赖风险

融资动能是优势,也提高了价格纪律和流程成熟度风险。Crunchbase 称 Agilink 在 2026 年 1 月分拆后完成四轮融资并达到 $1B 估值;官方关于页面同时显示团队向 2026 年第四季度 100+ 人扩张,且发布日程密集。如果需求质量已被证明,这种组合可能利好;但质量体系、伙伴管理和客户支持也必须快速扩张。本次审阅的公开材料仍未披露经审计收入、利润率或外部客户集中度,因此投资者还无法判断故事有多少依赖 Agibot 内部需求、生态伙伴或前瞻性市场定位。审慎的投资判断应把估值舒适区绑定到已验证的外部收入质量和集中度,而不是只看独角兽身份。[CR010, CR011, CR012, CR013, CR016]

伙伴 / 依赖风险登记表
依赖项对手方 / 锚点角色集中度失效情景严重程度缓释措施剩余风险敞口
对母公司及控制权的依赖Agibot / 与 Agibot 相关的股东起源平台、控股股东及可能的关联方锚点高治理、定价或 IP 决策偏向母生态,而非少数投资者或外部客户。高完成拆分并获得外部融资,说明公司已有一定机构化。在控制权和关联方边界形成文件前,风险维持高位。
工作流与集成伙伴MANUS 和系统集成商支持遥操作、重定向、数据采集和部署工作流中客户需要特定伙伴的工具或专业能力,导致上线变慢或支持成本上升。中开放接口和多条集成商路径可能逐步降低锁定。留存的伙伴冗余证据有限,因此风险为中。
外部买方对估值的支撑外部客户基础与内部 / 生态需求为独角兽叙事验证收入并支撑价格未知若外部客户占比低或过于集中,需求质量将弱于融资叙事。高除融资势头和产品里程碑外,留存公开证据中看不到任何缓释措施。在披露经审计的外部客户收入构成和集中度前,风险维持高位。

集中度反映留存来源公开揭示的内容;“未知”表示指定样本没有量化分母,不代表集中度低。

[CR003, CR006, CR010, CR013, CR014, CR016]

7.5 展示项

Chapter 08

08估值

8.1 估值背景与价格纪律

公开价格锚点足以确认 Agilink 已越过独角兽门槛,却不足以证明当前价格有吸引力。Dealroom 显示 2026 年 6 月一笔 $138M 融资,报告估值为 $1.4B;Crunchbase 和公司关联来源也一致把公司放在 2026 年 5 月至 6 月的独角兽群体中。因此,估值问题已不是市场是否愿意为资产支付 $1B 以上,而是今天的外部投资者是否掌握足够收入质量、稀释和可重复需求证据,能在没有折价或更强保护的情况下跟随该价格。[CV002, CV003, CV006, CV007, CV012, CV015]

建议摘要表
观察维度建议置信度风险评级估值判断决策含义
当前判断research-moremediumhighstretched继续推进尽调,但在获得更扎实的经济性和治理证据前,不要按当前价格做投资判断。
升级条件conditional buylowmediumfair-to-attractive需要验证收入质量、外部客户集中度数据,以及能防范平价或结构化后续轮融资的条款。

公开来源提供了估值标记和产品进展代理指标,但没有经过验证的收入、利润率或优先权结构细节;第二行是尽调门槛,不是当前判断。

[CV002, CV010, CV011, CV013, CV014, CV015]

8.2 正论与反论

看多逻辑首先基于 Agilink 不只是概念稿的证据:官方产品页显示已有灵巧手和夹爪在售,公司称分拆后在 2025 和 2026 年推出规模化产品线,独立报道也把当前融资与面向部署的硬件和数据项目联系起来。反论在于,最影响估值的变量仍被隐藏。公开来源未验证收入、毛利率、复购行为或优先权结构;更广泛的人形机器人生态仍面临商业化和安全摩擦,可能比叙事驱动的私募估值修复得更快地压缩倍数。[CV001, CV004, CV005, CV008, CV009, CV010, CV011, CV013]

正方 / 反方论点表
论点当前证据哪些事实会改变判断
正方:商业化是真实的官方和独立来源显示 OmniHand、OmniPicker 产品在售,2026 年 6 月完成独角兽融资,且灵巧手出货量据称超过 8,000 只。经验证的客户名称、复购节奏和按产品线拆分的收入,将显著强化这一判断。
正方:品类顺风有助于退出IFR 和 IDC 都将机器人及人形机器人采用描述为战略增长市场,同时中国政策提供支持,出货量预计快速增长。持续的出口需求或非 Agibot 授权证据,会让这一顺风更能落到公司本身。
反方:估值领先于已披露的经济性公开记录显示定价超过 $1B,但没有经审计的收入、毛利率、留存或复购质量。若在相近估值完成一轮有价格的融资,并由经审计的经营指标支撑,这一差距会缩小。
反方:采用与安全摩擦会压缩倍数IFR 和行业报道都警告,大规模采用并非近期之事,安全标准和触觉传感瓶颈仍在解决。第三方证明产品已可靠地大规模生产部署,将降低这一下行风险。

这张表混合了直接来源事实与投资判断,因此每一行都应视为综合论点,而非孤立的原始数据点。

[CV004, CV005, CV006, CV008, CV009, CV010, CV011, CV013]

