初创公司尽调
尽调报告 Robotics / hardware Late-stage private 2026-05-19

Unitree Robotics

腿式与人形机器人 — 商业牵引真实,地缘政治折价也真实

Unitree 是少数拥有公开产品定价、真实收入信号和明显规模的人形与足式机器人公司之一;但政策风险和披露缺口仍要求保持克制的观察立场,而不是激进买入。

封面要素

成立时间 01
2016 [CO001]
总部 02
Hangzhou, China [CO001]
2025 年人形机器人出货量 06
5500 units+ [CO029, CI016]

公司概况

Unitree Robotics 是一家总部位于杭州的腿式与人形机器人公司,由 Wang Xingxing 于 2016 年创立。公司的打法是用明显低于许多西方同行的价格,把高性能机器人商业化,同时把更多组件栈掌握在内部。公开证据显示,Unitree 的产品线已经很宽,覆盖 Go2 四足机器人、B2 工业机器人、G1 与 H1 人形机器人、Z1 机械臂、激光雷达、SDK 和强化学习工具。Unitree 的产品既直销,也通过面向开发者的渠道销售,并不只是停留在实验室原型;这一点把它和几家估值很高的人形机器人同行区分开来。公司仍为私有、财务不透明,但公开牵引力已经足够,应该把它视为真实运营中的机器人企业,而不是概念阶段故事。

官网
www.unitree.com
成立时间
2016-01-01
创始人
Wang Xingxing
创立地点
Hangzhou, China
总部
Hangzhou, China
产品
产品组合覆盖 Go2 消费者 / 开发者四足机器人、B2 工业四足机器人、G1 与 H1 人形机器人、Z1 机械臂、激光雷达,以及开放的开发者工具,包括 SDK、ROS2 包和强化学习代码库。
客户
研究实验室、开发者、教育、工业巡检与自动化用户、公共活动运营方,以及评估平价腿式与人形机器人的国际分销商或集成商。
商业模式
以硬件销售为主,组件、开发者工具及相关支持 / 服务补充收入;人形机器人主要面向开发者,四足机器人则覆盖消费者和工业工作流。
阶段
Late-stage private (post-Series C, IPO-prep reported)
融资情况
2025 年 6 月 Series C 轮据报道将 Unitree 估值约 RMB 12 billion(约 US$1.6B-US$1.7B),IPO 准备已启动;具体融资额和完全核对后的投资方名单仍未在公开来源中确认。
[CO001, CO002, CO003, CO006, CO020, CO021, CO025, CO029]

执行摘要

主要优势

  • Unitree 把机器人商业化信号做得异常透明:公开产品定价、直销渠道,以及 2024 年据报超过 RMB 1 billion 的收入。
  • Go2 和 G1 的性价比领先,使 Unitree 相对高成本西方同行具备颠覆性位置,也扩大了开发者采用。
  • 宽产品组合和自研组件 / 控制栈让产品面更有韧性,不只是单一机器人创业故事。
  • GitHub 上的公开开发者生态支撑研究人员、集成商和早期商业开发者采用。
  • 杭州及中国更广泛的机器人政策支持,有利于制造、标准和生态动能。

主要风险

  • 美国采购禁令、出口管制外溢和中国来源安全审查,可能在企业牵引成熟前就挡住西方需求。
  • 毛利率、现金、烧钱速度、保修负担和分部组合均未披露,限制了承销信心。
  • 公开客户证明在研究、演示和可见度上更强,在长期工业部署上更弱。
  • 围绕人形平台的网络安全和隐私担忧,可能放慢受监管环境中的采用。
  • 若从 US$1.6B-US$1.7B 私募估值迈向高得多的 IPO 估值,需要公开记录中尚未出现的证据。

未决问题

  • Series C 具体募资额和完全确认的投资人名单。
  • 经审计的 2024-2025 财务报表,包括毛利率、烧钱速度和现金跑道。
  • 四足机器人与人形机器人收入组合,以及渠道折扣后的实际平均售价。
  • 具名的重复采购工业客户,包括部署规模、留存和服务经济性。
  • 人形机器人 fleet 的详细安全姿态、遥测实践和长期支持承诺。

目录

Chapter 01

01公司概况

1.1 身份、创立故事与使命

官方与独立来源都一致把 Unitree Robotics 描述为一家总部位于杭州、由 Wang Xingxing 在 2016 年创立的机器人公司。创立故事重要,因为它解释了公司的商业 DNA:Wang 告诉 KrASIA,自己在研究生阶段做出 XDog 四足机器人,意识到电驱动能大幅降本并提升灵活性,随后短暂加入 DJI,又在约两个月后离开,于 2016 年 5 月创立 Unitree。低成本、快迭代的逻辑,今天仍体现在公司的定位里。官方材料把公司放在民用机器人赛道,强调技术驱动进步的使命,并把 Unitree 定义为一个覆盖四足机器人、人形机器人、机械臂和核心组件的宽平台,而不是一次性演示产品。表述偏理想主义,但主线很稳定:靠可负担价格和速度推动普及,而不是只做定制化研究系统。这种组合也解释了为什么 Unitree 会同时出现在消费者和开发者场景中。[CO001, CO002, CO003, CO004, CO005, CO006]

快照 KPI 表
指标数值 / 状态日期置信度缺口 / 注意事项
成立20162016官方、Xinhua 和 Forbes 来源一致
总部中国浙江杭州2026-05-19没有公开证据显示总部迁移
创始人 / 最高运营负责人Wang Xingxing — 创始人、CEO、CTO2026-05-19Wang 之外的公开管理层仍稀疏
产品范围四足机器人、人形机器人、机械臂、激光雷达、SDK2026-05-19未披露按产品划分的客户 / 收入组合
证据最充分的最新估值RMB 12B 至 12B+(约 US$1.64B–1.7B)2025-06 至 2025-07Series C 准确金额和投资方名单仍不一致
累计融资未完全确认;Humanoid Index 汇编为 $200M+2026-02-28需要股权结构表或监管文件支持
公开员工数背景1,000+ 名员工2026-02-28由媒体 / 聚合器汇编,并非公司申报
公开出货背景2025 年 5,500+ 台人形机器人;2026 年目标 10k–20k 台2026-02-17目标是前瞻性公司指引
IPO 准备进行中;CITIC Securities 据报道担任顾问2025-07-21正式申报文本在本会话中无法完全获取

混合了官方公司声明、可信媒体报道和聚合器背景。估值比累计融资或准确投资方组合更有支撑。

[CO001, CO003, CO006, CO021, CO022, CO024]
FO002: Unitree 公司快照逻辑

创始人主导的研发、平价产品策略、开发者工具、公开展示和资本形成,如何共同强化 Unitree 的当前位置。

[CO003, CO005, CO006, CO009, CO017, CO018]

1.2 领导力集中与开发者界面

管理层证据同时很强,也很薄。强在 Wang Xingxing:所审阅来源都一致显示他是创始人兼 CEO,Forbes 还把他标为 CTO,这与访谈和公司公开工程足迹中呈现的技术优先画像相符。薄在其余班底。已审阅材料没有披露董事会名单、独立董事,或单独具名的 CFO/COO 层级,因此围绕 Wang 存在真实的关键人依赖。部分抵消这种不透明的是,Unitree 作为硬件公司,却有异常公开的开发者界面。其 GitHub 组织公开了覆盖 G1、H1、Go2、B2、H2 及其他机器人家族的 SDK、ROS2 包、强化学习代码库和应用模板。这样的开源姿态支持一个判断:Unitree 不只是卖成品机器人;它还试图播种开发者生态,扩大采用面,并在研究人员和开发者中强化品牌。[CO003, CO009, CO010, CO011, CO012, CO013]

领导层和创始人表
人物 / 领导节点当前角色背景 / 证据点创始人-市场匹配或职能覆盖关键人依赖
Wang Xingxing创始人、CEO、CTO研究生项目中做出 XDog;2016 年创办 Unitree 前曾短暂在 DJI 工作在运动控制、商业化和公开战略上有直接的创始人-产品匹配非常高
未披露的更广泛高管团队董事会 / CFO / COO 未清晰公开已审阅材料未显示具名董事名单或完整财务 / 运营团队内部可能有职能覆盖,但公开证据不足

列举有意保持部分,因为 Unitree 公开材料未披露完整高管团队或董事会。该表只捕捉唯一有充分个人证据支持的领导节点,以及公开披露缺口本身。

[CO002, CO003, CO037, CO040]
FO003: Unitree 快照 KPI 与信号

最能锚定 Unitree 当前成熟度、规模和尽调注意点的六个核心事实。

混合使用官方说法、可信媒体报道和聚合器背景。出货量和累计融资数字不是审计披露。

[CO001, CO009, CO021, CO022, CO029, CO039]

1.3 融资、估值与 IPO 准备

目前最可靠的资本锚点不是投机性的 IPO 标题,而是 2025 年 6 月的 Series C 轮。TechNode 报道 Unitree 投后估值超过 RMB 12 billion(约 US$1.64 billion),Forbes 独立报道为 RMB 12 billion(约 US$1.7 billion),The Robot Report 也落在相同区间。关键限定是,具体融资额,甚至完整投资方名单都没有完全确认。TechNode 点名 China Mobile、Tencent、JinQiu Capital、Alibaba、Ant Group 和 Geely;Forbes 验证了 Geely、Ant Group 和 HongShan;Humanoid Index 汇总该轮约 US$96 million;The Robot Report 则提到第三方估算接近 US$97.6 million。这些信息在独角兽以上估值方向上相互印证,但还不够精确,不能把它当成已经坐实的股权结构表事实。已经坐实的是 IPO 准备:TechNode 称 CITIC Securities 担任顾问,后续报道称申报准备正在推进,这意味着 Unitree 这家私有公司正走向更正式的信息披露。[CO020, CO021, CO022, CO023, CO024, CO025]

利益相关方或投资人地图
利益相关方 / 投资人角色控制权或经济重要性当前证据状态尽调问题
Wang Xingxing创始人控制人核心运营中心,且可能是投票控制中心后续 IPO 准备报道称他控制 34.76% 股份核实投票权、超级投票条款和董事会控制
Geely / Geely Capital据报道的 Series C 投资方汽车 / 制造业战略信号TechNode 和 Forbes 点名确认实体、金额和权利
Ant Group据报道的 Series C 投资方大型金融科技战略背书信号TechNode 和 Forbes 点名确认规模和董事会 / 观察员权利
HongShan / Sequoia China据报道的投资方重要风险投资背书信号Forbes 和后续 IPO 准备报道点名确认这是新增资金还是既有股东延续持有
Tencent / Alibaba / China Mobile / JinQiu 投资方据报道的领投方群组潜在战略渠道分发和政治信号价值TechNode 和 / 或第三方报道点名,尚未完全对齐索取已签署的轮次公告和交割表
其他后续具名持有人(Meituan、Source Code 等)IPO 前股权背景可能更多反映早期轮次,而非 2025 年 6 月交割出现在后续 IPO 准备报道中,不能干净等同于 Series C 领投名单区分新股轮投资人与当前股权结构持有人

该表有意保留冲突。它区分 2025 年 6 月轮次报道内容,以及后续 IPO 准备报道对股东基础的说法。

[CO020, CO021, CO022, CO023, CO024, CO025]

1.4 里程碑、公众形象与新兴风险

作为私有机器人公司,Unitree 的里程碑记录异常公开。官方材料和独立报道显示,它登上过 2021 年春晚舞台、2022 年冬奥会开幕式、2023 年 Super Bowl 赛前秀、2023 年亚运会和 CES 2025;TIME 后来也把 Unitree 列入 TIME100 companies list。这些亮相既服务品牌,也提供商业证明,尤其是因为 Unitree 现在横跨消费级四足机器人、面向开发者的人形机器人、工业四足机器人和机械臂。但可见度也会带来双刃剑效应。Humanoid Index 汇总显示员工超过 1,000 人、2025 年出货超过 5,500 台,CnTechPost 报道 2026 年人形机器人目标为 10,000-20,000 台;这些是有用的规模信号,但不是经审计财务披露。更重要的是,反向证据已经出现。The Wire China 记录了华盛顿对 Unitree 美国销售的审查升温,一篇 2025 年 arXiv 论文则指称 G1 存在严重安全与遥测风险。叠加仍不清晰的治理细节,这些风险应该和里程碑故事一并呈现,而不是被放在故事之外。[CO014, CO015, CO016, CO017, CO018, CO019]

里程碑表
日期事件类型金额 / 状态参与方 / 来源背景影响
2016-05Wang 将 XDog 转为公司后,Unitree 在杭州成立成立公司启动KrASIA、Xinhua、Forbes创始故事把降本运动控制经验直接连到业务
2017官方材料称 Unitree 开始把足式机器人推向工业应用规模化商业化推进官方关于页面显示早期就有走出实验室演示的意图
2021-02机器人登上 CCTV 春节联欢晚会合作全国转播里程碑KrASIA 和官方历史大众认知度跃升
2022-02参与冬奥会开幕式合作官方声称官方关于页面增加主流背书
2023-02亮相 Super Bowl 赛前秀合作官方声称官方关于页面 / Wire China释放国际宣传价值信号
2023参与亚运会和亚残运会合作官方声称官方关于页面强化本地主城和国家级形象
2024GitHub 开发者触点扩展到 SDK、ROS2、RL 和 URDF 资产产品观察到的开源足迹GitHub 组织和仓库开发者生态成为公司故事的真实组成部分
2025-01蛇年春节联欢晚会人形机器人表演合作16 台人形机器人表演 YangBot官方关于页面、TIME、Global Times人形机器人的国内能见度峰值
2025-06Series C 以约 RMB 12B 估值完成融资证据最充分的私募市场估值约 US$1.6B–1.7BTechNode、Forbes、Robot Report独角兽以上估值锚
2025-07据报道由 CITIC Securities 辅导 IPO 准备治理IPO 前准备进行中TechNode 和后续 IPO 准备报道披露制度可能随时间趋严
2025-09安全研究指称 G1 存在攻击向量和遥测风险反向独立反向技术报告arXiv 安全论文治理和产品信任风险更加具体
2026-03中国发布人形机器人标准体系监管国家标准发布来源:People.cn / Xinhua部署政策背景更正式

日期反映本会话审阅的最佳公开证据。官方展示历史对品牌牵引有用,但财务披露仍远远落后于能见度。

[CO001, CO002, CO009, CO014, CO015, CO016]
FO001: Unitree 公司里程碑时间线

按日期梳理 Unitree 的成立、公开展示、融资、IPO 准备和政策 / 风险拐点。

时间线在日级公开证据对分析并非必要处采用月份级精度。

[CO001, CO002, CO014, CO015, CO016, CO018]

1.5 图表与要点

Chapter 02

02市场分析

2.1 市场边界:腿式机器人,不是所有自动化

Unitree 横跨两个重叠但仍然不同的市场:四足机器人已经服务巡检、公共安全、开发者和研究工作流;人形机器人则通过开发者、表演、接待和早期工业实验走向商业化。公司也销售机械臂和传感器,但这些产品今天的商业意义主要在于拓宽具身机器人栈,而不是创造一个独立的软件市场。因此,一个有用的市场边界应包括腿式系统上的机器人硬件、机载算力、感知包、集成、维护和开发者工具支出;不包括固定工业机械臂、传送带自动化、纯软件 AI 和无人机,除非买方明确在同一工作流中做替代。 这个边界重要,因为 Unitree 并不是先攻最成熟的自动化预算。它的直营网店、GitHub 代码库和平价 G1 入门点,显示出开发者、教育者和实验者中的强牵引;B2 巡检四足机器人则面向巡检、应急响应等更结构化的工业任务。公开证据还不支持把整个制造自动化市场都视为 Unitree 的可服务市场(SAM)。在许多工作流中,人力、AMR、协作机器人和固定工业机器人仍是强替代品,IFR 也明确表示,人形机器人在中短期不会成为万能帮手。正确框架是分层市场:四足机器人工作流今天已经商业化;人形机器人工作流也真实商业化,但仍集中在低风险、低问责场景。[CM001, CM002, CM003, CM004, CM005, CM006]