8.3 情景、可比公司与退出准备度

情景分析应以里程碑为基础,而不是假装公开记录包含一个干净的收入倍数。Dealroom 和 Crunchbase 暗示 Agilink 的定价更接近新晋机器人独角兽,而非已披露现金流基本面,因此可比集合应被视为方向性情绪,而不是精确基准。上行情景要求证明发货说法能转化为复购需求、健康利润率,并能在 Agibot 生态之外扩大许可。下行情景很直接:如果政策热情和品类稀缺性比单位经济性更能推动估值,即使公司继续发货,下一轮也可能平价、带结构条款或延期。[CV002, CV006, CV008, CV009, CV010, CV011, CV013, CV015]

乐观 / 基准 / 悲观情景表
情景假设估值 / 回报逻辑主要风险概率信号
乐观出货转化为可重复的外部需求,OmniHand 3 实现规模化生产,Agilink 也从封闭生态销售拓展出去。当前独角兽估值将证明偏保守,下一轮可凭经济性支撑的增长获得更高定价。执行延误、出口摩擦或利润率偏低,仍可能封顶上行空间。前提是验证当前出货叙事能转化为收入。
基准产品进展延续,但投资者仍缺少经审计的经济性数据和清晰的股权结构表。当前 $1.0-1.4B 区间更像持仓估值,而不是明显的便宜价格。外部客户增长停滞或治理复杂度上升,可能令回报维持平庸。与今天的证据最匹配。
悲观出货势头真实存在,但变现、客户多元化或安全准备度令人失望。下一轮融资需要结构化安排、估值下调,或耗时长于当前叙事所暗示的时间。若品类倍数在基本面追上前降温,下一轮降价融资风险会上升。若数据室证据无法填补当前缺口,这一情景最可能发生。
止损触发条件管理层无法验证收入质量、非 Agibot 客户构成或优先股堆叠。没有下行保护,就没有理由支付由叙事驱动的独角兽价格。信息不对称主导投资判断。应立即将判断从“继续研究”改为“放弃”。

这些是投资判断情景,不是模型输出;公开证据不足以支撑可辩护的 DCF 或收入倍数推导,因此区间以定性方式表达,而不制造虚假的精确度。

[CV002, CV006, CV008, CV009, CV010, CV011, CV013, CV014, CV015]

8.4 最终建议与尽调要求

证据支持继续接触,但不支持盲目接受价格。Agilink 有足够产品和融资证据留在进行中的尽调名单上,品类背景也足够有吸引力,完全放弃还为时过早。即便如此,本章尚未达到买入门槛,因为当前估值标记比其背后的经济性更有记录。投资判断应因此是“继续研究、估值偏高”:只有当管理层能够展示经审计或达到贷款人标准的收入质量、外部客户结构、股权结构表条款,并证明商业化扩张速度能超过标准、安全和采用瓶颈拖慢速度时,才应重新评估。[CV010, CV011, CV013, CV014]

FV001: 推荐逻辑

推荐仍为继续研究:真实的产品和市场证明,被估值不透明与采用风险抵消。

节点 / 连接节点 / 起点终点说明
节点 1 产品广度 [products]OmniHand 和 OmniPicker 是已上市的产品系列,官方定位覆盖操作、物体处理和研究。
节点 2 商业牵引代理指标 [traction]独立来源和公司周边来源称,2026 年第二季度有生产交付计划,累计手部出货量超过 8,000 只。
节点 3 市场顺风 [market]IFR 和 IDC 都描述了有利的机器人政策与出货增长背景。
节点 4 当前定价 [price]Dealroom 和 Crunchbase 认为,公司在 2026 年中处于独角兽估值水平。
节点 5 经济性与治理缺口 [gaps]公开证据仍缺少经验证的收入质量、优先权条款和客户构成细节。
节点 6 继续研究 / 估值偏高 [recommendation]保持关注,但在接受当前估值前,要求更扎实的经济性证据和下行保护。
连接 1 产品广度 [products]商业牵引代理指标 [traction]支持
连接 2 商业牵引代理指标 [traction]当前定价 [price]有助于论证
连接 3 市场顺风 [market]当前定价 [price]提高兴趣
连接 4 经济性与治理缺口 [gaps]继续研究 / 估值偏高 [recommendation]阻止买入
连接 5 当前定价 [price]继续研究 / 估值偏高 [recommendation]要求保持纪律
连接 6 商业牵引代理指标 [traction]继续研究 / 估值偏高 [recommendation]保持关注

产品广度 [CV004]

商业牵引代理指标 [CV005, CV006]

市场顺风 [CV008, CV009]

当前定价 [CV002, CV007, CV015]

经济性与治理缺口 [CV012, CV013]

继续研究 / 估值偏高 [CV014]

支持 [CV004, CV005]

有助于论证 [CV006, CV007]

提高兴趣 [CV008, CV009]

阻止买入 [CV012, CV013]

要求保持纪律 [CV002, CV015]

保持关注 [CV005, CV006]

[CV002, CV004, CV005, CV006, CV007, CV008, CV009, CV012, CV013, CV014, CV015]

免责声明

本报告是基于公开证据的尽调快照,不构成投资建议。重要的财务、法律、技术和合同事实仍未公开,任何投资决策前都应直接向管理层及主要文件核实。

证据索引

结论
编号陈述可信度来源
CO001 Agilink says it designs and manufactures dexterous manipulation systems, spanning high-DoF hands and adaptive grippers for research, industrial automation, and service robotics. 高 SO002, SO003
CO002 Current official surfaces present the active lineup as the OmniHand 3 family and OmniPicker 3, with selection centered on application fit and integration requirements. 高 SO002, SO003, SO006
CO003 Agilink's official about page says the dexterous-hand program began within AgiBot in August 2023 and that the company became independent in January 2026. 中 SO005
CO004 Agilink's about page says the team currently exceeds 60 people and is scaling toward 100+ by Q4 2026. 中 SO005
CO005 The official about page says more than half of the team has production experience at top robotics and AI companies and more than half comes from leading research institutions. 中 SO005
CO006 Agilink's homepage claims No.1 global dexterous-hand market share in Q1 2026, but no retained independent benchmark in the assigned pool explains the methodology. 低 SO002
CO007 Agilink's homepage claims 20,000+ products shipped and 50,000+ hours of real-world data. 中 SO002
CO008 Sina reported on 2026-05-20 that Agilink, spun out from AgiBot, had completed a new several-hundred-million-RMB financing round and crossed a US$1 billion valuation. 中 SO008
CO009 Baidu Baike says Agilink completed Series A financing in February 2026, Series A+ in May 2026, and a near-RMB1 billion financing announced on 2026-06-24, while cumulative OmniHand deliveries exceeded 8,000 units. 低 SO004
CO010 Official and third-party sources together support the view that Agilink is a 2026 spinout with a very short standalone operating history but deep roots inside AgiBot's dexterous-hand program. 高 SO001, SO004, SO005
CO011 Baidu Baike identifies Wang Chuang as Agilink's legal representative. 低 SO004
CO012 Sina's 2026-05-20 financing article identifies Deng Taihua as Agilink's legal representative, conflicting with the Baidu Baike summary. 低 SO008
CO013 Both Baidu Baike and Sina attribute the founder or core technical leadership signal to Xiong Kun, citing Tencent Robotics X and prior AgiBot-linked dexterous-hand work. 中 SO004, SO008
CO014 New Atlas described an October 2026 embroidery demonstration using Agilink's OmniHand 3 Ultra and characterized Agilink as a spin-off of AgiBot. 中 SO001
CO015 Humanoids Daily reported that Beijing warned China's humanoid robotics sector against repetitive products and bubble dynamics, raising a relevant sector-risk backdrop for Agilink's category. 中 SO007
CO016 The assigned public-source pool does not disclose verified revenue, named-customer count, board composition, investor rights, or parent-related-party commercial terms for Agilink. 中 SO002, SO004, SO005, SO008
CM001 Agilink's official positioning centers on robotic manipulation systems, specifically dexterous hands and grippers, rather than the full robot platform. 中 SM004
CM002 The AGIBOT store frames purchase decisions around the object, movement, and host robot, implying the addressable spend sits at the end-effector and integration layer. 中 SM001
CM003 Compatibility depends on the mounting interface, power supply, communication protocol, and control software, so integration work is part of the adoption path. 中 SM001
CM004 Baidu Baike says AGILINK's business targets algorithm and data collection, humanoid or general robot integration, and industrial or scientific deployment. 中 SM003
CM005 Those scenario descriptions imply three early buyer-payer clusters: robotics R&D labs, robot OEM or integrator programs, and industrial automation teams. 低 SM001, SM003
CM006 IFR says China has around 2 million operational industrial robots and accounted for 54% of annual global industrial-robot installations. 中 SM006
CM007 IFR says local suppliers' share of domestic industrial-robot installations rose from 30% in 2020 to 57% in 2024. 中 SM006
CM008 Humanoids Daily reported more than 50 financing events and roughly RMB 20 billion flowing into China's embodied-intelligence sector in Q1 2026, nearly 60% higher year over year. 中 SM002
CM009 Humanoids Daily said dexterous hands historically account for roughly 20% of a humanoid robot's bill of materials. 中 SM002
CM010 Humanoids Daily and Baidu Baike both reported that Agilink had surpassed 8,000 dexterous hands and 10,000 grippers in cumulative deliveries or shipments by mid-2026. 中 SM002, SM003
CM011 Agilink's official about page claims the company held number-one market share in China in Q1 2026. 低 SM004
CM012 Crunchbase described Agilink as a dexterous-hand spinout using fresh funding for model development, data, and hardware that can be licensed to the broader robotics market. 中 SM005
CM013 IFR says mass adoption of universal humanoid factory helpers or household robots will not happen in the near- or medium-term future. 中 SM006
CM014 The assigned public evidence does not isolate Agilink's TAM, SAM, or SOM with revenue-grade precision because it is built from installed-base, capital-flow, and shipment proxies rather than disclosed ASP, revenue, or budget data. 中 SM002, SM003, SM005, SM006
CM015 The nearer commercial opportunity is more likely to come from integration-ready research, OEM, and industrial workflows than from immediate broad-based humanoid labor replacement. 中 SM001, SM003, SM006
CP001 Agilink's official storefront positions OmniHand for multi-finger manipulation and OmniPicker for simpler gripping tasks, so a practical substitute is often a simpler gripper rather than another dexterous hand. 中 SP001
CP002 Agilink's official sales flow says buyer fit depends on mounting, payload, power, communication, and software interfaces, making integration work a meaningful switching-cost lever. 中 SP001
CP003 Agilink says it was formed within AGIBOT in 2023 and became an independent company in January 2026. 高 SP002, SP005
CP004 Agilink claims it held the number-one dexterous-hand market share in China for Q1 2026. 中 SP002