市场定义表
细分市场 / 类别纳入支出排除支出主要买方 / 付款方为什么对 Unitree 重要
四足机器人工业巡检和应急响应机器人硬件、感知栈、电池、维护、集成、远程运营仅固定 CCTV、人工巡逻劳动力、设施重设计公用事业运营商、工业安全团队、公共安全机构B2 已围绕该细分市场的地形、续航和载荷做规格定义
四足机器人研究和开发者市场机器人采购、SDK 访问、配件、实验室集成、实验载荷纯仿真软件和非足式研究平台大学、实验室、机器人创业公司、资助项目GitHub 活动加上 Go1 学术研究,显示这是当前真实预算线
人形机器人开发套件和具身 AI 实验机器人硬件、开发工具、计算模块、数据采集、操作员培训未绑定当前能力的通用劳动力替代叙事开发者、AI 实验室、大学、创业原型团队G1 价格和二次开发定位,使其成为当前真实可触达的切入市场
人形机器人表演、接待和品牌体验机器人租赁或采购、编舞 / 集成、活动支持、内容管理广义企业自动化项目活动运营商、品牌、场馆、公共展示IDC 称,表演和接待仍是 2025 年实际需求最大的场景之一
人形机器人工业试点机器人硬件、工厂集成、受监督部署、安全配置、维护缺乏合同证据的大规模劳动力替代假设制造商、仓库、集成商长期具战略重要性,但相较今天的展示活动或开发者需求,仍处更早阶段
排除的相邻自动化无——本行标记排除项固定工业机械臂、协作机器人、AMR、无人机、纯软件 AI现有自动化买家这些类别仍是替代品,默认不应计入 Unitree 可服务支出,除非有工作流级替代证据

边界依据产品目录、第三方对买方组合的报道,以及 IFR 替代逻辑。纳入支出专指足式机器人部署;排除支出列出不应在缺乏工作流级替代证据时滚入 Unitree TAM 的自动化类别。

[CM001, CM002, CM004, CM006, CM016, CM018]
FM001: 市场规模测算视角

四层金字塔展示长期 TAM 叙事、当前可测人形机器人收入、Unitree 可触达 SAM,以及 Unitree 可观察 SOM 下限之间的差异。

这座金字塔有意混合已发布和推断层。只有前三层直接来自来源;第四层是证据约束下的 SAM 估计,第五层是公司的可观察收入底盘。

[CM007, CM008, CM011, CM013, CM014, CM035]

2.2 用彼此冲突但有用的视角测算市场

关于 Unitree 机会的公开市场规模证据相互矛盾,因为不同来源在衡量不同东西。Goldman Sachs 给出很宽的长期人形机器人区间,从 US$6B 到 US$154B,取决于成本、用例和公众接受度障碍能否清除。相比之下,IDC 支持的报道给出更落地的当前快照:2025 年约 18,000 台人形机器人、约 US$440M 收入。Xinhua 引用 Morgan Stanley,指向一个大得多的 2050 年中国单一市场场景:价值 6 trillion yuan、出货 59 million 台。Forbes 又加了一层,把 Goldman 的硬件市场框架和 ARK Invest 的劳动力价值框架放在一起比较,而后者根本不是硬件收入预测。 落到 Unitree 身上,最能支持决策的视角既不是蓝天 TAM,也不是最保守的当前市场快照。可观察的可获取市场(SOM)地板已经能从公司据报道 2024 年超过 1 billion yuan 的收入和 2025 年人形机器人出货规模中看到;相对 2025 年测算的人形机器人市场,这个规模不小,但仍几乎不说明毛利率或经常性收入质量。短期可服务市场(SAM)比纯人形机器人更宽,因为 Unitree 今天就能从四足巡检和研究预算中变现;但它又比“全球劳动力替代”叙事窄,因为实际需求仍偏向表演、教育、数据采集和早期工业用途。也因此,本章保留相互矛盾的估算,而不是强行制造虚假的精确度:市场显然真实,显然在增长,但还远没有一个定数。[CM007, CM008, CM009, CM010, CM011, CM012]

TAM / SAM / SOM 和规模测算视角表
视角发布方 / 依据地域 / 时间范围数值方法置信度限制
长期人形机器人 TAM(基准情景)Goldman Sachs全球 / 10–15 年US$6B 下限如果采用仍受约束,分析师给出的人形机器人市场情景范围很宽,且前瞻性强
长期人形机器人 TAM(蓝天情景)Goldman Sachs全球 / 2035US$154B 上限假设设计、可负担性和接受度重大障碍均被清除的情景不是中心情景
可测量的当前人形机器人市场IDC 经 CGTN全球 / 202518,000 台;~US$440M 收入报道的出货量和收入快照仅人形机器人;不含四足机器人市场
中国人形机器人长期情景Morgan Stanley 经 Xinhua中国 / 20506T yuan;59M 台机器人自上而下的长期情景地域和时间范围不同于 Goldman 与 IDC
四足机器人结构性市场信号IDC全球 / 2024中国公司成为领先者全球四足机器人市场份额研究公开摘要未披露完整收入总量
可观察的 Unitree 收入下限TechNode + The Robot ReportUnitree / 2024>1B yuan 收入报道的公司上一年收入无细分拆分
可观察的 Unitree 人形机器人份额视角IDC 支持的出货量 + Unitree 报道交付Unitree / 2025Unitree 交付 5,500+ 台人形机器人,对照 18,000 台全球市场快照基于出货量的当前市场切片交付数量几乎不能说明已实现收入质量
近期可触达 SAM(分析师约束)本章综合Unitree 可触达的 2026–2028 细分市场约 US$0.3B–US$1.0B结合可测量的人形机器人收入基数,以及当前四足巡检和研究需求不是已发布的第三方市场估计

这些已发布视角有意不合并成一个数字,因为口径不兼容。最后一行是本章受约束综合,不是第三方市场报告;应把它当成证据边界内的规划区间,而不是已验证市场规模。

[CM007, CM008, CM010, CM011, CM013, CM015]
FM002: 市场估计区间

区间图保留相互矛盾的市场估计,而不是假装它们能收敛成单点估计。

纳入 Morgan Stanley 的中国 2050 年情景,是为了保留矛盾,不是暗示它能与 Goldman 或 IDC 做近期比较。Unitree 可触达 SAM 区间是本章估计,使用今天可测得的需求结构。

[CM008, CM011, CM013, CM014, CM038]

2.3 买方分层:工业巡检、开发者、研究与推广用途

Unitree 的买方地图,比长期人形机器人叙事暗示的更碎片化。四足机器人中,工业巡检、应急响应、公共安全和公用事业式工作流,是当前预算最清晰的场景,因为 B2 产品围绕载荷、续航和复杂地形明确设计。The Wire China 也记录了美国大学、地方警察和陆军的真实需求,其中可负担价格是决定性因素。学术实验室和开发者构成第二个清晰分层:GitHub 工具、SDK 文档,以及外部学术论文中对 Go1 的使用,都指向可重复的研究和教育需求。这类买方往往来自 R&D、实验室或院系资本预算,而不是企业自动化预算。 人形机器人已经商业化,但当前买方组合不如巡检分层成熟。IDC 称 2025 年出货主要由表演、研究与教育、数据采集和接待拉动。这与 Unitree 在春晚上的高可见曝光相符,也与公司围绕 G1 的开发者导向信息相符。未来家庭使用是管理层公开愿景的一部分,但已抓取证据还看不到规模化家庭收入市场。最重要的实践区分因此不是抽象的“消费者还是企业”,而是买方购买机器人,是为了可重复、ROI 可测的工作流,还是为了开发、实验、数据采集或推广可见度。Unitree 当前两边都参与,这种混合需求画像同时带来机会和估值模糊性。[CM003, CM004, CM005, CM016, CM017, CM018]

细分市场 / 买方地图
细分市场买方用户付款方 / 预算负责人采用触发因素当前证据
工业巡检与应急响应公用事业管理者、工业安全负责人、公共安全采购方巡检员、救援人员、一线团队运营或安全预算危险地形、长时巡逻、降低人员风险B2 产品定位与 Wire China 公共安全采购案例
开发者与具身 AI 原型开发机器人初创公司、AI 实验室、独立开发者工程师、ML 研究员、机器人操作员研发预算可负担的人形或四足硬件,加上 SDK 访问G1 定价与 GitHub SDK 生态
学术与研究机构实验室负责人、PI、院系管理员学生、研究员、博士后课题经费或实验室资本预算需要低成本移动平台做运动控制和导航研究RomiLab 和 Zenodo 上的 Go1 论文
表演与接待品牌方、场馆、活动运营方、城市展示项目表演者、导览员、访客营销或运营预算吸引注意力、新鲜感或导览互动春节联欢晚会曝光度与 IDC 用例排名
海外市场的公共安全与高校警察部门、校园、国防或市政项目安保人员、学生、研究团队部门采购预算预算受限但需要能干活的足式机器人Wire China 报道的警察、高校和 U.S. Army 案例
未来家庭与家务辅助消费者或照护服务运营方家庭和照护者消费或辅助生活预算在安全价位上可用的通用自主能力管理层野心可见,但当下规模化证据偏弱

买方分层基于公开产品界面、外部学术使用案例,以及独立报道中的实际客户。“未来家庭”被纳入,是因为它会影响估值叙事;但相对今天可见的购买证据, 它仍偏愿景。

[CM003, CM004, CM017, CM018, CM019, CM024]
FM003: 买家 / 细分市场图

矩阵把 Unitree 当前和未来最相关的六类买家,映射到用户画像、付款方、预算权和采用触发因素。

[CM016, CM017, CM018, CM019, CM024, CM036]

2.4 中国顺风真实存在,但采用约束仍然有效

Unitree 受益于两个异常强的顺风。第一,中国制造生态和密集供应商基础压缩组件成本和生产周期,帮助本土机器人厂商比许多西方同行更快商业化。第二,政策栈正在变得更明确:2026 年人形机器人与具身 AI 国家标准、杭州“全球总部港”集群,以及已经把机器人视为战略基础设施的本土环境,都提高了更快本地测试、认证和采购的概率。这些不是抽象政策口号;它们在试生产、测试和供应链寻源中降低摩擦。 但这些顺风并不会消除核心约束。IFR 仍警告,人形机器人近期并不是传统工业机器人的万能替代品。IDC 的 2025 年市场组合显示,行业真实需求中心仍比炒作周期暗示的更早期。Unitree 还面临超出工程问题的信任摩擦:网络安全研究凸显了广泛攻击面,美国政治审查也已经在公共部门语境中把公司框定为安全风险。低价今天有助于赢单,但如果软件和服务尚未成熟、市场就先商品化,低价也可能预示未来毛利率压力。实际结论是,中国供应链和政策顺风会加快商业化节奏,但采用仍取决于具体工作流的 ROI、安全信心,以及地缘政治上可接受的采购路径。[CM020, CM021, CM022, CM023, CM026, CM027]

增长驱动因素与约束表
因素方向时间对 Unitree 的影响有证据支撑的解读尽调需索取
中国密集的零部件生态驱动因素当下支撑更低 ASP 和更快迭代供应商竞争和集聚让 Unitree 在成本与速度上占优要求披露 COGS 趋势和供应商集中度
2026 年人形机器人标准与具身 AI 政策支持驱动因素当下至 2030改善国内测试、合规和采购路径中国已从泛泛鼓励转向具体标准梳理 Unitree 已满足哪些标准
杭州机器人集群驱动因素当下至 2027提升本地测试和产业化能力集群基础设施可缩短商业化闭环跟踪 Unitree 是否使用新测试评估基础设施
低入门价与直销驱动因素当下扩大开发者、学术和准专业消费者漏斗价格是客户选择 Unitree 而非西方替代品的直接原因评估低价是在增厚还是摊薄利润率
劳动力短缺与以人为中心场景适配驱动因素中期让人形机器人投资逻辑仍具战略意义IFR 认可真实需求逻辑,但不认同短期普及把 Unitree 试点与量化的劳动力替代挂钩
当前需求集中在表演、研究和数据采集约束因素当下收入质量可能落后于叙事质量已测得的用例结构,比家庭或工厂热叙事更早期要求按终端市场拆分出货结构
传统机器人仍是强替代品约束因素当下限制高吞吐、结构化工作流中的 SAM形态优势只在特定以人为中心任务中成立按工作流将 Unitree 与协作机器人、AMR 和无人机对标
网络安全与责任担忧约束因素当下拖慢敏感环境采用人形机器人扩大攻击面,也放大安全责任问题要求第三方安全审计和事故历史
美国及盟友安全审查约束因素当下至中期压低海外公共部门 TAM,并可能外溢到企业采购围绕中国机器人的政治叙事已经启动跟踪采购限制和出口管制动作
生态与服务转向做不到是约束 / 做到则是驱动2026 年以后Unitree 必须走出只卖便宜硬件IDC 预计竞争会转向服务和生态索取软件附加率、服务收入和开发者留存指标

本表区分有证据支撑的驱动因素与约束因素。几行是双刃剑:低价拉动需求,也可能挤压利润率;政策支持利好国内市场, 但无法抵消海外审查。

[CM020, CM021, CM022, CM027, CM028, CM029]
FM004: 采用漏斗 / 价值链图

示意性漏斗:长期机器人需求会收窄到今天 Unitree 在技术、经济和地缘政治上都能触达的任务。

漏斗数值只指示方向,并非来源原生数据。它综合了 IFR 对采用节奏的谨慎判断、Wire China 关于采购摩擦的证据,以及 Unitree 当前以产品驱动的商业化基础,说明总可用市场(TAM)不能混同于眼前可获取收入。

[CM020, CM031, CM032, CM033, CM035]

2.5 图表与要点

Chapter 03

03竞争格局

3.1 竞争版图 — 直接同行、现有玩家与现状替代方案

Unitree 并不是在一个干净的绿地市场里竞争。直接人形机器人同行包括追求工业或家庭部署叙事的西方初创公司(Figure、1X、Agility、Boston Dynamics),以及密集的中国队列,覆盖全尺寸人形机器人、娱乐机器人和四足替代品(Agibot、Fourier、UBTech、DEEP Robotics)。公开证据还显示,买方经常可以完全避开人形机器人:Agility 明确把 Digit 定位为连接既有自动化孤岛,Boston 已经把 Spot 卖进巡检工作流,DEEP 把四足机器人卖进公用事业和救援任务,Goldman / IFR 都认为可负担性和用例成熟度仍制约采用。这意味着 Unitree 不仅要打败其他人形机器人,也要打败固定工业自动化、AMR、四足机器人和劳动密集型人工工作流等已被验证的现状栈。赛道已经足够拥挤,低成本本身不会带来赢家通吃。[CP010, CP015, CP018, CP022, CP023, CP024]

竞争对手画像表
公司 / 替代品类别规模 / 融资信号目标客群产品范围核心差异化核心限制
Unitree Robotics中国低成本具身机器人厂商2025 年出货 5,500+ 台;声称收入 $140M;1,000+ 名员工开发者、研究机构、价格敏感的企业边缘用例G1、H1、Go2、机械臂、激光雷达、灵巧手、SDK公开市场上最便宜的人形机器人标杆,加上广泛开放的开发者栈公开企业验证薄,保修期短,安全说明明确强调早期阶段
Figure美国 AI 主导的人形机器人初创公司$39B 估值;BMW 试点;商业化前工业试点和家庭辅助叙事Figure 02 劳动力;Figure 03 家庭;Helix AI初创公司中资本最厚、AI 品牌最强价格不透明,披露的部署规模有限
1X Technologies挪威 / 美国家庭人形机器人初创公司$125M+ 融资;200+ 名员工;预购阶段家庭机器人和远程操作辅助自主NEO、NEO Gamma软体安全家庭设计,远程操作兜底资本有限,公开工业客户验证有限
Agility Robotics美国工业人形机器人初创公司$641M+ 融资;约 100 台商用单位仓储、物流、制造Digit + Arc + 服务公开工作流栈最完整,且有具名企业用户价格未披露,对开发者爱好者市场不够可及
Boston Dynamics美国工业机器人老牌公司Hyundai 支持;Atlas 选择性推出;Spot 已实地部署工业自动化、巡检、物料搬运Atlas、Spot、Orbit企业信任、载荷和集成纪律Atlas 不能公开购买,价格不透明
Agibot中国人形机器人和表演机器人竞争对手声称部署 1,000+ 台 A2;公开电商价格商业表演、展览、具身 AI 平台A2 Ultra、X2、A2 Lite、平台软件价格和认证透明,表演用途组合更广支持责任仍推回买方;西方验证有限
DEEP Robotics中国四足机器人替代品在公用事业、救援和公共安全中有工业部署巡检、巡逻、危险环境四足系统机动性比手更重要时,它是务实替代品不是直接的通用人形平台
UBTECH中国机器人老牌公司商业和教育覆盖面广人形服务、教育、养老、消费硬件人形及更广机器人组合纯人形初创公司之外的装机基础经验公开材料覆盖广,但对当前旗舰人形机器人的经济性深度不足
Fourier中国聚焦灵巧操作的人形机器人竞争对手官方推进 GR-2 产品,配 SDK 和触觉手开发者和企业实验GR-2 人形机器人相较 Unitree G1 基础版,公开的灵巧与触觉叙事更强价格未披露,商业部署规模不公开
现状替代品固定自动化、AMR、四足机器人和人工成熟行业,装机基础深厚仓库、工厂、巡检、安全机械臂、AMR、Spot 类四足机器人、人工工作流ROI、支持和操作流程更清晰比成功人形机器人可能提供的通用移动性和具身能力更弱