CP005 C Capital reported in September 2026 that Agilink had delivered more than 8,000 OmniHand units and over 10,000 grippers overall, while also claiming first-quarter shipment leadership and first-quarter profitability. 中 SP003
CP006 Unite.ai wrote that Agilink's shipments had passed 15,000 dexterous hands by September 2026 and that the company's June 2026 timeline targeted production capacity of 10,000 units per month. 中 SP004
CP007 Agilink's shipment and share narrative shows momentum, but the retained public claims use inconsistent denominators and mostly company-aligned issuers, limiting clean peer benchmarking. 中 SP002, SP003, SP004
CP008 Baidu Baike says Shanghai Zhiyuan New Innovation Technology Co., Ltd. holds roughly 67% of Agilink's shares, with Tencent- and SAIC-linked entities also listed as shareholders. 中 SP005
CP009 Agilink's flagship positioning centers on the OmniHand Ultra-M, a direct-drive model that the company says debuted in June 2026 and that Unite.ai describes as a frontier-research hand. 高 SP002, SP004
CP010 Unite.ai reports that the OmniHand Ultra-M lists 20 active degrees of freedom, fingertip and palm tactile sensing, a 3-kilogram stable grasping payload, and software assets including C++/Python SDKs, simulation packages, and ROS 2 reference nodes. 中 SP004
CP011 C Capital says AGIBOT's G2 robot is already operating on factory floors at Longcheer Technology and SAIC, which gives Agilink a parent-led deployment path into live systems. 中 SP003
CP012 ROBOTERA reported a May 2026 financing round of more than USD 200 million and claimed product-market fit with deployments across more than ten logistics centers plus thousand-unit deliveries in Q2 2026. 中 SP006
CP013 ROBOTERA also says it built more than 95% of core components in-house and is pushing a full direct-drive dexterous hand architecture, showing that adjacent robotics vendors can internalize the same subsystem Agilink sells. 中 SP006
CP014 Humanoid Index presents Shadow Robot as a commercial company founded in 1987 whose dexterous hand remains positioned as a 24-degree-of-freedom research standard. 中 SP007
CP015 Public pricing and packaging transparency are weak across the retained alternatives: Agilink asks buyers to contact sales after task review, while the retained ROBOTERA and Shadow Robot sources disclose no list pricing. 中 SP001, SP006, SP007
CP016 Agilink's clearest edge in this retained pool is ecosystem velocity rather than independently benchmarked moat durability, because parent-linked distribution proof is stronger than external win/loss, pricing, or reliability evidence. 中 SP003, SP004, SP005, SP006, SP007
CI001 Agilink says its dexterous-hand program was formed within AGIBOT and became an independent company in January 2026. 中 SI004
CI002 Agilink's current public catalog points to a hardware-led revenue model centered on OmniHand dexterous hands and OmniPicker grippers rather than a disclosed recurring software subscription. 中 SI003, SI004
CI003 The visible sales workflow is consultative because buyers are asked to confirm application fit, mounting, power, communications, software, budget range, and timeline with sales before selecting a configuration. 中 SI003, SI008
CI004 Agilink says it has 60+ team members today and is scaling toward 100+ by Q4 2026. 中 SI004
CI005 Agilink says OmniHand and OmniHand Pro launched in August 2025 and that the 1,000th dexterous hand rolled off the production line in December 2025. 中 SI004
CI006 Agilink says the OmniHand 3 Lite, OmniHand 3 Ultra-T, and OmniPicker 3 launched in April 2026 and OmniHand 3 Ultra-M debuted in June 2026. 中 SI004
CI007 Company-issued September 2026 coverage says Agilink has shipped more than 15,000 dexterous hands to date. 高 SI001, SI002
CI008 Baidu Baike reports that cumulative OmniHand deliveries exceeded 8,000 units and cumulative gripper shipments exceeded 10,000 units as of June 2026. 中 SI005
CI009 None of the reviewed official commerce pages disclose list prices, realized ASPs, discounting, or contract terms for OmniHand or OmniPicker, and the store routes buyers to sales instead of publishing pricing. 中 SI003, SI008
CI010 Crunchbase News reports that Agilink reached a $1 billion valuation after four funding rounds since its January 2026 spinout and that the new capital would fund model development, data, and hardware. 中 SI007
CI011 Agilink's about page says a June 2026 funding round pushed the company's post-money valuation above US$1 billion. 高 SI004, SI007
CI012 Baidu Baike reports that Shanghai Zhiyuan New Innovation Technology Co., Ltd. holds about 67% of Agilink's shares. 中 SI005
CI013 Humanoids Daily reports that Chinese planners warned humanoid-robotics companies about "highly repetitive products" and sector "bubbles," which raises differentiation and financing risk for fast-scaling vendors. 中 SI006
CI014 The reviewed public source set does not disclose revenue, gross margin, burn, runway, debt obligations, or the split between Agibot-internal demand and third-party customers. 中 SI003, SI004, SI005, SI007
CE001 Agilink's official surfaces define the company around dexterous hands and grippers for robotic manipulation rather than a broad robot-platform offering. 中 SE003, SE004
CE002 The current AGIBOT dexterous-hand collection page visibly lists three models: AGIBOT OmniHand 2025(w.Tactile), OmniPicker 3, and AGIBOT OmniHand 2025. 中 SE003
CE003 The public catalog separates dexterous hands from grippers and frames selection around objects, movements, host-robot fit, and gripping requirements. 中 SE003, SE006, SE007
CE004 Agilink's official timeline says OmniHand and OmniHand Pro launched in August 2025, the Omni 3 series launched in April 2026, and OmniHand 3 Ultra-M debuted at ICRA in June 2026. 中 SE004, SE005
CE005 Independent September-October 2026 coverage says OmniHand 3 Ultra publicly demonstrated silk-thread separation, needle threading, hooping, and stitching. 中 SE001, SE002
CE006 The same demo coverage describes OmniHand 3 Ultra as a direct-drive hand with 21 active degrees of freedom and tactile sensing on fingertips or palm. 中 SE001, SE002
CE007 The demo-training narrative says operators first controlled the hand remotely and then refined behavior through reinforcement learning. 中 SE001
CE008 Baidu Baike describes Agilink's business scenarios as embodied-intelligence algorithm and data collection, robot integration, and industrial or scientific-research deployments. 中 SE005, SE004
CE009 The AGIBOT store tells buyers to validate mechanical mount, payload or power needs, communication, and control-software compatibility before selecting a configuration. 中 SE003, SE006
CE010 Humanoids Daily says APC 2026 presented OmniHand 3 Ultra-T alongside AGIBOT's GO-2 (VILLA), Genie Sim 3.0, WITA Omni, and AIMA or Link-U OS stack components. 中 SE008
CE011 Because Agilink says it was formed within AGIBOT and unveiled the Omni 3 series at AGIBOT's partner conference, the public record points to material ecosystem dependence rather than a separately documented standalone software stack. 中 SE004, SE008
CE012 Public volume disclosures rise from more than 8,000 OmniHand deliveries by June 2026 to more than 15,000 dexterous hands by September 30 and then to more than 20,000 products shipped on the homepage, but those denominators are not directly aligned. 中 SE005, SE001, SE004
CE013 Agilink's about page says the team had more than 60 members and was scaling toward more than 100 by Q4 2026. 中 SE004
CE014 Public materials describe different OmniHand 3 variants rather than one reconciled spec: demo coverage says OmniHand 3 Ultra has 21 active degrees of freedom and direct drive, while APC coverage says OmniHand 3 Ultra-T is a 22+3-DOF tendon-driven system. 中 SE001, SE002, SE008
CU001 Public company materials position AGILINK around dexterous hands and grippers for both research and industry rather than a single end market. 高 SU001, SU002
CU002 The public buying process is application-engineered, with fit determined by the object, robot platform, and integration requirements rather than by self-serve checkout. 高 SU001, SU008
CU003 AGILINK says the program was formed within AGIBOT as the dexterous-hand supplier for AGIBOT's humanoid robotics platform, making the parent ecosystem the clearest named early-demand surface. 高 SU002, SU003
CU004 AGILINK's about page shows first product launch in August 2025 and a 1,000th-hand production milestone in December 2025, so the visible commercialization window is still short. 中 SU002
CU005 Baidu Baike reports that cumulative OmniHand deliveries exceeded 8,000 units and gripper shipments exceeded 10,000 units by June 2026. 中 SU003
CU006 The assigned pool does not surface named third-party end customers with production outcomes, so public proof is materially stronger on use-case breadth and shipment milestones than on account-level validation. 中 SU001, SU002, SU003
CU007 AGILINK's about page shows an ecological-partner and location footprint across Europe, the US, the Middle East, Southeast Asia, Japan, Beijing, Shenzhen, and Singapore without naming customers or volumes by region. 中 SU002
CU008 Yicai reports that AgiBot expects CP Robotics to help develop business applications in life sciences, new retail, new consumption, and health and wellness, suggesting Agilink's vertical expansion may still rely on parent-led channels. 中 SU002, SU004
CU009 Humanoids Daily's summary of Morgan Stanley's China survey says 62% of respondents are likely to adopt humanoid robots within three years but only 23% are satisfied with current products. 中 SU005
CU010 The same survey summary says only about 10% of respondents are currently evaluating or launching pilot projects, which signals slow deployment conversion even among early adopters. 中 SU005
CU011 Search discovery around the name 'Agilink' is noisy enough to surface unrelated software and cable manufacturers, so external customer proof requires domain-level verification before attribution. 中 SU006, SU007
CU012 None of the assigned sources provides NRR, GRR, churn, renewal, contract length, or top-customer concentration data for AGILINK. 中 SU001, SU002, SU003, SU004, SU005
CU013 The public evidence supports real demand and shipment momentum, but it still does not separate affiliate demand from third-party repeat purchasing. 中 SU002, SU003, SU004
CR001 AGILINK says the business was formed within AGIBOT before being established as an independent company in January 2026. 中 SR001
CR002 Baidu Baike reports that Shanghai Zhiyuan New Innovation Technology Co., Ltd. holds approximately 67 percent of Agilink's shares. 中 SR005
CR003 Agibot origin plus reported majority ownership make parent and control concentration a high residual governance risk until board, IP, and related-party terms are verified. 高 SR001, SR005
CR004 AGIBOT says OmniHand compatibility depends on the host robot's mechanical mount, payload, power supply, communication, and software interfaces. 中 SR002
CR005 MANUS says OmniHand control can require fingertip retargeting and inverse-kinematics translation because several joints are mechanically coupled. 中 SR003
CR006 The combination of explicit host-platform compatibility requirements and partner-described retargeting complexity makes deployment reliability a high residual execution risk. 高 SR002, SR003
CR007 The retained coverage includes an Agilink-sponsored IEEE Spectrum demo and an independent New Atlas recap of the embroidery demonstration, but neither is independently benchmarked production reliability data. 中 SR004, SR008
CR008 IFR says China's robotics push is strategically important while actual humanoid capabilities in production scenarios remain concentrated in demonstrators or pilot projects. 中 SR007
CR009 Together, the sponsored demo evidence and IFR's pilot-stage market context imply material risk in converting showcase manipulation into repeatable commercial deployments. 中 SR004, SR007
CR010 Crunchbase says Agilink reached a $1 billion valuation after four funding rounds since the January 2026 spinout from Agibot. 中 SR006
CR011 Agilink's official about page says the company had 60-plus team members and was scaling toward 100-plus by Q4 2026 after multiple launches and funding rounds. 中 SR001