规模与融资字段混合公司直接披露、分析师或媒体报道;它们只是尽力收集的公开信号,不是经审计财报。纳入现状替代品, 是因为很多买方仍可完全绕开人形机器人。

[CP010, CP012, CP014, CP016, CP018, CP019]
FP001: 竞争定位图

基于证据的序位定位:横轴是价格可及性,纵轴是企业级就绪度 / 信任。x 越高,公开购买越容易、前期成本越低;y 越高,已披露的工作流验证、支持和信任越强。

坐标轴是作者基于已发布价格透明度、工作流软件、具名客户、支持披露和安全 / 政策姿态做出的序位判断,并非独立基准测试分数。

[CP015, CP018, CP019, CP024, CP031, CP039]

3.2 价格 / 性能 — Unitree 的基准具有颠覆性,但不等同于企业就绪

Unitree 最强的公开证据是经济性,而不是具体工作流。G1 标价约 $13.5K-$16K,Go2 为 $2,800,而西方人形机器人同行仍依赖未披露的商业条款、订金或选择性部署漏斗。即便 Agibot 透明定价的人形机器人,标价也高于 Unitree。这让 Unitree 成为面向开发者、实验室和边缘用例企业买家的行业内最可见价格性能锚点。但规格表也说明,为什么不能把公司误认为最企业就绪的平台。G1 公布的手臂载荷只有约 2-3 kg,灵巧手是选配而非标配,保修窗口较短,并明确提示行业仍处于早期。Fourier 公布了更强的灵巧操作叙事,Atlas 公布了更重载荷和工业集成,Agility 公布了工作流控制和支持。因此,Unitree 更常赢下首次采购对话,而不是整支车队的生产部署对话。[CP001, CP002, CP003, CP004, CP005, CP017]

功能 / 能力矩阵
能力 / 采购标准UnitreeFigure1XAgilityBoston DynamicsAgibotFourier
公开标价仅定金 / 目标价
官网直接购买路径线索收集消费者候补 / 定金可用线索收集线索收集线索收集
开放开发者 SDK / 代码仓库Unknown公开资料有限公开资料有限公开资料有限公开资料有限
公开载荷 / 扭矩规格部分有限有限有限
公开灵巧手叙事可选 / 部分强叙事通用叙事功能可用但公开细节不多聚焦工业操作因型号而异
具名企业部署研究 / 公共安全买方BMW 试点家庭测试 / 公开验证有限Amazon / GXO / SchaefflerSpot 已有装机基础;Atlas 早期公开部署声称未公开详述
工作流软件 / 服务层开发者工具多于企业栈Unknown聚焦消费者 UXArc + 支持Orbit + 支持平台 / AIMMASTER开发者工具包
西方市场公开信任 / 合规姿态弱到中弱到中

强 / 中 / 弱标签是基于公开产品页和报道的有证据支撑的顺序判断,不是第三方基准分。未知或有限意味着能力可能存在, 但公开记录不足。

[CP003, CP007, CP011, CP013, CP015, CP017]
价格 / 打包对比
厂商 / 产品公开价格或套餐信号包含 / 承诺内容分销模式含义
Unitree G1$13.5K-$16K 标价基础人形机器人、可选灵巧手、面向开发者的二次开发电商直销 + 销售联系把入门级人形机器人的价格下限打下来
Unitree Go2$2,800 标价配激光雷达的四足机器人、App 引导建图、可选更大电池电商直销在人形机器人以下价位挤压巡检和研究替代品
Agibot X2$24,240 标价娱乐 / 商业人形机器人,可选自主套件电商直销价格透明度上最接近 Unitree 的中国竞争者
Agibot A2 Lite$44,560 标价全尺寸表演机器人,一年 / 3000 小时保修电商直销显示即便是价格透明的中国同业,也常明显高于 G1
1X NEO$200 定金;据报道目标价约 $20K家庭机器人预购路径,靠远程操作辅助学习消费者候补 / 定金在可负担叙事上比西方工业同业更接近 Unitree
Figure 03价格未披露由 Helix 驱动的家庭人形机器人叙事线索收集 / 选择性推出AI 差异化公开,经济性不公开
Agility Digit价格未披露机器人加 Arc 工作流控制和支持企业销售更高切换成本或可支撑更高但不透明的定价
Boston Atlas价格未披露带 Orbit 集成的工业物料搬运机器人选择性接触企业客户竞争点在就绪度和信任,而不是标价透明度
Fourier GR-2价格未披露53 个关节的人形机器人,配 12 自由度手部和 SDK 支持联系销售灵巧操作更靠前,但价格透明度低于 Unitree
现有替代方案通常按项目、工时或机器人工作站定价ROI 和运营模式已有验证由集成商或现有供应商主导人形机器人在 TCO 和可维护性上要赢过的不只是同类竞品

本表只比较公开标价或套餐信号。信息不透明的行不默认昂贵或便宜;没有公开价格本身就有信息量,因为 Unitree 和 Agibot 更像产品化购买,而不是定制工业项目。

[CP001, CP002, CP014, CP019, CP031]
FP002: 功能广度 / 能力图

对与 Unitree 最相关的主要人形机器人竞争者,按公开能力披露做序位比较。

强 / 中 / 弱取值概括的是公开披露密度和部署证据,不是单一实验室基准。“未知”表示公开材料证据不足,无法有把握评分。

[CP007, CP011, CP013, CP015, CP018, CP024]

3.3 分销、制造规模与开发者准入

作为人形机器人供应商,Unitree 的销售路径异常开放。它直接在线销售,在公开代码库发布开发者工具;第三方报道称,其人形机器人面向会自定义运动和感知的开发者,而不是购买封闭黑盒的客户。这种开放降低了进入门槛,也解释了为什么 Unitree 在研究、大学、公共安全试点和分销商主导的海外销售中有牵引。它也可能支撑了据报道 2025 年 5,500+ 台出货和公司 1,000+ 人员工规模。取舍在于,开放准入并不等于企业锁定。Agility 把 Digit 包进 Arc 加服务 / 支持,Boston 强调 Orbit 集成和工业推广纪律,Figure / 1X 试图拥有封闭 AI 或家庭用户体验。因此,公开渠道宽度看起来是优势,但也可能掩盖了高接触部署和售后支持中仍有多少工作回落到客户和分销商身上。Unitree 能比同行更快做宽漏斗顶部,但其当前分销模型似乎更适合开发者采用和渠道宽度,而不是高接触企业部署支持。[CP006, CP007, CP008, CP009, CP010, CP015]

3.4 护城河耐久性 — 成本曲线和生态真实存在;信任、灵巧操作和政策仍是弱点

Unitree 护城河中耐久的部分很直接:中国制造快、组件成本低、机器人品类宽、开放开发者生态让平台容易试用。这些属性很难被西方初创公司快速复制。薄弱部分同样清晰。Unitree 尚未披露与 Agility 和 Boston 相当的工作流软件、具名企业部署深度或安全 / 合规姿态。华盛顿的政策话术越来越常把中国机器人框定为安全问题,围绕 G1 的独立安全研究又放大了低价不等于企业信任这一点。实践中,这意味着 Unitree 很适合把入门级具身机器人商品化,尤其是在研究和价格敏感用例中;但当买方最在意灵巧操作可靠性、支持、隐私或采购审查时,公司仍然暴露。因此,护城河在成本和宽度上强,在生态上中等,在信任驱动的锁定上弱。[CP027, CP032, CP033, CP034, CP036, CP037]

护城河耐久性 / 竞争风险登记表
护城河主张主要威胁严重性当前证据缓释措施 / 尽调问题
人形机器人最低可见价位同行降价或补贴部署G1 标价仍是公开基准,但 Agibot 和 1X 也在讲低成本故事跟踪同行价格披露,并确认 Unitree 扩量时能否守住毛利率
宽产品线的具身机器人组合产品线更窄的同行靠旗舰产品执行力追上Unitree 覆盖四足、人形、机械臂、激光雷达和开发者资产验证交叉销售是否真的转化为经常性平台收入
开放开发者生态开放工具降低锁定,也方便复制SDK2、ROS2、RL 和模板让 Unitree 容易上手,也容易离开索取能加深留存的应用商店、云或经常性软件指标
中国供应链成本优势政策冲击或零部件限制削弱供给、推高成本第三方报道把主要成本优势归因于中国零部件生态验证供应商集中度、出口管制敞口和备选采购路径
借助经销商快速出海分销支持质量和合规深度落后于企业客户预期Wire China 显示美国销售由经销商主导,但直接本地运营存在感有限按地区索取 SLA 覆盖、备件仓和企业直服人员数量
研究和边缘部署的上市速度受监管买家转向更重信任的西方替代方案安全批评和采购禁令话术给 Unitree 造成非价格阻力压测政府相关或隐私敏感行业的胜率

本登记表评估竞争耐久性,而非纯技术优劣。即使护城河有价值,只要它依赖政策、信任或支持假设,而这些变量变化快于硬件成本,护城河仍可能脆弱。

[CP032, CP033, CP034, CP036, CP037, CP038]
FP003: 护城河 / 就绪度 KPI

这些压缩事实最能说明 Unitree 的颠覆性来自哪里,以及这份颠覆仍有哪些硬约束。

[CP001, CP007, CP010, CP015, CP017, CP037]

3.5 图表与要点

Chapter 04

04财务情况

4.1 收入模式与价格架构

公开证据强烈指向一个硬件主导的收入模式。Unitree 销售消费级和工业级四足机器人、面向开发者的人形机器人、机械臂、激光雷达和相关工具;KrASIA 在 2021 年引用 Wang 的说法称,大部分收入已经来自机器人销售,表演只占小部分。这个历史评论仍符合当前目录。TechNode 补充了一个重要分层线索:四足机器人主要面向消费者,人形机器人则面向会在 Unitree 硬件上修改感知和运动算法的开发者。价格梯度也强化了这种分层。Go2 在官方页面起价 US$1,600,在商店页面为 US$2,800;G1 产品页标为 US$13.5K,商店页标为 US$16K;H1 标价 US$90K;STEMfinity 将 Z1 机械臂列为 US$13,387.50;B2 则似乎处在报价驱动的工业工作流中,而不是完全透明的电子商务路径。这些都是标价锚点,不是实际 ASP,因为税费、运输、海关、渠道折扣和售后义务均未披露。[CI001, CI002, CI003, CI004, CI005, CI006]

收入流表
收入流机制单位当前数值 / 状态质量尽调问题
四足机器人销售通过官网页面、商城和经销商直接销售硬件每台机器人有公开证据(Go2、B2 系列)按渠道和地区获取实际 ASP
人形机器人销售 / 开发者版本向开发者和高阶买家销售硬件每台机器人有公开证据(G1、H1),带开发者定制角度拆分研究 / 开发销售和企业试点
机械臂 / 外设硬件附件销售(Z1 机械臂、激光雷达、组件)每台设备有公开证据,但结构未披露量化配件搭售率和毛利率
SDK / 软件赋能支持第三方定制和部署的工具链N/A开发者信号强;收入模式未披露澄清软件是打包销售、授权收费,还是仅用于促成采用
服务 / 支持 / 表演售后支持和少量表演收入项目或支持合同KrASIA 称表演在 2021 年占比很小分别拆出维护、集成和活动收入

收入流由目录证据、媒体报道和观察到的开发者工具链推断。公司未披露经审计的收入结构。

[CI001, CI002, CI003, CI004, CI025, CI026]
定价 / 变现表
产品 / 档位可见公开价格主要核验来源置信度关键限制
Go2US$1,600 官方预告价;US$2,800 商城入门价;也展示更高档位官方 Go2 页面 + 官方商城不含运费、税费、关税和实际折扣
G1官方页面 US$13.5K;商城标价 US$16K官方 G1 页面 + 官方商城不同页面可能对应基础配置与电商配置
H1US$90,000 商城标价官方商城页面不含关税;企业实际成交价未知
B2询价驱动 / 核验正文未清楚给出价格官方和经销商页面工业销售动作比消费级 SKU 更不透明
Z1经销商标价 US$13,387.50STEMfinity 商品页转售商价格;官方经销商页面确认渠道,但正文未给出价格

本表有意把可见标价和询价状态放在一起,因为 Unitree 没有发布一套干净的全球价格表。B2 仍是公开定价上最大的缺口。

[CI005, CI006, CI007, CI008, CI009, CI010]
FI001: Unitree 收入模型桥

Unitree 的产品系列和销售渠道如何连接硬件、配套件和支持收入。

[CI001, CI002, CI003, CI004, CI025, CI026]

4.2 公开牵引与收入背景

Unitree 有足够的公开牵引信号,说明它不是商业化前公司,但这些来源仍不足以拼出干净财务模型。TechNode 报道 Unitree 2024 年收入超过 RMB 1 billion(US$137 million),The Robot Report 也转述了约 US$140 million 的粗略数字。Humanoid Index 补充称,截至 2026 年 2 月,公司员工超过 1,000 人,2025 年出货超过 5,500 台;CnTechPost 报道 Wang 对 2026 年人形机器人出货的目标为 10,000-20,000 台。正确用法是把这些数字当作中等置信度背景,而不是审计事实。它们来自可信报道和汇总方,但不是财务报表。它们之间也不能整齐映射:2024 年收入早于被引用的 2025 年人形机器人出货爬坡,Unitree 还同时销售价格低得多的四足机器人和外设,以及更贵的人形机器人。因此,投资者应克制,不要试图从少数公开数字倒推出单一混合 ASP 或毛利率故事。公开数据说明的是动能,不是完整透明度。[CI014, CI015, CI016, CI017, CI018, CI030]

单位经济性表
指标数值 / 状态置信度重要性尽调问题
2024 年收入参考值收入:RMB 1B / US$137M–140MIPO 前最佳公开收入锚点用经审计的 FY2024 利润表验证
公开员工数背景1,000+ 名员工暗示薪酬和支持负担不小确认当前员工数,以及制造与 R&D 的拆分
2025 年人形机器人出货背景5,500+ 台有用的需求 / 规模信号将人形机器人与四足机器人、配件分开
2026 年人形机器人出货目标10,000–20,000 台勾勒管理层野心和产能需求索取已订订单和月度生产计划
粗略人均收入单台约 ≈US$137k–140k仅作方向性硬件效率替代指标用经审计收入和实际平均 FTE 重新计算
毛利率未披露资本效率的核心决定因素提供产品线毛利率和质保准备金数据
现金 / 消耗 / 债务未披露决定资金可支撑期和下一轮融资风险提供最新现金余额、现金消耗和债务明细
实际 ASP 和渠道利润率未披露标价不等于净收入按 SKU 提供已订 ASP 和转售商经济性

公开数字存在时,多数来自媒体报道或聚合器编制。核心单位经济性字段仍未披露。

[CI014, CI015, CI016, CI017, CI018, CI023]
FI002: 财务估算区间

公开可观察的 Unitree 关键财务和规模指标区间。

区间混合了报道数据和情景框架。估值和收入区间基于多家独立报道;出货目标来自管理层指引。

[CI009, CI010, CI014, CI015, CI017, CI019]

4.3 资本强度与融资依赖

即便没有公开现金流量表,Unitree 看起来也资本密集。卖机器人而不是纯软件,意味着营运资本、制造规模、执行器、电池、售后支持和渠道经济性都很重要。公开材料没有披露毛利率、现金、债务或烧钱速度,所以核心问题不是 Unitree 是否有牵引,而是这些牵引能否在下一次融资触发前转化为自我供血的经济性。资本时间线有助于界定这个风险。TechNode 和 Forbes 把 2025 年 6 月 Series C 轮锚定在约 RMB 12 billion 估值;后续 eHangzhou 和 Global Times 报道描述了活跃的 IPO 准备工作和预期申报阶段披露。这个顺序意味着,公司仍依赖外部资本和公开市场准备,来弥合可见产品动能与完整财务透明度之间的缺口。因此,融资故事可信但不完整:估值锚点真实,IPO 路径真实,但现金充足性或毛利结构仍没有硬公开证据。[CI019, CI020, CI021, CI022, CI023, CI024]