CR012 The combination of rapid headcount expansion and fast financing cadence raises execution risk around process maturity, quality systems, and go-to-market discipline. 高 SR001, SR006
CR013 The retained public package reviewed for this chapter still does not provide audited revenue, margin, or customer-concentration disclosure needed to underwrite a $1 billion valuation cleanly. 高 SR001, SR006
CR014 Baidu says Tencent- and SAIC-linked investors became shareholders, but the same source still shows majority control with the Agibot-linked shareholder. 中 SR005
CR015 The retained public evidence for this chapter does not surface named licenses, certifications, enforcement actions, or litigation outcomes for Agilink, so legal and regulatory clearance remains a diligence item rather than a cleared risk. 中 SR001, SR005, SR007
CR016 Financing and valuation risk remains high until audited external-customer revenue quality and concentration data support the post-money unicorn narrative. 高 SR001, SR006
CV001 Agilink was incubated within AGIBOT in 2023 and had become an independent company by January 2026. 高 SV001, SV003
CV002 Dealroom shows a June 2026 financing event of $138M at a reported $1.4B valuation, confirming a mid-2026 unicorn-level private mark. 中 SV001
CV003 Public sources consistently describe a rapid 2026 financing cadence spanning Series A, A+, and a later midyear round. 中 SV003, SV004, SV005
CV004 Official company-controlled pages show Agilink is selling both OmniHand dexterous hands and OmniPicker grippers for research and object-handling use cases. 高 SV002, SV003
CV005 Gasgoo reported that mass production and delivery of OmniHand 3 Lite were slated to begin in the second quarter of 2026. 中 SV007
CV006 Dealroom and Baidu Baike both report cumulative OmniHand deliveries above 8,000 units by June 2026, which indicates early commercial traction but not revenue quality. 中 SV001, SV004
CV007 Crunchbase places Agilink among May 2026 unicorns and says the financing would fund model development, data, and hardware for the broader robotics market. 中 SV005
CV008 IFR says China's 2026-2030 industrial plan places robotics at the center of national modernization, supporting investor appetite for the category. 中 SV006
CV009 IDC forecasts global humanoid-robot shipments above 510,000 units by 2030 and says competition will increasingly center on real-world deployment capability and value delivery. 中 SV008
CV010 IFR explicitly warns that mass adoption of humanoid robots in factories and households is not a near- or medium-term expectation. 中 SV006
CV011 Independent industry reporting says China only published its first national humanoid standard system in February 2026 and still faces tactile-sensing and safety bottlenecks. 中 SV009
CV012 Baidu Baike reports that Shanghai Zhiyuan New Innovation Technology holds about 67% of Agilink's shares, indicating control remains concentrated around the Agibot ecosystem. 低 SV004
CV013 The reviewed public sources disclose valuation marks and shipment claims but do not verify revenue, gross margin, repeat-order quality, or the preference stack behind the current price. 中 SV001, SV003, SV004, SV005
CV014 The best-supported current call is research-more with a stretched valuation stance because upside evidence is real but price-sensitive diligence gaps remain material. 中 SV002, SV005, SV006, SV008, SV009
CV015 Relative to other 2026 robotics unicorn disclosures, Agilink's $1.0-1.4B range looks investable enough to stay engaged but not obviously cheap enough to offset the current information asymmetry. 中 SV001, SV005
来源
编号出版方标题引文
SO001 New Atlas Video: Dexterous robot hand learns the delicate art of silk embroidery Agilink, a spin-off of Chinese robotics giant AgiBot, asked Fu Xianghong, a master of traditional Suzhou embroidery, to train its OmniHand 3 Ultra.
SO002 AGILINK AGILINK | Make Dexterity Real We design and manufacture dexterous manipulation systems for the next generation of robots from high-DoF dexterous hands to industrial-grade adaptive grippers, supporting research, industrial automation, and service robotics.
SO003 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Explore OmniHand dexterous hands and OmniPicker grippers. Choose an end effector around the objects, movements and robot platform in your application.
SO004 Baidu Baike AGILINK AGILINK, officially Shanghai AGILINK Innovation Intelligent Technology Co., Ltd., is a developer of robotic dexterous hands founded on January 14, 2026, focusing on providing dexterous manipulation solutions for
SO005 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK was formed within AGIBOT, developing dexterous hands as core components for AGIBOT’s humanoid robotics platform.
SO006 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Compatibility depends on the mechanical mount, payload, power supply, communication and software interfaces.
SO007 Humanoids Daily China’s Robot Fever Meets Reality: Beijing Warns of “Bubbles” and “Repetitive” Clones Li Chao, a spokesperson for the NDRC, cautioned companies to prevent the risk of "highly repetitive products" entering the market.
SO008 Sina Finance 智元拆出一家“灵巧手”独角兽,成立不足5月估值破10亿美元 由智元机器人拆分出的灵巧手研发商临界点(AGILINK)已完成新一轮数亿元人民币的融资,公司估值突破10亿美元,正式成为具身智能赛道又一只“光速独角兽”。
SM001 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Confirm the mounting interface, power supply, communication protocol and control software for the host robot.
SM002 Humanoids Daily AGILINK Hits Unicorn Status: AGIBOT's Dexterous Hand Spin-off Raises Hundreds of Millions Dexterous hands, which historically account for roughly 20% of a humanoid robot's total bill of materials, are widely viewed as the ultimate bottleneck for practical utility.
SM003 Baidu Baike AGILINK The company's business primarily targets three major scenarios: embodied intelligence algorithm and data collection, Humanoid Robot and general robot integration, and dual-scenario implementation in industrial and scientific research settings.
SM004 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK secured #1 market share in China for Q1 2026.