资本充足性表
资本 / 披露事项日期公开数值 / 状态主要来源置信度含义
Series C 完成2025-06已完成;确切金额未完整披露TechNode / Forbes / Robot Report确认新增资本,但不说明现金余额
Series C 后估值2025-06 至 2025-07RMB 12B 至 12B+(~US$1.64B–1.7B)TechNode / ForbesIPO 前最有支撑的私募市场估值标记
后续 IPO 准备阶段估值背景2025-09eHangzhou 报道为人民币 12B–15BeHangzhou显示申报前估值区间有所上移
IPO 申报准备2025-07 起据报 CITIC Securities 为顾问;预计进入申报阶段披露来源:TechNode / Global Times / eHangzhou支撑正在推进外部融资和披露路径
官方申报文件访问2025-07-18已识别监管相关 PDF 路径,但当前抓取画像返回 405CSRC 披露 URL仍需直接获取一手申报文件
现金 / 资金可支撑期2026-05-19未公开披露无已核验公开来源资本充足性中最核心的未解问题

本表刻意收窄范围:它从最后一轮有支撑的私募融资到 IPO 准备披露预期来框定资本充足性,而不是完整历史融资年表。

[CI019, CI020, CI021, CI022, CI034, CI035]
FI003: 资本强度 / 现金流图

为什么可见产品牵引力尚未消除融资依赖。

[CI023, CI024, CI027, CI028, CI034, CI040]

4.4 尽调阻碍与财务结论

本章结论很直接:Unitree 有真实商业证据,但毛利率可见度仍太弱,难以干净承销。公司显然在硬件上变现,展示了异常宽的产品阶梯,也有足够的据报道收入和出货背景,值得重视。但缺失变量正是决定可投资性的变量——BOM、服务成本、保修负担、渠道折扣、积压订单转化、现金跑道、债务,以及申报阶段文件中的实际披露。可访问的申报路径本身在本次会话中仍不完美:监管机构链接的 PDF 路径可以识别,但对抓取配置返回 405 响应;这提醒我们,公开报道和一手披露不是同一回事。在 Unitree 申报或以其他方式开放数据室之前,投资者应把公开收入和出货数字视为牵引的方向性证明,而不是质量证明。这让融资依赖和未解决单位经济性成为主要阻碍,即便收入线故事仍有吸引力。[CI023, CI034, CI039, CI040, CI041, CI042]

公开财务缺口表
缺失指标对投资测算的影响目前公开可得内容缺口为何存在精确尽调路径
按 SKU 的毛利率无法判断硬件经济性只有标价和粗略收入参考值私营公司;尚无申报阶段财务数据按产品系列索取经审计的毛利率桥
在手现金 / 消耗 / 资金可支撑期无法精确评估融资依赖只有 Series C 和 IPO 准备信号已核验公开来源中没有现金流披露索取最新资产负债表、现金流和融资计划
在手订单 / 订单转化无法把出货雄心转化为预测信心只有媒体披露的出货目标目标来自采访,不是正式申报文件索取已订订单、取消情况和交付排期
实际 ASP 和渠道利润率标价阶梯可能夸大变现质量官方页面展示部分 SKU 标价渠道折扣和到岸成本未披露按 SKU、渠道和地区索取已订 ASP
地区 / 客户结构无法测算海外集中度或风险敞口只有零散的经销商轶事证据没有收入分部披露按地区、渠道和客户类型索取收入结构
债务、担保或项目融资义务杠杆和或有负债未知无已核验公开证据私营公司不透明索取债务明细、银行授信、担保和契约条款摘要

这个缺口登记表就是操作性尽调清单,用来把公开叙事信心推进到可投资测算。

[CI023, CI027, CI028, CI034, CI039, CI040]
FI004: 公开财务披露矩阵

核心财务问题哪些已公开披露、只能粗略估算,哪些仍被挡住。

矩阵只反映截至本报告运行日,本次研究审阅到的证据。

[CI014, CI017, CI023, CI039, CI040, CI041]

4.5 图表与要点

Chapter 05

05产品与技术

5.1 产品家族与工作流匹配

作为一家私有机器人供应商,Unitree 公开产品界面异常宽:Go2 面向低成本开发者用途、教育和轻载野外场景;B2 面向工业巡检和重载移动;G1 与 H1/H1-2 面向人形机器人 R&D、遥操作、灵巧操作和展示部署;Z1 把产品栈延伸到机械臂和机载操作。共同商业模式不是单一交钥匙工作流,而是硬件平台策略:买方先选择一种机体形态,再围绕它叠加二次开发、RL 策略、ROS2 控制或集成商支持。这一点重要,因为公司证据最强的工作流仍由工程牵引,而不是终端消费者自动化。Go2 页面突出建图、侧跟随、4D lidar、OTA 更新和二次开发;B2 强调全地形巡检、爬楼梯、越障、载重和可选充电;G1 与 H1/H1-2 强调灵巧性、高扭矩密度、激光雷达加深度感知和开发者扩展。Z1 公开信息更薄,但它明确被定位为面向移动机器人载荷工作、物流和研究的开放编程机械臂。公开活动亮相提升认知,但最强的工作流证据仍来自研究、巡检和集成商主导部署,而不是大众市场自主家庭使用。 [CE001, CE002, CE003, CE004, CE005, CE006]

产品模块 / 资产矩阵
产品系列主要用户状态 / 成熟度关键组件差异化尽调缺口
Go2 / Go2 Pro / Go2 EDU 系列开发者、教育、轻负载现场用户已出货的零售四足机器人;商业化最成熟的系列自研 4D 激光雷达、OTA、建图、可选 Jetson Orin基础价很低,开发栈可见独立正常运行时间或车队运营参考案例很少
B2 / B2-W工业巡检、重载移动、公共安全试点工业级平台,联系销售部署高扭矩关节、3D 激光雷达、深度相机、可选轮足形态载荷、续航、爬楼梯和越障能力具名生产运营方和服务指标稀少
G1 / G1 EDU人形机器人 R&D、遥操作、操作开发者早期商用人形机器人,零售与联系销售并行23–43 DOF、激光雷达 + 深度相机、灵巧手选项人形机器人实验入门价低人形机器人工作流仍需谨慎,部分仍在开发中
H1 / H1-2高阶人形机器人 R&D、展示、高端集成商大规模商用前的人形机器人商业化阶段360 N.m 级关节声称、激光雷达 + 深度相机、可选 Orin NX高功率密度和全尺寸形态现场可靠性或客户成效的公开证明很少
Z1 机械臂移动机器人制造商、物流和研究用户已成型的机械臂模块,但公开文档偏薄开放编程接口、机载机械臂用例将 Unitree 从移动延伸到操作公开规格深度和参考部署有限

状态反映截至 2026-05-19 运行日期的公开文档;这是商业化视角,不是经审计的出货台账。

[CE001, CE002, CE003, CE004, CE005, CE006]
工作流 / 用例表
用户任务当前工作流痛点Unitree 匹配度可衡量收益信号限制 / 注意事项
研究实验室的运动实验西方机器人平台通常太贵,实验室很难广泛使用Go1/Go2 和人形套件提供成本更低的具身平台,并开放 SDK / ROS / RL 路径实验室和学术论文明确提到价格可承受和适应性公开证据强调实验,而不是长期生产工作周期
工业巡检或危险地形通行人和轮式机器人在楼梯、碎石、陡坡或狭窄空间里吃力B2 和 Go2 把地形机动、相机、激光雷达、远程 / 二次开发结合起来官方页面宣传楼梯、坡道、建图和载荷工作流独立运营方 ROI 和维护数据仍稀缺
人形机器人操作和遥操作 R&D开发者需要全身机器人,以及仿真 / 策略工具链G1、H1/H1-2、Dex 灵巧手、应用模板和世界模型 / RL 仓库拼出连续开发路径GitHub 上可见内容覆盖部署、数据集、仿真和策略训练许多能力仍靠演示驱动,尚未在商业环境验证
公开演示 / 活动主办方要吸睛的机器人表演,并且动作能力一眼可见H1/H1-2、Go1/Go2 及相关平台出现在头部活动中借助晚会、冬季运动会、Super Bowl 和 CES 获得全球曝光曝光不等于可重复的企业级正式采用
系统集成商定制买家要的不只是裸硬件,还要上手培训、支持和软件接口零售页面、分销商和 Docker 应用模板支撑集成商主导的落地官方与分销商页面都强调支持、培训和二次开发伙伴执行力和支持质量是重大依赖

收益项只是公开工作流信号,不保证客户结果。除学术实地测试和公开展示外,独立 KPI 披露有限。

[CE003, CE004, CE005, CE006, CE014, CE017]
FE002: 客户工作流 / 运行流程

开发者或集成商通常从选择 Unitree 平台开始,到在真实硬件上部署策略或应用。

流程综合了公开产品页、SDK README、RL 指南和应用模板说明,并非来自某个官方参考流程。

[CE012, CE013, CE014, CE015, CE016, CE017]

5.2 技术栈、内部组件与开发者工作流

最深的产品技术证据位于机器人外壳之下。Unitree 的介绍页面明确声称内部研发电机、减速器、控制器、激光雷达以及感知和运动控制算法;A1 电机页面把这个主张具体化:执行器把控制器、行星减速器、交叉滚子轴承、传感器、RS485 通信和即插即用维护能力集成到单一模块中。在公开机器人页面中,感知同样处于中心位置:Go2 宣传自研 4D lidar,G1 和 H1/H1-2 则用激光雷达加深度相机做人形感知。软件上,Unitree 公开的栈远多于许多硬件同行。GitHub 代码库覆盖 SDK2、ROS2、Isaac Gym RL、Isaac Lab 仿真、世界模型学习、应用模板、URDF 资产和部署助手。SDK2 支持 Ubuntu、x86_64 和 aarch64 上的当前家族;ROS2 使用 CycloneDDS,并公开运动、低层状态、远程控制和激光雷达的直接话题;RL Gym 规范了 Train-to-Play-to-Sim2Sim-to-Sim2Real 闭环;Isaac Lab 复用来自实体机器人的 DDS 话题;UnifoLM 世界模型代码库为 G1 和 Z1 操作任务加入更有野心的仿真加策略层。结果是,从原始硬件到开发者应用打包和策略部署,有一条可见路径,即便官方文档中心页面仍难以通过静态抓取检查。 [CE008, CE009, CE010, CE011, CE012, CE013]

技术 / 运营架构表
层级 / 组件作用公开证据关键依赖风险
执行器模块将策略和控制指令转成关节运动A1 电机页面披露集成控制器、减速器、传感器、RS485、1 KHz 控制自研执行器设计与制造质量故障率和生命周期数据未公开
具身感知深度感知、建图、避障、状态估计Go2 页面提到自研 4D 激光雷达;人形机器人页面提到激光雷达 + 深度相机传感器校准与加固演示之外的真实环境稳健性只有部分证据
机器人中间件底层状态、运动控制、无线控制、激光雷达 topicSDK2 和 ROS2 仓库暴露 DDS 与 ROS 兼容的控制路径CycloneDDS、Ubuntu 工具链、本地网络开发者能否跑通,取决于 Linux 网络和中间件配置
仿真与策略训练部署前训练、验证并回放运动或操作策略RL Gym、Isaac Lab 和 world-model 仓库展示 sim-to-real 与数据集工作流GPU 资源、Isaac / MuJoCo 环境、数据集质量高计算资源假设和 sim-to-real 偏移仍是关键风险
应用打包与部署将自定义控制逻辑打包成 Unitree 生态服务unitree-app-templates 使用 Docker 和 UniStore 打包HTTP 服务打包、元数据、应用商店审核通过商业部署路径仍受伙伴支持和文档不足制约
OTA 与生命周期管理推送机器人软件更新,让设备持续迭代Go2、B2、H1 和 G1 页面均宣传 OTA 或升级版 OTA云连接、操作者授权、发布纪律回滚、机队分组和补丁节奏未公开细化
制造 / 试生产层将原型转成可重复交付的系统About 页面和杭州生态文章提到产线改进与试生产基础设施供应商质量、测试系统、认证能力未公开良率、缺陷或 RMA 指标

这张表映射公开可见内容,不代表完整披露的内部架构。证据只停留在营销或仓库 README 层面时,风险列会明确写出。

[CE008, CE009, CE010, CE011, CE012, CE013]
FE001: 产品架构图

Unitree 技术栈从一体化驱动与感知出发,经 DDS 中间件,延伸到仿真、策略训练、打包和现场更新层。

由于 Unitree 尚未发布单一权威架构图,层级边界由产品页和代码库 README 推断。

[CE008, CE009, CE010, CE011, CE012, CE013]

5.3 部署成熟度与制造姿态

公开证据显示 Unitree 正在走出实验室原型,但产品组合内部并不均衡。公司反复宣称自己在四足机器人公开零售上领先,商店页面通过公布 Go2 和 G1 的直销价格、运输披露和二次开发选项支持这个判断。外部 CES 报道确认,Unitree 愿意在 2025 年初同时向全球买方展示 Go2、G1、H1 和 B2-W,这意味着它有一套连贯产品线,而不是孤立概念车。介绍页面还称 Unitree 正在持续改善产线,并覆盖 ToB 和 ToC 腿式、人形和机械臂品类;杭州机器人枢纽文章又提供了更广的生态信号,描述围绕 Unitree 等公司的试生产、测试、认证和培训能力正在成形。不过,成熟度并不一致。四足机器人更接近可部署产品,已经有载荷、续航和建图流程文档;人形机器人销售时仍突出对局限、开发中特性和安全使用的提醒。二级来源称 Unitree 是量产领导者,应作为方向性背景,而不是经审计制造证据。最可辩护的成熟度判断是:Unitree 已在四足机器人和开发者硬件上实现可信商业化;人形机器人仍是快速推进、但尚未完全验证的平台家族。 [CE019, CE022, CE023, CE024, CE028, CE029]

路线图 / 发布 / 开发阶段表
日期 / 阶段功能或里程碑状态含义来源依据
2024G1 公开发布并亮相 CES已发货 / 已演示说明人形机器人项目从概念推进到可销售平台官方 G1 页面;CES 报道
2025H1/H1-2 登上 CCTV 春晚已演示提升曝光,也验证协同运动演示About 页面;Global Times;Xinhua
Sep 2025UnifoLM-WMA 训练、推理和部署代码已开源发布将 Unitree 从运动演示推进到 world-model 实验GitHub world-model 仓库
2025应用模板和 Isaac Lab 工作流已开源发布改善开发者打包和仿真优先迭代GitHub 仓库和开源页面
2025About 页面时间线显示 R1 和 A2 发布已宣布将产品线拓展到当前核心组合之外About 页面时间线
2025-2026杭州的试生产、测试和标准生态规模化支撑支撑一个判断:商业化不只靠演示,还依赖测试 / 认证基础设施China Daily;RAN 标准

里程碑混合了已发货硬件、公开演示和生态支撑,不应都解读为等价的收入里程碑。

[CE015, CE017, CE018, CE022, CE024, CE025]
FE004: 产品成熟度 / 能力图

Unitree 主要产品族在商业化、开放开发者界面、野外耐用性和证明质量上的相对成熟度图。

评分是基于公开证据集的 1–3 定性判断,不是内部经营指标。

[CE003, CE004, CE005, CE006, CE007, CE019]

5.4 安全、差异化与技术风险

Unitree 的核心差异化,是低价公开硬件、对关键机器人组件的垂直整合,以及相对开放的开发者界面的组合。这个组合很可能解释了为什么它的机器人频繁出现在研究实验室、公开展示和成本敏感买方讨论中。但同一组公开材料也凸显了实质风险。安全免责声明不是埋在条款里的样板脚注;它们直接嵌在 G1、H1、B2、Go2 和 A1 电机页面中,提醒强力运动、全球人形机器人行业的不成熟、保持安全距离的必要性,以及禁止危险改装。性能限定同样重要:B2 的最高速度需要特殊配置,Go2 提到部分功能需要人工操作或二次开发,G1 称部分样例功能仍在开发中。Correll Lab 的外部现场证据刺破了营销演示的光滑表面:Go1 可以在崎岖地形中导航,但在真实环境中仍会打滑、缠住,并需要手动重启。最严厉的反向信号来自网络安全:一篇 2025 年研究论文指称 G1 存在 BLE 配网和加密弱点。即便该论文中的部分主张仍需独立复现,它也足以支撑一个结论:Unitree 的技术优势并不会消除严肃尽调需求。在承销广泛企业部署前,仍必须核查安全、环境坚固性、安全认证和生产可靠性。 [CE020, CE021, CE025, CE026, CE027, CE030]

信任 / 质量 / 合规表
控制项或信号状态范围证明什么未解缺口
民用用途免责声明在公开产品页面明确出现Go2、B2、G1、H1/H1-2、A1 电机Unitree 公开提示动力和危险改装风险不等同于第三方安全认证
人形机器人谨慎表述明确G1 和 H1/H1-2 营销页面管理层承认人形机器人仍处早期未公开可靠性基准或 MTBF 披露
标准参与可见但间接2026 年中国人形机器人标准委员会报道Unitree 参与行业规则制定委员会成员身份不证明终端产品合规
开放开发者仓库SDK、ROS2、RL、仿真、应用模板开发者可以查看真实可用的软件接口并基于它开发文档可发现性和支持一致性仍参差
安全响应证据弱 / 反向公开记录包含攻击面指控网络安全仍是尽调中的活问题,不是已解决的信任能力已审阅来源中未找到公开修复公告
制造态势在改善,但未经审计About 页面和杭州试生产生态文章Unitree 正在投入生产和测试能力未公开工厂 QA 指标、召回历史或认证清单

如果没有经审计的公开认证或 QA 指标,缺口就直接列出,不推断消除。

[CE020, CE021, CE023, CE024, CE025, CE031]
FE003: 关键依赖图

Unitree 的产品执行依赖纵向整合的机器人组件、外部计算生态、合作伙伴渠道,以及不断演进的标准和安全控制。

[CE008, CE012, CE015, CE016, CE017, CE024]

5.5 图表与要点

Chapter 06

06客户情况

6.1 分层地图与采购路径

理解 Unitree 的客户宇宙,最好把它看作一组公开证据簇,而不是一份披露清晰的收入分层。最强的证据簇是研究和开发者用户、教育相关买方、工业巡检前景客户、公共活动或媒体组织方,以及服务大学、公共安全和企业实验的西方分销商或集成商。研究和教育尤其突出,因为实验室和渠道伙伴反复点名这些场景,而且相对西方替代品,Unitree 公开低价显著降低了实验室采购门槛。工业巡检经常出现在公司和分销商营销中,尤其围绕电力巡检和全地形移动,但独立具名运营方很少。公开活动证据很多——春晚、冬奥会、Super Bowl、亚运会、CES——但这些证据更多说明认知度和技术展示,而不是经常性客户经济性。采购路径也会按买方成熟度分裂:Go2 和部分 G1 配置可以直接在线购买,或至少可以直接查看价格;更重载或更复杂的部署,则依赖授权分销商、培训、保修和集成支持。这种分裂塑造了几乎所有西方证据点。 [CU001, CU004, CU005, CU020, CU021, CU022]

客户分层表
细分客群买方 / 用户 / 付款方使用场景具名证据质量战略价值缺口
研究实验室实验室 PI / 机器人团队 / 大学预算运动控制、自主、建图、操作研究真实使用和技术迭代的最佳证据未公开支出或续约指标
教育和开发者买家教授、学生实验室、开发者、小型机构教学、原型开发、具身 AI 实验低零售价扩大漏斗顶部采用公开结果多为轶事或分销商主导
工业巡检潜在客户企业运营方或集成商电力巡检、测绘、危险区域穿越低-中可能是价值最高的真实工作流具名运营方和 ROI 证据稀少
公共安全 / 政府警方、国防、公共机构侦察、进入危险场景、研究显示买家愿意基于成本 / 性能采购安全和政策审查可能阻断采购
公共活动 / 媒体主办方广播电视机构、活动制作方、赞助商晚会表演、体育赛事、CES 演示带来全球认知和阶段性变现曝光不等于经常性客户收入
西方分销商 / 集成商渠道合作伙伴销售、导入、保修、集成、培训面向北美和企业买家的关键出口路径渠道依赖掩盖终端客户集中度

细分客群证据质量反映本章审阅的公开来源集,不代表 Unitree 内部收入结构。

[CU001, CU003, CU004, CU020, CU030, CU033]
出口与渠道路径表
路径典型买方公开来源显示的内容商业含义缺口
直营网店开发者或小型实验室Go2 和 G1 的定价、运输和清关摩擦公开可见实验和小额订单的低摩擦入口没有购物车到激活的转化数据
官方 / 授权经销商高校、企业、政府、公共安全经销商把保修、培训、集成和支持作为核心价值来卖说明复杂买方需要硬件之外的服务经销商经济性和排他权未知
人工跟进销售路径工业巡检或高级人形机器人买方B2、EDU 和部分人形机器人配置会把买方导向销售专家较高 ASP 销售很可能依赖顾问式采购未披露销售周期时长或赢单率
受出口限制的连接能力欧洲和部分亚洲现场用户Go2 页面限制了部分 4G 支持地区本地化和运营商支持会影响部署质量没有完整出口市场支持矩阵
政府 / 联邦资助渠道陆军实验室、公共机构Futurology 明确宣传联邦承包商能力即便缺少直接运营落点,也可能撬动美国机构需求采购审批可能迅速收紧

商业路径很重要,因为 Unitree 的客户证据离不开中国以外的支持、合规和集成交付方式。

[CU014, CU015, CU016, CU017, CU020, CU021]
FU001: 客户旅程图

从认知到活跃使用的典型旅程显示,今天有两条主路径:低摩擦的开发者直接购买,以及渠道辅助的企业或机构采购。

旅程综合自官方商店流程、分销商话术和具名公开采购案例,并非来自某个已披露的公司销售手册。

[CU020, CU021, CU026, CU033, CU034, CU040]

6.2 按分层列出的具名证据

具名公开证据在学术和研究使用中最深。RoMI Lab 称,一台 Go1 自 2022 年春季起就在其实验室,是敏捷运动研究的关键平台;RTU MIREA 的 Zenodo 支持论文记录了一次具体的 Go1 Air 与 ArduPilot 集成,并报告了实测延迟;Correll Lab 的现场测试提供了异常有价值的负面证据,展示 Unitree 四足机器人在哪里会打滑、缠住并需要人工恢复。西方公共安全和政府证据更薄,但仍有意义:The Wire China 报道,Central Connecticut State University 运营一台 Go2,Topeka Police Department 因为西方替代品太贵而购买了三只 Unitree 机器狗,美国陆军通过分销商为四只 Unitree 机器狗预留了超过 $60,000。分销商页面强化了西方机会由渠道主导这一点:TVC、Drones Plus Robotics、Futurology、RoboStore 和 US Robot Systems 都把 Unitree 营销到研究、教育、工业巡检、政府或安防用例,并明确使用支持和集成话术。公开活动提供了最宽的可见度证明——尤其是 2025 年春晚和多次全球活动亮相——但这些应归类为展示证据,而不是生产客户背书。 [CU006, CU007, CU008, CU010, CU011, CU012]

客户增长 / 采用轨迹表
指标或证据点数值日期来源质量含义缺失分母
RoMI Lab 使用 Go1机器人自 2022 年春季起就在实验室使用2022-2026 年公开持续存在公开来源中最强的直接重复使用证据实验室支出和机队规模未知
RTU MIREA Go1 论文集成 ArduPilot,平均延迟 206.7 ms2023-08-31科研买家在积极改造 Unitree 硬件并发表成果缺少类似实验室数量数据
Correll Lab 野外部署Go1 崎岖地形测试,并记录失效模式2024-07-05独立野外使用验证真实运行,也暴露真实限制单一实验室结果,不是广泛装机基数研究
春晚16 台 H1 人形机器人参与 Yang Bot 节目演出2025-01-28大众曝光,并证明可完成协同舞台部署未披露经济性或续约条款
CES 2025 多产品展示Go2、Go2-W、B2-W、H1 和 G1 在 Las Vegas 展示2025-01-08释放出积极获取全球买家的信号未披露展位客流和转化
美国警察采购Topeka Police 购买三台 Unitree 机器狗2025 年对 2024-2025 年采购的报道公共安全买家会基于成本 / 性能采购尚不清楚采购是否扩展为更大机队
美国陆军采购自 2023 年以来,为四台 Unitree 机器狗划拨 >$60,0002023-2025政府科研买家愿意通过分销商采购没有续购或更广泛正式项目证据
海外销售占比Wang 称年度销售额一半来自海外2025 年文章引用的 2025 年访谈出口需求重要,而非边缘未披露按国家拆分或细分客群结构

这张表混合了直接部署证据和采用相邻的商业信号,因为 Unitree 不发布标准化客户 KPI。

[CU006, CU007, CU008, CU011, CU012, CU013]
具名客户证据表
客户 / 群组细分客群部署 / 使用场景正式使用 vs 试点 vs 曝光结果 / 证据限制
RoMI Lab研究实验室Go1 用于足式与敏捷运动研究正式科研使用实验室页面称,机器人自 2022 年春季起投入使用,是关键平台未披露支出、续约或论文数量
RTU MIREA科研 / 教育Go1 Air 任务控制和 GNSS 路线跟随正式科研使用Zenodo 论文记录集成路径和 206.7 ms 平均延迟单篇论文,不是机队基准
Correll Lab研究实验室崎岖地形野外部署正式科研使用独立实验室同时记录地形能力和失效模式反向证据凸显越野稳健性限制
Central Connecticut State University大学 / 教育美国文章展示 Go2 操作正式使用证据具名运营方照片和文章证据显示真实校园使用数量和合同条款未知
Topeka Police Department公共安全机器狗先于警员进入危险场景正式部署部门发言人被引用,说明采购数量和理由缺少长期使用率或续购数据
美国陆军科研买家政府科研为 ARL 研究通过分销商采购 Unitree 机器狗正式科研采购文章引用四台机器人的豁免和支出记录仍是科研采购,不是广泛运营部署
2025 年春晚公共活动 / 媒体16 台 H1 人形机器人在 Yang Bot 表演中跳舞仅曝光 / 展示多个来源确认规模和所用平台不能证明稳定的付费客户关系
亚运会场地作业公共活动 / 体育物流机器狗在赛场搬运铁饼和标枪仅曝光 / 展示Go2 页面和 About 页面展示了具体活动任务未披露运营经济性和合同安排
西方分销商群组渠道 / 经销商TVC、Drones Plus、Futurology、RoboStore、USRS 销售 Unitree,并提供支持和保修正式销售渠道显示可重复触达西方科研、工业和政府买家的路径终端客户赢单率和收入集中度未知

各行有意混合真实客户部署和仅用于曝光的活动证据,因为本章目标就是把二者明确区分。

[CU003, CU006, CU007, CU008, CU010, CU011]
FU002: 采用 / 部署流向

具名证据通常从兴趣表达和渠道接触,走向实验室或机构部署;只提供曝光的活动亮相则并行留在认知路径上。

流程把付费使用证据与活动曝光证据拆开,因为公开来源常把这两类混在一起。

[CU003, CU014, CU020, CU025, CU028, CU029]

6.3 耐久性、集中度与证据质量缺口

本章的核心尽调结论是:Unitree 有真实采用,但公开证据质量不均,留存证据偏弱。研究使用是置信度最高的分层,因为它包括直接运营方页面和论文,其中一些跨越多年仍持续存在。西方分销和采购证据提供方向性支持,因为它显示了保修、集成、培训以及政府或大学销售路径,但它仍离终端客户运营结果隔了一层。公开活动和 CES 亮相具有战略价值,因为它们制造认知,并且如 Xinhua 所说,带来阶段性商业价值;但它们不能证明买方会续约、扩张或大规模部署车队。工业巡检处在两端之间:Unitree 显然在营销该场景,分销商也积极卖进去,但具名第三方运营方和量化结果仍少。集中度风险很可能偏向研究、演示和渠道辅助出口市场,只是因为这些分层拥有最密集的公开证据。同时,西方采用还面临来自安全担忧、政治审查、分销商依赖,以及缺乏公开 NRR、GRR、头部客户或队列数据的非小摩擦。实际读法是,Unitree 已经有足够客户证据支持更深入尽调,但披露的客户经济性还不足以证明企业、工业和政府队列中的可重复性。投资者可以承销需求生成和低摩擦进入的证据,但还不能承销一个证据充分的经常性客户基础。 [CU003, CU009, CU020, CU024, CU028, CU030]

留存 / 重复使用 / 满意度表
指标数值 / 证据细分客群置信度尽调追问
净收入留存率未公开披露所有付费客户N/A要求按细分客群和地区提供 NRR
总收入留存率未公开披露所有付费客户N/A要求提供流失和降级数据
装机客户数未公开披露所有细分客群N/A要求按产品系列提供活跃账户数
重复实验室使用RoMI Lab 自 2022 年春季起公开使用研究实验室索取使用超过 12 个月的实验室数量
公开活动重复曝光2021 至 2025 年多次亮相标志性活动展示场景索取同一买方是否重复采购活动、还是一次性展示的证据
渠道支持连续性经销商页面强调上线、培训和持续支持西方渠道索取经销商续约、附加销售和服务收入数据
客户满意度未发现可比 B2B SaaS 评论网站的大规模独立评价语料所有细分索取 NPS、工单解决率和客户背调访谈包
采购粘性Topeka 购买了 3 台机器人;陆军自 2023 年以来购买了 4 台公共安全 / 政府研究索取复购记录和维护合同

该表有意以 null 占位符和代理指标为主,因为公开耐久性数据本就是本章的主要尽调缺口之一。

[CU009, CU020, CU034, CU035, CU040]
扩张与集中风险表
扩张驱动或风险类型影响公开证据尽调路径
相较西方同行,零售定价更低扩张驱动Wire China 和官方商店页面都明确显示价格优势索取分渠道利润率和 CAC
授权经销商网络扩张驱动TVC、Drones Plus、Futurology、RoboStore 和 USRS 都主打重支持销售索取经销商收入分成和区域权利
从研究向产业迁移扩张驱动同一平台出现在实验室、教育和巡检营销中索取实验室买方向企业集成商转化的数据
曝光驱动的需求生成扩张驱动Gala、Super Bowl、Winter Games、Asian Games 和 CES 带来全球认知索取与标志性活动挂钩的线索生成数据
证据集中在研究 / 展示场景集中风险这些场景之外,具名公开证据很少索取按细分市场和产品家族拆分的活跃客户结构
美国安全与政治审查集中风险Wire China 记录了国会和采购层面的担忧审查出口管制、采购和数据治理风险敞口
北美渠道依赖集中风险中高Wire China 称 Unitree 在美国缺少直接运营落点索取直营与渠道收入、服务指标
留存不透明集中风险没有公开 NRR、GRR 或队列数据在认可其规模化主张前,将队列数据设为尽调门槛

这张风险表只看公开证据能支撑什么、不能支撑什么,不尝试估算隐藏的收入结构。

[CU014, CU020, CU024, CU028, CU037, CU038]
FU003: 客户证据矩阵

科研实验室的公开证据质量最强,公共活动的经常性价值证据最弱;西方经销商居中,因为它们能证明渠道准备度,却不能证明终端客户的持续性。

分数采用 1-3 级,3 代表公开证据强,1 代表公开证据弱。

[CU003, CU009, CU020, CU028, CU030, CU036]

6.4 图表与要点

Chapter 07

07风险

7.1 地缘政治、采购禁令与监管风险

风险栈从 Unitree 的出身开始。美国议员已经不再只谈无人机和电信设备;American Security Robotics Act 明确针对来自受关注外国实体的地面机器人,IEEE 把该提案界定为两党脱钩议程的一部分。即便法案永远不会成为法律,它的话术也很重要,因为它给受监管和政府邻近买方一个可辩护理由,让他们可以先发制人地避开中国机器人。出口管制风险会叠加放大这个问题。美国政府来源显示,对中国半导体限制已经很宽、还在扩张,并且行政执行复杂。公开披露并未显示 Unitree 是制裁目标,但它仍依赖一个生活在政策冲击半径内的硬件生态。中国国内标准和杭州政策支持确实有助于近期规模化和安全治理,但它们无法解决西方买方眼中“中国来源本身”带来的外交政策反对。[CR001, CR002, CR003, CR004, CR005, CR006]