SM005 Crunchbase News AI Services And Robotics Lead Diverse Crop Of 29 New May Unicorns Shanghai-based Agilink has raised four funding rounds since it was spun out of Agibot in January, and reached a valuation of $1 billion. Agilink is focused on dexterous hand technology. The funding will be used for model development, data and hardware with the spinout able to license to the broader robotics market.
SM006 International Federation of Robotics China Makes AI-powered Robots Core of National Strategy But mass adoption as universal humanoid factory helpers or in private households will not happen within the near- and medium-term future. The 15th Five-Year plan sees the commercialization of humanoid robots rather towards the end of the plan´s period. Wide adoption of AI with traditional industrial robotics is expected over the next five to ten years.
SM007 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Explore OmniHand dexterous hands and OmniPicker grippers. Choose an end effector around the objects, movements and robot platform in your application.
SM008 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Explore OmniHand dexterous hands and OmniPicker grippers. Choose an end effector around the objects, movements and robot platform in your application.
SP001 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT A dexterous hand offers multiple fingers and joints for a broader range of hand movements. A gripper is designed around grasping tasks.
SP002 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK was formed within AGIBOT, developing dexterous hands as core components for AGIBOT’s humanoid robotics platform.
SP003 C Capital C Capital-Backed AGIBOT Tops World Humanoid Robot Games with 46 Medals; AGILINK Wins 7 of 8 Golds Founded only in January 2026, it has already delivered more than 8,000 OmniHand units and over 10,000 grippers overall, led the Chinese market in dexterous-hand shipments in the first quarter of the year, and turned a profit in its very first full quarter of operation
SP004 Unite.AI AGILINK Robot Hand Takes Up Suzhou Embroidery as Shipments Pass 15,000 Published specifications for AGILINK’s flagship direct-drive model, the OmniHand Ultra-M, list 20 active degrees of freedom in a 630-gram, human-scale build
SP005 Baidu Baike AGILINK In the company's equity structure, Shanghai Zhiyuan New Innovation Technology Co., Ltd holds approximately 67% of the shares.
SP006 RoboticsTomorrow ROBOTERA Raises Over USD 200 Million in New Round Led by SF Group, HSG and IDG Capital The company has achieved the first product-market fit (PMF) in the embodied intelligence sector, with deployments across more than ten logistics centers in collaboration with China Post and SF Group. In Q2 2026, ROBOTERA initiated thousand-unit deliveries, with growth exceeding 300%.
SP007 Humanoid Index Shadow Robot Company: Funding, Valuation, Robot Specs & More | Humanoid Index Dexterous Hand. World-leading robot hand with 24 degrees of freedom. Research standard.
SP008 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Confirm the mounting interface, power supply, communication protocol and control software for the host robot.
SI001 Humanoid Robotics Technology AGILINK Ships 15,000 Dexterous Robotic Hands Spun out from AGIBOT in January 2026, the company has shipped more than 15,000 dexterous hands to date.
SI002 PR Newswire via Yahoo Finance Chinese robot hand learns Suzhou embroidery as AGILINK surpasses 15,000 dexterous hands shipped Spun out from AGIBOT in January 2026, the company has shipped more than 15,000 dexterous hands to date.
SI003 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Share your task requirements, location, budget range and project timeline.
SI004 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK was formed within AGIBOT, developing dexterous hands as core components for AGIBOT’s humanoid robotics platform.
SI005 Baidu Baike AGILINK As of June 2026, cumulative deliveries of its OmniHand Series Dexterous Hands have exceeded 8,000 units, and cumulative shipments of grippers have surpassed 10,000 units.
SI006 Humanoids Daily China’s Robot Fever Meets Reality: Beijing Warns of “Bubbles” and “Repetitive” Clones With over 150 humanoid robot companies now operating in China, Li noted that "speed and bubbles" are perennial issues in frontier industries that must be "managed and balanced."
SI007 Crunchbase News AI Services And Robotics Lead Diverse Crop Of 29 New May Unicorns Shanghai-based Agilink has raised four funding rounds since it was spun out of Agibot in January, and reached a valuation of $1 billion.
SI008 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Compatibility depends on the mechanical mount, payload, power supply, communication and software interfaces.
SE001 The Rundown AI Agilink teaches a robot hand the basics of embroidery Agilink says the hand has 21 active degrees of freedom, or independently controlled movements, and relies on direct drive throughout. Sensors at the fingertips read how the contact surface changes shape, giving the system feedback as it works.
SE002 Rocking Robots AGILINK Demonstrates Robotic Hand on Suzhou Embroidery Tasks - Rockingrobots AGILINK said the OmniHand 3 Ultra has 21 active degrees of freedom and uses a direct-drive architecture. The system incorporates tactile sensors on the fingertips and palm, including image-processing-based sensors designed to detect changes in the contact area while handling thin and deformable materials.
SE003 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Compatibility depends on the mechanical mount, payload, power supply, communication and software interfaces. Share the host robot and intended task with our team before selecting a configuration.
SE004 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK is a global robotics company dedicated to robotic manipulation systems — the dexterous hands and grippers that allow robots to interact with the physical world.