监管 / 法律风险登记表
风险或义务司法辖区状态可能性严重性缓释措施剩余风险敞口尽调路径
American Security Robotics Act 针对受关注外国实体地面机器人的采购和使用禁令美国联邦2026 年 3 月由两党议员提出严重除了避开受影响细分市场、证明用途不敏感,几乎没有真正缓释手段法案通过前,涉政府和受监管买方就可能自我排除跟踪法案进展、众议院 / 参议院采纳情况和企业客户反应
出口管制外溢到先进芯片和机器人供应美国 / 中国跨境EAR 规则已扩大,合规和执法仍在推进若已有供应商多元化和芯片替代方案,可能有所帮助成本、性能或交付可能在没有公开预警时恶化索取具名芯片供应商、EAR 分类和替代采购路径
人形机器人伤害或故障的产品责任与问责边界不清美国 / 欧盟 / 主要运营市场具身 AI 的法律框架仍不成熟合同和保险可转移部分风险一次严重事故就可能引发远超事件本身的法律和声誉损害审查保险、赔偿条款、安全手册和事故响应预案
音频、视频和传感器数据的隐私 / 生物识别数据合规欧盟 / 英国 / 其他隐私监管较重市场只要机器人数据跨境流动,风险就实时存在数据最小化、本地化和更充分披露可降低风险遥测或云架构不透明,可能直接挡住受监管客户梳理每个目标地区的数据流、留存和托管选项
中国标准与西方符合性预期分化中国及出口市场2026 年发布新标准国内标准化应能改善基线测试和文档标准分化会带来重复认证工作和买方困惑确认认证、公告机构计划,以及与西方合规要求的对应关系
政策集中于中国地方支持和机器人产业集群优先事项中国国内目前支持力度大,但由政策驱动杭州集群支持降低近期执行摩擦一旦优先事项转向,Unitree 会失去部分成本和商业化优势按省份核验补贴、设施和政策依赖

各行按当前严重性排序,聚焦会实质影响可服务市场或法律敞口的公开义务或政策动向。未发现针对 Unitree 的活跃公开诉讼,因此本登记表转而强调监管和潜在责任风险。

[CR001, CR003, CR004, CR006, CR007, CR008]
FR001: 风险热力图

Unitree 主要风险在考虑已披露缓释措施后的发生概率—严重程度位置。

[CR001, CR004, CR009, CR012, CR018, CR032]

7.2 安全、网络安全、质量与产品责任风险

Unitree 的公开产品页面和独立安全文献指向同一方向:公司仍足够早期,技术进步与真实运营不确定性并存。G1 和 H1 页面告诉用户保持安全距离、避免危险使用,并理解早期人形机器人的局限;G1 基础产品也只公布了不高的手臂载荷和较短保修条款。仅这一点,就足以支持在工业场景中保持谨慎。arXiv 安全评估走得更远,指称存在具体 BLE 攻击路径和反复出现的遥测行为;在受监管或隐私敏感环境中,这可能变得不可接受。Hill Dickinson 补上法律框架:人形机器人一旦离开实验室,制造商、运营方和软件栈之间责任不清,就会成为现实产品责任问题。Unitree 的 OTA 节奏和开发者速度是有用缓释项,但不能替代经审计安全保证、长期支持承诺或披露的现场可靠性数据。[CR007, CR008, CR009, CR010, CR011, CR012]

运营 / 质量 / 安全风险登记表
失效模式可能性严重性缓释成熟度剩余风险敞口未解决缺口
与 G1 / Unitree 遥测相关的漏洞利用或数据外泄事件严重低至中 —— OTA 更新存在,但未披露公开第三方安全保证材料安全事件可能冻结面向隐私敏感行业的销售,并放大政策审查未发现独立修复验证、安全 SDLC 披露或企业安全认证
早期人形机器人部署中的物理安全事故低至中严重中 —— 公司已有安全警示,但法律问责仍未厘清严重伤害会引发诉讼、客户暂停和监管审查保险、事故预案和安全认证披露不足
保修 / 支持窗口短,已部署机器人实际缺少支持低 —— 已公布的保修窗口很短没有长期支持承诺的平台,企业买方可能回避,或大幅压低估值未披露公开长期支持计划或全球零部件 SLA
操作和载荷限制压缩真实生产任务空间中 —— 可选灵巧手和高端型号有所帮助,但基础版 G1 仍受限前期成本低,也未必能转化为真实劳动力替代没有独立基准测试覆盖真实工业流程中的任务成功率
规模扩张缺少已披露的 MTBF、退货率或质量指标中 —— 出货量和 OTA 节奏暗示存在学习闭环,但硬质量数据缺失现场质量意外可能同时打击利润率、保修准备金和品牌信任需要按型号和队列拆分的 MTBF、退货率和服务工单数据
芯片限制或组件瓶颈引发供应中断或重新设计低 —— 未披露公开冗余计划关键增长窗口内,成本、性能或交付可能恶化具名供应商和替代路径未公开
开放开发者栈加速仿制或 IP 外泄中 —— 开放确实带来社区价值和更快迭代如果生态价值不能转化为经常性软件收入或服务控制,护城河会被压缩没有关于付费软件、云附加销售或开发者变现的公开数据

本表合并网络、产品、质量和支持风险,因为公开证据显示这些风险会相互叠加:一个支持窗口短、安全状态仍偏早期的平台,更容易受到安全和可靠性冲击。

[CR009, CR010, CR011, CR012, CR013, CR015]
FR002: 风险传导图

Unitree 在政策、网络安全、质量和客户证据上的上游风险,如何传导到收入、可触达市场和估值结果。

[CR001, CR004, CR009, CR012, CR024, CR029]

7.3 供应链、分销与依赖风险

Unitree 极大受益于以中国为中心的供应网络,而同一张网络也制造了依赖风险。第三方报道把公司的价格优势归因于中国组件生态;如果出口管制规则或国内瓶颈收紧,这个优势会很快被压缩。公司的海外销售 路线也在结构上依赖分销商和早期采用买方,而不是直接的西方运营足迹。Wire China 显示,当前美国需求真实存在,但集中在大学、警察部门和陆军研究,而不是大型经常性工业合同。这让 Unitree 暴露于渠道脆弱性:在公司建立独立本地支持基础之前,分销商、客户或监管机构都可能打断动能。Agility 和 Boston Dynamics 等同行在这里不只是竞争者,也是证明软件集成、现场服务和具名工业用户本身就是风险缓释项的参照。Unitree 尚未披露同等运营缓冲。[CR014, CR016, CR022, CR023, CR024, CR025]

合作伙伴 / 依赖风险登记表
依赖项交易对手 / 系统角色集中度失效情景严重性缓释措施剩余风险敞口
海外销售渠道RobotLab 等持牌经销商在美国进口、测试并转售机器人经销商退出、政策冲击或支持不佳会拖累转化和声誉拓宽渠道基础,并建设区域直营支持能力经销商主导销售在结构上仍弱于直接运营落点
中国组件生态电机、芯片和其他本地供应商压低价格、拉高速度出口管制或供应冲击会推高成本、拖慢出货关键组件多源化,并披露应急计划缺少公开替代采购证据
研究 / 公共安全买方结构高校、警方、陆军研究买方当前可见的海外需求池这些用户未能转化为经常性工业生产合同将试点和开发者关系转化为具名工业客户公开记录中的商业证据仍然薄弱
政策生态支持杭州 / 中国产业政策集群效应、试点、基础设施、关注度政策优先事项变化,或在海外变成政治成本分散地域,并降低对类补贴优势的依赖支持有用,但政治上是双刃剑
开放开发者生态SDK、ROS2、RL、集成商驱动早期采用的漏斗入口开发者在 Unitree 上做原型,随后迁移到其他机器人打造付费服务、云端挂钩和生态留存闭环当前公开证据更能证明开放,而非留存
同行信任基准Agility / Boston / Figure / 1X严肃企业买方的备选供应商即便赢下研究需求,Unitree 也会失去最高价值机会补齐支持、安全和客户证据缺口的速度,要快过同行拉大差距同行优势今天已公开;Unitree 的回应不够可见

合作伙伴和依赖风险异常重要,因为 Unitree 当前护城河依赖的是生态——供应商、经销商、地方政策支持和开发者——而不是封闭工作流垄断。

[CR014, CR016, CR022, CR023, CR024, CR025]
FR003: 依赖关系图

关键外部依赖决定 Unitree 在供应链、渠道、政策和对标同业上的风险画像。

[CR016, CR020, CR023, CR024, CR025, CR026]

7.4 执行、客户证据、IP / 护城河与 IPO 时点风险

执行风险不只是多造机器人。IPO 审视升温之前,Unitree 还得把开发者驱动、价格驱动的增长,转成企业客户撑得住的业务。开放 SDK 和仿真生态是很强的采用引擎,但也削弱专有锁定,让护城河看起来更像速度和成本,而不是受保护的工作流控制权。如果客户证据来得够快,这可以接受;如果同业继续抓住信任层,问题就会放大。公开报道显示,Unitree 确有收入规模和私募市场热度,但若公开市场要把 IPO 估值抬到更高水平,投资人就必须相信客户留存、合规和质量能持续——这些领域公开记录仍然很薄。高速增长、披露不完整、地缘政治阴影叠在一起,估值和时点风险就不是旁枝;它是投资逻辑最直接的断裂点之一。[CR015, CR027, CR028, CR029, CR030, CR032]

人员 / 执行风险登记表
角色 / 职能依赖或缺口可能性严重性缓释措施尽调路径
创始人 / 高级管理层快速执行现在必须配上合规和支持纪律在创始人层之下补上可见的安全、法律和政府事务负责人索取组织架构图,以及合规、安全和客户成功的高管责任归属
客户成功和现场支持开发者优先的姿态可能低配企业级 SLA 运营大规模工业部署前,先建设区域支持团队和零部件仓按地域询问支持人员配置、响应时间和备件策略
安全工程公开漏洞利用叙事跑在已披露修复叙事前面发布安全响应政策、补丁节奏和外部审计索取 CVE 处理流程、红队工作和安全更新控制
法律 / 合规建设IPO 路径提高披露和治理负担上市前,将政策、隐私、出口管制和产品安全团队制度化要求公司提供 IPO 就绪治理计划,以及外部律师 / 审计准备情况
IP / 生态策略如果软件商业化跟不上,开放工具会削弱留存在开放工具之上叠加付费云、工作流或企业集成要求披露开发者商业化、附加率和软件收入

这里的人员风险,与其说是某位高管离职,不如说是公司能否赶在估值预期见顶之前,从高速硬件创新者成熟为服务撑得住、政策风险也看得懂的全球供应商。

[CR014, CR015, CR032, CR038]

7.5 缓释因素、剩余暴露与投资逻辑断裂触发点

Unitree 的缓释逻辑很直接:中国正主动支持人形机器人标准化和机器人产业集群,Unitree 有真实销量和低定价,迭代快,开发者开放策略也让实验覆盖面很广。这些都是有意义的缓冲。但缓释之后,剩余暴露仍集中在四处。第一,政策:中国背景是美国采购禁令言论里无法修复的根因。第二,网络安全 / 隐私:一次可信的漏洞或数据流发现,就可能立刻让买家犹豫。第三,客户证据:在 Unitree 展示具名、持续使用的工业客户之前,西方买家仍可把平台视为强大的研究工具,而非规模化企业系统。第四,估值时点:IPO 重估或推迟会立刻检验规模叙事是否转成耐久经济性。这些才是最重要的监控点。投资人可以承受硬件风险;政策关门、安全冲击和商业验证缺席同时出现,就难承受得多。[CR034, CR035, CR036, CR037, CR038, CR039]

缓释措施与否决标准表
风险可监测触发因素阈值 / 事件行动含义
采购禁令 / 地缘政治封闭ASRA 或类似限制法案通过、适用范围扩大,或受监管买家大规模主动排除除非海外逻辑不再是核心,否则转为回避
出口管制外溢新芯片限制或披露的供应中断交付延迟、规格下调,或突然需要重新设计重新测算成本曲线和交付可信度
网络安全 / 隐私冲击独立漏洞、确认的遥测滥用,或监管调查具名客户暂停、召回,或公开事件在补救获得独立验证前,按投资逻辑击穿处理
质量 / 支持失败保修索赔、故障或支持积压显著上升现场保有量扩大,但没有公开扩充支持能力下调企业转化和利润率假设
客户验证缺口具名工业客户持续缺位IPO 进程推进前仍无披露的经常性生产部署将故事重估为开发者 / 研究硬件,而不是企业机器人平台
IPO 时点 / 估值风险IPO 延期、重新定价,或治理短板政策审查升温时,CSRC / 交易所流程滑坡披露改善前,不按峰值估值承销
护城河侵蚀同行缩小价差,或以相近成本在服务上超过 UnitreeAgibot / 1X / 其他公司公布明显更低价格或更好支持重新评估 Unitree 是否仍配得上成本和产品广度溢价

这张表把宽泛风险转成明确的监测触发因素。对这家公司而言,否决标准大多是外部事件或证据点缺失,而不是季度间产品功能漏项。

[CR028, CR029, CR034, CR035, CR036, CR037]

7.6 图表

Chapter 08

08估值

8.1 建议、投资逻辑与反向逻辑

以当前约 US$1.6B–US$1.7B 的隐含估值看,Unitree 值得观察。投资逻辑成立:在人形和四足机器人公司里,很少有公司能同时拿出已发布的产品价格、直销渠道、真实报道收入、覆盖四足和人形的广产品组合,以及可见的海外需求。Unitree 不是零收入概念股。它的低价格和出货节奏,给了它许多估值更高的同业仍缺的商业化轮廓。 反向逻辑同样重要。目前公开证据更多说明 Unitree 的销量和能见度,而非经济性。毛利率、出货组合、保修负担、经常性软件或服务收入都未披露。IDC 支持的报道还显示,2025 年可测量的人形机器人市场只有约 US$440M,需求由表演、教育、数据采集和接待主导,并非规模化工业 ARR。这意味着投资人仍在押注:可负担硬件销量会转成更耐久的工业经济性。安全审查和中国注册地折价,也进一步解释了为什么 Unitree 的估值远低于 Figure 或 Apptronik。因此,公开证据支持的是对价格敏感的观察姿态,而不是按当前估值确信买入。[CV001, CV005, CV006, CV007, CV017, CV019]

建议摘要表
参数评估证据支持的说明
总体建议观察商业化和收入底盘真实存在,但经济性披露太薄,还不足以支撑坚定买入
置信度公开证据方向上积极,但利润率、股权结构表和经常性收入仍不完整
风险评级硬件收入结构、中国折价和安全审查都可能压制上行兑现
估值立场合理至偏高低于 Figure 或 Apptronik,但对照 Unitree 自身公开证据,并不明显便宜
支持入场的因素真实产品、可见价格、已发货量、海外需求Unitree 已不是还没产品的故事
阻碍买入的因素没有毛利率、分部结构或优先股堆叠披露当前价格要求投资者相信尚未验证的收入质量
上调条件更好的经济性披露或更好的价格工业经常性需求证明加上下行保护,会改变判断

这张表只总结本章判断,而且有意保持价格敏感:更好的价格或更充分的披露,即使不改变公司的技术质量,也可能改善观点。

[CV019, CV020, CV027, CV042, CV045]
投资逻辑 / 反向逻辑表
论点重要性反向逻辑 / 削弱因素改变观点的条件
真实的产品市场验证Unitree 已在四足机器人和人形机器人上公开产品页、价格和出货证据单纯销量仍可能只是低毛利硬件收入披露毛利率和重复订单率
中国成本和政策优势密集供应链和标准支持加快规模化海外买家仍可能套用结构性中国折价证明海外商业需求在审查下仍然稳定
相对最贵同行的估值折价当前估值远低于 Figure,也低于 Apptronik也可能是同行估值过高,而不是 Unitree 被低估证明 Unitree 能把销量转成优于同行预期的经济性
双平台敞口四足机器人让业务不只押注人形机器人热潮收入结构复杂会让可比对象更难选,也可能掩盖低质量收入按产品家族和终端市场披露收入拆分
IPO 可选性IPO 准备带来真实的重估上行空间US$7B 目标不是已完成轮次的事实发布更正式的 IPO 准备披露,或更新私募市场条款

投资逻辑有意与直接的反向逻辑并列。只有当「改变观点的条件」一栏开始获得真实证据,而不只是叙事支撑时,建议才会改善。

[CV002, CV017, CV018, CV019, CV020, CV029]
FV001: 建议逻辑

流程图把当前证据支柱和风险向量,推导到最终「观察」建议,以及上调所需条件。

[CV019, CV020, CV021, CV022, CV042, CV045]

8.2 融资背景与可比公司组

以私募市场标准看,Unitree 当前估值锚相对扎实:TechNode 和 The Robot Report 都把公司放在 2025 年 Series C 背景后的约 US$1.6B–US$1.7B,Humanoid Index 则更保守,给出 US$1.4B+,同时也收录了更带宣传色彩的 US$7B IPO 目标。这种冲突不是致命问题,反而有用。它提醒投资人,已交割轮次证据和前瞻叙事证据不能混成一个估值锚。更保守的定价方式,是用已交割轮次估值约束进入价格,把 IPO 数字只当乐观情景叙事。 相对同业,Figure 约 US$39B 是行业异常值,不是所有人形机器人公司的出清价格。Apptronik 约 US$5B 明显高于 Unitree,但背后有近 US$1B 融资和高度战略化的投资人基础。Agility 和 1X 提供更接地气的中档参照,Agibot 和 UBTech 对中国语境有用,但在本次抓取资料里数字锚更弱。关键对比点不是 Unitree 「因为更低所以便宜」。而是 Unitree 今天的产品证据强于许多同业,同时在毛利率和细分组合披露质量上明显更低,还叠加清晰的中国地缘政治折价。这个组合让当前估值可以理解,但看不出明显向投资人一侧错定价。[CV001, CV003, CV004, CV009, CV010, CV011]

可比估值表
可比对象当前可支撑估值或状态已抓取证据能支撑什么可比理由主要局限
UnitreeUS$1.6B–US$1.7B 已完成轮次锚点;US$7B IPO 目标TechNode 和 The Robot Report 支持当前锚点;Humanoid Index 给出较低的四舍五入估值和 IPO 野心主要标的估值锚点利润率、结构和优先股堆叠未披露
Figure AI约 US$39BTech Market Briefs 和 Humanoid Index 支持极高的后期私募估值展示该行业最昂贵的叙事倍数异常案例,公开经济性披露有限
Apptronik约 US$5BApptronik 官方信息和 Reuters-linked US News 支持 2026 年 A-X 轮估值,以及接近 US$1B 的累计融资Figure 之下最接近的美国高估值私募可比公司战略投资者占比重;经济性披露仍然很轻
Agility RoboticsHumanoid Index 摘要高于 US$2.1B+官方定位加 Humanoid Index 支持其商业物流重点和近期大额融资更贴近落地的商业仓储可比对象抓取集里没有完整披露精确投后估值和经济性
1X TechnologiesHumanoid Index 摘要高于 US$500M+官方 NEO 定位加 Humanoid Index 支持家用机器人逻辑和较低估值层级可作为消费者 / 家庭叙事的低端同行地域不同,且更偏消费者
Agibot / UBTech定性中国背景;抓取文本里没有干净的可比估值AGIBOT 证实中国同行野心;UBTech 证实公开市场参照,但报价页在文本模式下无法机器读取提供有用的中国语境和公开市场透明度基准当前抓取集的数字支撑较弱

这组可比公司有意保留不均衡,因为不同同行的可支撑度不同。Figure 和 Apptronik 的公开估值线索比 Agibot 或 UBTech 更干净;表格保留这种不对称,而不是强行制造虚假精度。

[CV001, CV003, CV009, CV011, CV012, CV013]
FV004: 投资 KPI

投委会风格 KPI 看板汇总 Unitree 在市场、证据、经济性可见度、风险和估值上的评分。

[CV017, CV018, CV019, CV020, CV027, CV045]

8.3 含明确估值区间的乐观、基准和悲观情景

情景分析应锚定在公开收入下限上,而不是英雄式 TAM 捕获假设。合理乐观情景是假设 Unitree 把产品驱动的出货量转成工业和商业重复需求,同时保持海外渠道开放。在这种情况下,2028 年收入可能达到 US$500M–US$800M,公司也许配得上约 6x–8x EV/revenue 区间,对应估值约 US$3.0B–US$6.4B。该情景要求的不只是出货增长,还需要可信的需求质量改善,以及一些证据表明软件、服务或集成附加率在上升。 基准情景更克制,也可能更接近当前公开证据。Unitree 继续增长,但组合仍以开发者、研究、表演和早期商业用途为主,而不是高质量工业 ARR。这支撑约 US$300M–US$450M 的 2028 年收入和约 4x–6x 倍数,对应约 US$1.2B–US$2.7B。悲观情景假设地缘政治压缩、ASP 压力和低毛利硬件组合,把 2028 年收入压在约 US$180M–US$250M,并把倍数推向 3x–4x,估值约 US$0.6B–US$1.0B。在当前约 US$1.6B–US$1.7B 的价格上,基准情景回报只是中等,悲观情景会亏本金;因此,这里价格纪律比行业兴奋更重要。[CV023, CV024, CV025, CV026, CV027, CV028]

乐观 / 基准 / 悲观情景表
情景核心假设2028 年收入估值区间从当前入场价看概率信号
乐观工业需求放量,海外渠道保持开放,软件 / 服务附加改善,IPO 路径保持可信US$500M–US$800MUS$3.0B–US$6.4B上行空间有吸引力,但经济性质量必须追上出货故事需要看到工业重复订单和利润率进展
基准销量继续增长,但结构仍以开发者、研究、表演和早期商业用途为主US$300M–US$450MUS$1.2B–US$2.7B相对当前估值只有中等上行最符合今天可测得的市场结构
悲观安全审查升温,ASP 受压,Unitree 仍主要是低毛利硬件商US$180M–US$250MUS$0.6B–US$1.0B当前入场会产生资本损失由出口 / 采购限制或工业转化疲弱触发

收入区间是分析师估计,锚定公开收入底盘和人形机器人市场的可测现状。倍数不是上市公司可比估值,而是为一家资本密集、经济性披露不完整的机器人公司设定的情景带。

[CV023, CV024, CV025, CV026, CV027]
FV002: 估值敏感性

柱状图展示不同 2028 年收入与倍数组合如何映射到企业价值,并与当前入场锚点对比。

[CV023, CV024, CV025, CV026, CV027]
FV003: 估值 / 回报区间

区间图保留当前入场价格下的下行与上行分散度,而不是用单一点估计把差异抹平。

回报按总额口径计算,且不计任何未来稀释,因为公开资料没有优先权或稀释细节。图表意在展示分布,而非精确值。

[CV026, CV027, CV028, CV045]

8.4 为什么公开证据还不足以支撑全额建仓

公开证据让 Unitree 向可投资走了一大步,但还没走完。它确认公司真实存在、正在出货、有价格体系,商业化渠道也比大多数同业更广。它还支持一个强有力的相对论点:相对于最膨胀的私募可比公司,Figure 看起来贵得多,Apptronik 的定价也显然押注了一个大于当前公开证据所能证明的未来。但关键缺口是经济性质量。抓取到的来源没有一个披露 Unitree 的毛利率、产品族组合、服务附加、保修负担或清算优先顺位。投资人因此知道公司有需求,却不知道这些需求会复利成有吸引力的企业价值,还是只会变成难服务的硬件收入。 这就是本章停在观察的核心原因。问题不是公开证据为负,而是当前价格已经要求投资人相信尚未公开证明的收入质量。安全审查还带来一层独立且真实的倍数折价,西方同业受影响没这么大。因此,正确读法是合理到偏高:Unitree 不是显而易见的泡沫,但公开证据仍不足以支持按高质量工业 ARR 转型已经完成来加价买入。[CV017, CV018, CV019, CV020, CV021, CV031]

8.5 最后尽调问题、投资逻辑断裂触发点和退出路径

当前公开资料指向少数会实质改变判断的尽调项。第一是收入质量:投资人需要细分组合、毛利率和重复订单证据,判断 Unitree 是否正从低价机器人供应商进化为有防守能力的机器人平台。第二是下行结构:没有股权结构表和清算细节,即使企业价值公平,股权回报也可能没有吸引力。第三是地缘政治耐久性:美国及盟友的采购或出口限制可能迅速收窄海外 TAM,尤其是公共部门或敏感商业买家。 退出路径存在,但近期最能支撑的路径仍是中国境内 IPO 或 IPO 前重估,而不是溢价全球战略并购。IPO 准备叙事真实存在,但不应把它写成必然事件。IPO 搁置、下轮融资降价,或有证据显示报道的出货增长没有转成重复工业合同,都会打破乐观叙事。反过来,如果 Unitree 披露细分经济性、展示可信的工业经常性需求,并给投资人更好的价格或更好的下行保护,建议就会改善。在此之前,审慎姿态是密切跟踪、围绕尽调沟通,不要假设仅靠销量领先就能抹平与最贵同业的估值差。[CV002, CV022, CV027, CV028, CV038, CV039]

投资逻辑击穿与否决触发因素表
触发因素阈值 / 事件重要性行动含义
出口或采购限制显著扩大公共部门或敏感商业禁令从零散审查扩散开来会收窄海外 TAM,并压缩投资者愿意为 Unitree 支付的倍数除非仅靠国内经济性也能支撑价格,否则转为负面观点
出货增长没有转成工业重复需求到 2027 年仍没有可信的工业经常性需求证据会把 Unitree 从平台故事重估为低毛利硬件商不在当前或更高价格加仓
下调估值融资或 IPO 叙事搁置下一轮融资低于当前估值,或 IPO 准备故事停滞会削弱公开证据中可见的主要重估机制按悲观情景区间重新承销
毛利率被证明结构性疲弱尽调显示,服务负担之后产品家族毛利率低或为负会推翻「规模自然改善价值」这个逻辑把低价视为商品化,而不是护城河
重大安全事件或审计失败公开泄露、安全事故,或关键采购审计失败会强化中国折价,并推迟企业采用暂停或退出,直到补救得到证明

否决触发因素是可观察事件,不是泛泛担忧。每一项都直接攻击 Unitree 的可服务市场、估值倍数,或守住当前定价叙事的能力。

[CV039, CV040, CV041, CV042]
最终尽调要求表
主题缺失证据重要性负责人 / 尽调路径
按产品家族和终端市场拆分收入没有公开拆分四足机器人、人形机器人、部件、服务、工业、研究或表演需求决定当前增长是否在复利成耐久价值要求管理账或 IPO 准备材料
毛利率、保修和现场故障数据没有公开硬件单位经济性或服务负担区分规模护城河和低价商品陷阱要求产品家族贡献利润率和保修准备金明细
股权结构表和优先股堆叠没有公开清算、参与权或反稀释细节股权回报可能与企业价值大幅背离要求当前轮条款的法律摘要
工业经常性需求证明没有公开合同金额、续约或 ARR 式披露当前估值需要比单纯出货量更高质量的需求要求工业客户队列、重复订单和服务附加
海外敞口和政策预案海外销售在哪里、受采购限制影响多大,细节有限海外 TAM 消失速度可能快于国内 TAM按地理和客户类别映射收入
IPO 准备实质IPO 准备已公开,但正式申报内容尚未公开IPO 可选性是当前最清晰可见的重估路径要求时间表、审计就绪情况和顾问工作流细节

前四项要求最重要,因为它们直接影响当前价格的估值支撑。没有这些证据,最有支撑的姿态仍是观察,而不是买入。

[CV036, CV037, CV038, CV042]

8.6 图表

免责声明

本报告依赖截至 2026-05-19 可获得的公开来源。审阅材料未包含私营公司财务、客户合同、董事会材料、股权结构条款和安全审计;任何投资决策前,都应在一手尽调中验证。