SE005 Baidu Baike AGILINK The company's business primarily targets three major scenarios: embodied intelligence algorithm and data collection, Humanoid Robot and general robot integration, and dual-scenario implementation in industrial and scientific research settings.
SE006 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Consider OmniPicker for gripping tasks. Check object dimensions, mass, surface and required gripping force.
SE007 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Consider OmniHand for multi-finger movements, human-like gestures and manipulation research.
SE008 Humanoids Daily Deployment Year One: AGIBOT Unveils Massive Fleet and AI Model Stack at APC 2026 OmniHand 3 Ultra-T: A next-generation dexterous hand featuring a 22+3 DOF tendon-driven system and 3D tactile sensing, aimed at matching human-level precision in industrial assembly.
SU001 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Compatibility depends on the mechanical mount, payload, power supply, communication and software interfaces. Share the host robot and intended task with our team before selecting a configuration.
SU002 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK was formed within AGIBOT, developing dexterous hands as core components for AGIBOT’s humanoid robotics platform.
SU003 Baidu Baike AGILINK As of June 2026, cumulative deliveries of its OmniHand Series Dexterous Hands have exceeded 8,000 units, and cumulative shipments of grippers have surpassed 10,000 units.
SU004 Yicai Global Mirae Asset-LG Electronics Fund Invests in Chinese Humanoid Robot Startup AgiBot, Report Says AgiBot announced on July 15 that it had received an investment from CP Robotics, with the Danish robotics software firm expected to help it develop business applications in vertical sectors such as life sciences, new retail, new consumption, and health and wellness.
SU005 Humanoids Daily Morgan Stanley Survey: Chinese Companies Want Humanoids, But the Products Aren't Ready According to the survey, a staggering 62% of respondents indicated they are likely to adopt humanoid robots within the next three years. However, this enthusiasm crashes into a reality wall: only 23% of those surveyed expressed satisfaction with current products, citing issues with dexterity, functionality, and price.
SU006 Agilink Technologies Agilink Technologies — Custom Software, Built · Modernized · Run For 24 years, hundreds of companies — from Fortune 500 energy leaders to sole proprietors — have brought us the systems their business runs on: to build them, to take them over, and to keep them running.
SU007 Agilink Group Home - Agilink Agilink Group designs and produces cables and cords for many sectors of activity, such as telecommunications, transport, defense, manufacturing and aviation.
SU008 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Share your task requirements, location, budget range and project timeline.
SR001 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK was formed within AGIBOT, developing dexterous hands as core components for AGIBOT’s humanoid robotics platform.
SR002 AGIBOT Store Dexterous Robotic Hands & Grippers | AGIBOT Compatibility depends on the mechanical mount, payload, power supply, communication and software interfaces.
SR003 MANUS Agilink OmniHand Retargeting and Multimodal Data Collection with MANUS Several joints are mechanically coupled, and the distribution of active degrees of freedom differs from the anatomical hand model captured by MANUS.
SR004 AGILINK via IEEE Spectrum sponsored article Beyond Dexterity: Why Contact May Define the Next Era of Robotics At ICRA 2026, AGILINK demonstrated the capabilities of its OmniHand dexterous hand by folding balloons.
SR005 Baidu Baike AGILINK
SR006 Crunchbase News AI Services And Robotics Lead Diverse Crop Of 29 New May Unicorns Shanghai-based Agilink has raised four funding rounds since it was spun out of Agibot in January, and reached a valuation of $1 billion.
SR007 International Federation of Robotics China Makes AI-powered Robots Core of National Strategy the actual capabilities in real-world production scenarios are currently limited to demonstrators or pilot projects.
SR008 AGIBOT Dexterous Robotic Hands & Grippers | AGIBOT Confirm the mounting interface, power supply, communication protocol and control software for the host robot.
SV001 Dealroom AGILINK - Unicorn company profile | Dealroom Within approximately five months of its founding, AGILINK completed multiple funding rounds, raising hundreds of millions of yuan and achieving a post-money valuation of over $1 billion, granting it unicorn status.
SV002 AGIBOT Store Dexterous Robotic Hands & Grippers A dexterous hand offers multiple fingers and joints for a broader range of hand movements. A gripper is designed around grasping tasks.
SV003 AGILINK About AGILINK | Dexterous Manipulation Systems AGILINK completed a new funding round, reaching a post-money valuation exceeding US$1 billion.
SV004 Baidu Baike AGILINK In the company's equity structure, Shanghai Zhiyuan New Innovation Technology Co., Ltd holds approximately 67% of the shares.
SV005 Crunchbase News AI Services And Robotics Lead Diverse Crop Of 29 New May Unicorns Shanghai-based Agilink has raised four funding rounds since it was spun out of Agibot in January, and reached a valuation of $1 billion.
SV006 International Federation of Robotics China Makes AI-powered Robots Core of National Strategy But mass adoption as universal humanoid factory helpers or in private households will not happen within the near- and medium-term future.
SV007 Gasgoo Seeds | AGILINK Raises Hundreds of Millions With its valuation surpassing $1 billion, the company has officially joined the unicorn club.
SV008 IDC From Task Execution to Value Creation: What the 2026 Humanoid Robot Half Marathon Reveals About Industry Progress Humanoid robotics is transitioning from technically feasible to commercially viable.
SV009 Robotics & Automation News Why China's New Humanoid Robot Standards Could Change the Industry In other words, a robot might know it shouldn’t crush a human hand, but without reliable tactile sensing, it may not realize it is crushing one until it is too late.