证据索引

结论
编号陈述可信度来源
CO001 Unitree Robotics is a Hangzhou-based robotics company founded in 2016. SO002, SO004, SO016
CO002 Wang Xingxing founded Unitree after turning his XDog quadruped thesis work into a company and leaving DJI after roughly two months. SO003
CO003 Wang Xingxing is publicly described as Unitree's founder, CEO, and CTO. SO003, SO016
CO004 Official messaging frames Unitree's mission around “Technology drives world progress” and “To plant the science and technology tree of the world.” SO002
CO005 Wang told KrASIA that Unitree's commercialization thesis was to lower robot cost and target individual consumers rather than copy Boston Dynamics' government-heavy market path. SO003
CO006 Official materials say Unitree focuses on quadrupeds, humanoids, six-axis manipulators, lidar, and core robot components. SO002, SO011
CO007 Unitree says it was the first company to publicly retail high-performance quadruped robots and that its sales have led globally over multiple years. SO002, SO004
CO008 The official about page says Unitree has filed more than 200 patent applications and holds more than 180 authorized patents. SO002
CO009 Unitree's GitHub organization exposes SDK, ROS2, reinforcement-learning, URDF, and app-template repositories across multiple robot families. SO011, SO012, SO013, SO014
CO010 The G1 humanoid page advertises 23–43 joints, about 35 kg weight, and a headline price of US$13.5K before tax and shipping. SO006
CO011 The Go2 official page advertises pricing from US$1,600, reinforcing Unitree's low-entry quadruped positioning. SO008, SO005
CO012 The H1 official page positions H1 as Unitree's first universal humanoid robot with continuing OTA updates. SO007
CO013 B2 and Z1 product pages show Unitree sells an industrial quadruped and a separate robotic arm line beyond consumer quadrupeds and humanoids. SO009, SO010
CO014 KrASIA documented Unitree robot performances at the 2021 Spring Festival Gala, giving the company an early nationwide visibility boost. SO003
CO015 The official about page says Unitree also appeared at the 2022 Winter Olympics opening ceremony. SO002
CO016 The official about page and later press coverage place Unitree robots at the 2023 Super Bowl pre-game show and the 2023 Asian Games and Asian Para Games. SO002, SO021
CO017 CES 2025 coverage shows Unitree using G1, H1, Go2, B2, and dexterous-hand demos to market itself internationally. SO019, SO020
CO018 TIME named Unitree to the TIME100 Most Influential Companies list in 2025. SO005
CO019 Xinhua said Unitree launched its humanoid project in 2023 and produced a first bipedal prototype within six months by leveraging its quadruped know-how. SO004
CO020 TechNode reported that Unitree completed Series C in June 2025 and began IPO preparation with CITIC Securities as advisor. SO015
CO021 TechNode reported 2024 revenue of more than RMB 1 billion (US$137 million) and a post-money valuation above RMB 12 billion (about US$1.64 billion). SO015
CO022 Forbes separately reported the June 2025 Series C round at RMB 12 billion (about US$1.7 billion) valuation and verified only a subset of investors including Geely, Ant Group, and HongShan. SO016
CO023 The Robot Report also placed Unitree around RMB 12 billion (about US$1.7 billion) valuation and noted that some estimates put round size near US$97.6 million. SO017
CO024 Humanoid Index compiles Unitree's last round as roughly US$96 million and total funding at US$200 million-plus, but the site presents those figures as reported aggregates rather than official disclosure. SO022
CO025 The public source set does not reconcile to one exact Series C amount or a fully confirmed investor list. SO015, SO016, SO017, SO022
CO026 Later IPO-prep coverage still points back to the same June 2025 Series C as the last concrete private-market pricing anchor. SO015, SO026
CO027 Wire China reported that U.S. universities, police departments, and Army research buyers were purchasing Unitree robots through distributors while political scrutiny intensified. SO021
CO028 Wire China says Unitree claims 60% of worldwide quadruped sales and cites Yole for a 37% share of global humanoid sales, but the reviewed materials do not include a directly auditable market-share methodology. SO021
CO029 Humanoid Index lists Unitree at 1,000-plus employees and 5,500-plus units shipped in 2025 as of its February 2026 update. SO022
CO030 CnTechPost reported Wang's target of 10,000 to 20,000 humanoid shipments in 2026, explicitly as a management forecast rather than a filed order book. SO023
CO031 People.cn said China released a national standard system for humanoid robotics and embodied AI in March 2026. SO025
CO032 China Daily's Hangzhou robotics-hub coverage positions Hangzhou as an increasingly policy-backed home port for robotics manufacturers such as Unitree. SO026
CO033 The arXiv security paper described command-injection, telemetry-exfiltration, and cloud-plane risk around the Unitree G1, making security governance a real diligence issue. SO024
CO034 Washington scrutiny over military linkage and data collection is a material risk to Unitree's U.S. expansion path. SO021, SO024
CO035 KrASIA said Unitree's revenue was already majority robot-sales driven in 2021, with only a minor share from paid performances. SO003
CO036 Official materials say Unitree has pursued industry applications since 2017 across agriculture, power inspection, survey and exploration, public rescue, and industrial settings. SO002
CO037 Publicly reviewed materials do not disclose a board roster, independent directors, or a separately identified CFO/COO bench. SO002, SO016
CO038 Unitree's public narrative consistently emphasizes affordability and faster iteration than higher-priced Western rivals. SO003, SO005, SO016
CO039 At the report date, the best-supported published private-market mark remains TechNode's roughly US$1.6 billion post-money valuation rather than later speculative IPO targets. SO015, SO016, SO017
CO040 IPO preparation is confirmed, but governance detail, exact round amount, and audited operating data still require direct diligence. SO015, SO016, SO017, SO022
CM001 Unitree operates across quadrupeds, humanoids, robotic arms, and sensors rather than a single robot category, so its addressable market spans multiple legged-robot subsegments. SM001, SM002
CM002 Public company and media descriptions show Unitree’s quadrupeds are marketed into consumer and industrial workflows while humanoids are still positioned more toward developers and early adopters. SM010, SM011
CM003 Unitree’s G1 humanoid has a published entry price of US$13.5K before tax and shipping, which materially lowers the threshold for developer and education buyers versus most Western humanoid peers. SM003, SM006
CM004 Unitree’s B2 quadruped is positioned as an industrial inspection and emergency-response robot with more than 6m/s top speed, more than 40kg continuous walking load, and more than 5 hours unloaded endurance. SM005
CM005 Unitree’s Go2 product and storefront make quadruped robots available through direct purchase channels, reinforcing a prosumer and developer market that is already transacting rather than purely hypothetical. SM004, SM007
CM006 Human labor, fixed industrial robots, AMRs, drones, and software-only automation all remain credible substitutes depending on whether the task requires dexterity, mobility, inspection reach, or pure throughput. SM013, SM021
CM007 Goldman Sachs frames the long-run humanoid robot market at at least US$6B in 10–15 years and as much as US$154B by 2035 if affordability and acceptance hurdles are overcome. SM012
CM008 IDC-backed reporting says global humanoid shipments reached about 18,000 units and roughly US$440M of revenue in 2025, which is still a very small actual market relative to headline TAM narratives. SM014, SM016
CM009 IDC-backed reporting says entertainment and performance were the largest humanoid shipment segment in 2025, followed by education and research, data collection, and reception, with industrial manufacturing still earlier in adoption. SM014, SM016
CM010 Unitree stated that its humanoid shipments exceeded 5,500 units in 2025 and that orders ran ahead of deliveries, implying real current commercial demand but also backlog and mix uncertainty. SM014, SM025
CM011 TechNode and The Robot Report both report that Unitree exceeded 1 billion yuan of revenue in 2024 and cleared a roughly US$1.6B–US$1.7B post-money valuation in the 2025 Series C context. SM010, SM011
CM012 Humanoid Index rounds Unitree to a US$1.4B-plus valuation and says the company is eyeing a US$7B IPO valuation, preserving a more promotional forward-looking framing than the latest closed-round anchor. SM025
CM013 Xinhua’s interview with Wang Xingxing cites Morgan Stanley forecasting China’s humanoid market could reach 6 trillion yuan and 59 million units by 2050, a far more expansive lens than near-term shipment data. SM022
CM014 Forbes explicitly contrasts Goldman Sachs’s roughly US$38B hardware-market framing with ARK Invest’s much larger labor-value framing, demonstrating that headline market numbers are often measuring different things. SM021
CM015 IDC’s quadruped market-share study says China manufacturers emerged as leaders in the worldwide quadruped market in 2024, which matters because Unitree’s current commercial strength still leans heavily on quadrupeds. SM015
CM016 Unitree maintains a visible GitHub organization and SDK repository, which indicates an active secondary-development and developer ecosystem rather than a purely closed enterprise sales model. SM008, SM009
CM017 Academic and lab sources show Unitree Go1 is used in navigation, mission-control, and locomotion research, confirming education and research as real budget-bearing segments for Unitree hardware. SM026, SM027
CM018 The Wire China reports that universities, police departments, companies, and even the U.S. Army bought Unitree quadrupeds largely because their price points were far below Western alternatives. SM024, SM007
CM019 Unitree’s official surfaces and Wang Xingxing’s Xinhua interview both point to future home and household aspirations, but present-day evidence still concentrates in developer, research, inspection, and showcase use cases. SM001, SM022
CM020 IFR argues humanoids are attractive because they fit human-centric environments and can address labor shortages, but it also warns that mass adoption as universal helpers is not a near-term outcome. SM013
CM021 China’s 2026 humanoid and embodied-AI standard system covers six components including application plus safety and ethics, giving domestic vendors a clearer compliance path than most foreign rivals currently have. SM018, SM019, SM017
CM022 China Daily says Hangzhou already hosts more than 200 robotics-related enterprises and targets more than 50 billion yuan of sector output by 2027, making Unitree part of a dense local industrial cluster. SM020
CM023 Xinhua describes Unitree as the first company in the world to publicly retail high-performance quadruped robots and as a leader in robot sales volume, which helps explain its unusually broad current market reach. SM022, SM002
CM024 The Spring Festival Gala exposure turned Unitree’s H1 into an entertainment and brand-recognition asset, which supports short-term commercial value even if it does not by itself prove industrial ROI. SM023, SM022
CM025 The Wire China reports that Unitree claims 60 percent of worldwide quadruped sales and 37 percent of global humanoid sales according to Yole Group, but those share figures are company-adjacent and not independently audited in the fetched set. SM024
CM026 The Wire China also reports Wang said roughly half of Unitree’s annual sales now come from overseas, showing international demand exists while also increasing exposure to export and security policy shifts. SM024
CM027 The Wire China attributes Chinese robot cost and speed advantages to dense component ecosystems and aggressive supplier competition, reinforcing a structural cost tailwind for Unitree versus many U.S. peers. SM024
CM028 Unitree’s official product pages emphasize OTA updates and secondary development, implying that the addressable market includes software, maintenance, and integration layers rather than hardware boxes alone. SM003, SM005
CM029 IDC-backed reporting says competition in humanoids is shifting from pure hardware toward services and ecosystems, which matters because Unitree’s long-term market position will depend on more than low ASPs. SM014, SM016
CM030 China’s 2026 standards coverage and 2025 surge to more than 140 manufacturers and 330 humanoid products imply fierce domestic competition even as the policy environment remains supportive. SM018, SM019
CM031 IFR explicitly says traditional industrial robots retain clear function advantages in many repetitive high-speed tasks, so humanoids should not be assumed to replace existing automation across the board. SM013
CM032 The arXiv security paper argues that humanoid robots expand cyberattack surfaces through microphones, cameras, networked control, and physical actuation, adding trust friction to adoption in sensitive environments. SM028
CM033 The Wire China says U.S. lawmakers and defense-oriented critics are increasingly framing Unitree as a security concern, creating procurement risk in public-sector and critical-infrastructure segments. SM024
CM034 Price elasticity is high in Unitree’s current quadruped market because customers cited by The Wire China bought Unitree robots precisely because Boston Dynamics alternatives were unaffordable. SM024, SM005
CM035 Unitree’s near-term obtainable market is broader in quadruped inspection, research, and public-safety workflows than in full humanoid labor replacement because those quadruped use cases already repeat today. SM005, SM015, SM024
CM036 IDC’s 2025 market mix creates a real mismatch between futuristic household narratives and today’s actual deployment base, which still skews toward performances, research, data collection, and reception. SM014, SM016
CM037 Unitree’s own G1 page cautions that humanoid robots remain in an early exploratory stage and that individual buyers should understand their limitations before purchase. SM003
CM038 The contradictory market numbers from Goldman, IDC, Morgan Stanley, and ARK should be treated as bounding scenarios because they differ on geography, time horizon, and whether they measure hardware revenue or labor value. SM012, SM014, SM021, SM022
CM039 Published prices plus direct sales channels mean Unitree is already attacking the developer and prosumer legged-robot market at price bands many Western humanoid peers have not made public. SM003, SM006, SM007
CM040 Unitree’s combination of direct ecommerce, developer repositories, academic usage, and overseas reseller-style sales implies a multi-channel go-to-market model rather than a single enterprise-integrator path. SM006, SM008, SM024, SM027
CP001 Unitree lists the G1 humanoid at roughly $13.5K-$16K before tax and shipping, making it the clearest published low-price benchmark in commercial humanoids. SP001, SP003
CP002 Unitree lists the Go2 quadruped at $2,800, extending its low-price positioning below humanoids into inspection and developer robot dogs. SP004
CP003 The G1 base platform weighs about 35 kg, offers 23 DoF (23-43 for G1 EDU), and publishes only about 2-3 kg arm load with optional dexterous hands rather than standard manipulation hardware. SP001
CP004 Unitree's H1/H1-2 line adds higher torque, faster locomotion, optional dexterous hands, and up to about 7 kg rated arm load on H1-2, showing broader humanoid coverage than G1 alone. SP002
CP005 Unitree publicly warns that humanoid robots remain in an early exploratory stage, asks users to keep a safe distance, and prohibits hazardous modifications or dangerous use. SP001, SP002
CP006 Unitree's GitHub organization spans humanoids, quadrupeds, dexterous hands, robotic arms, lidar, world-model code, app templates, and URDF assets, signaling unusually broad product and developer breadth. SP005, SP006, SP007, SP008
CP007 Unitree exposes SDK2, ROS2 interfaces, reinforcement-learning repos, simulation tooling, teleoperation, and data/robot learning frameworks, making developer access materially deeper than most hardware peers disclose publicly. SP005, SP006, SP007, SP008
CP008 TechNode reports that Unitree exceeded RMB 1 billion of 2024 revenue and that its humanoids are oriented toward developers who customize perception and motion algorithms. SP009
CP009 The Robot Report says Unitree grew to more than 1,000 employees and about $140 million of annual revenue while staying focused on affordable robots. SP010
CP010 Humanoid Index says Unitree shipped 5,500+ units in 2025, raised about $96 million in a June 2025 Series C, and was already being discussed as an IPO candidate. SP011
CP011 Figure's public story has pivoted from factory validation with Figure 02 to home assistance with Figure 03 and Helix, emphasizing AI-led generalization rather than transparent price or deployment disclosures. SP012, SP013
CP012 Humanoid Index characterizes Figure as pre-commercial, BMW-pilot validated, about 800+ employees, and the sector valuation leader at roughly $39 billion. SP014, SP015
CP013 1X positions NEO as a home robot with expert teleoperation fallback, quiet operation, soft-safe design, and a low-friction preorder path rather than a disclosed industrial service stack. SP016, SP017
CP014 1X Technologies has far less capital than Figure or Agility but still advertises a roughly $20K target price point, making it one of the few peers trying to meet Unitree on affordability. SP017, SP018
CP015 Agility says Digit is the first humanoid to prove real value at scale and pairs the robot with Arc workflow controls, on-site service, online support, and real-time monitoring. SP019, SP020
CP016 Agility publicly names Amazon, GXO, and Schaeffler as users and Humanoid Index says roughly 100 commercial units are deployed, giving it stronger enterprise proof than Unitree has disclosed. SP019, SP020
CP017 Boston Dynamics Atlas is engineered for industrial material handling with 50 kg instant payload, 30 kg sustained payload, 4-hour battery life, self-swappable batteries, and Orbit system integrations. SP021
CP018 Boston Dynamics already has field-proven enterprise deployment through Spot in inspection and monitoring, so the company competes on trust, service, and installed-base credibility even if Atlas remains early access. SP021, SP022
CP019 Agibot competes directly against Unitree on published pricing: X2 is listed at $24,240 and A2 Lite at $44,560, both above Unitree G1 but still far more transparent than western peers. SP025, SP026
CP020 Agibot's A2 Ultra page claims over a thousand units in real-world deployment, 24/7 autonomous walking proof, and top-tier CR, CE-MD, CE-RED, and FCC certifications. SP024
CP021 Agibot's ecommerce pages also reveal an enterprise support gap of their own: manual support is limited, technical issues are largely the user's responsibility, and only hardware warranty is explicitly promised on X2. SP025, SP026
CP022 DEEP Robotics is not a direct humanoid peer but is a real substitute in inspection, rescue, utility, and public-safety mobility jobs where a lower-risk quadruped can replace a humanoid. SP027
CP023 UBTECH competes from a different starting point as a broader commercial, education, elderly-care, and consumer robot vendor rather than a pure-play low-cost developer humanoid specialist. SP028
CP024 Fourier positions GR-2 around dexterous manipulation, tactile sensing, 53 joints, and a developer toolkit for Isaac Lab, ROS, and Mujoco, giving it a stronger published manipulation story than Unitree G1 base hardware. SP029
CP025 Forbes lists Unitree, Figure, Agility, Boston Dynamics, 1X, Agibot, and Fourier among the leading humanoid manufacturers, confirming that Unitree is competing inside a dense field rather than a winner-take-all opening. SP030
CP026 Wire China says Unitree robots are being bought by universities, companies, police departments, and even U.S. Army research buyers because they cost a fraction of western alternatives. SP031
CP027 Wire China quotes customers saying Boston Dynamics offers a better overall product but Unitree wins when budgets make the alternative unaffordable. SP031
CP028 Wire China also says Unitree sells in the United States through licensed distributors such as RobotLab rather than a direct local operating presence, limiting support depth but widening access. SP031
CP029 Goldman frames humanoid adoption as constrained by affordability, useful task design, and commercialization timing, which keeps human labor, fixed automation, AMRs, and quadrupeds relevant substitutes today. SP032
CP030 IFR says regional differences and commercial reality still matter more than hype, supporting the view that western enterprise trust and integration proof are not interchangeable with low unit cost. SP033
CP031 Unitree is one of the only humanoid vendors that lets buyers see near-term public list prices and place direct orders, while Figure, Agility, and Boston Dynamics still route buyers through lead qualification or selective deployment processes. SP003, SP012, SP019, SP021
CP032 Agility's Arc and service layer creates stronger workflow lock-in than Unitree's open hardware-plus-SDK model, because it embeds the robot into WMS/MES and fleet operations rather than developer experimentation alone. SP019, SP007
CP033 Unitree's main competitive moat is manufacturing speed plus China-centric component economics, not a demonstrated enterprise workflow monopoly or uniquely closed software stack. SP009, SP010, SP031
CP034 Unitree's broad portfolio across quadrupeds, humanoids, hands, arms, and lidar does create a cross-sell and developer-ecosystem advantage that narrower humanoid specialists lack. SP005, SP006, SP007, SP008
CP035 Publicly available Unitree materials point more strongly to research, developer, and price-sensitive edge deployments than to recurring Fortune-500-style workflow wins with disclosed ROI metrics. SP009, SP031, SP005
CP036 The same openness that helps Unitree win labs and developers also lowers switching cost for integrators who can port work onto other open or ROS-friendly platforms. SP005, SP006, SP008, SP029
CP037 Procurement-ban politics and privacy rhetoric create a competitive wedge that can push government-adjacent or regulated western buyers toward Agility or Boston even when Unitree is cheaper. SP031, SP034
CP038 Independent security criticism around the G1 strengthens buyer concerns that Unitree's low price does not by itself equal enterprise safety, privacy, or compliance readiness. SP035, SP031
CP039 Unitree is strongest where buyers value immediate access, low upfront cost, and customizable development more than fully validated manipulation, workflow software, or western trust posture. SP001, SP003, SP005, SP009, SP031
CP040 Unitree is weakest in manipulation-heavy, compliance-sensitive, or service-intensive enterprise settings where rivals advertise stronger support, tactile hands, workflow software, or established field deployments. SP019, SP021, SP024, SP029, SP035
CI001 Official surfaces show that Unitree sells quadrupeds, humanoids, robot arms, lidar, and related robotics components rather than a single flagship robot. SI004, SI011, SI027
CI002 KrASIA quoted Wang in 2021 saying most revenue came from robot sales and only a minor share from paid performances. SI021
CI003 TechNode said Unitree's quadrupeds are mainly targeted at consumers while its humanoids are geared toward developers who customize perception and motion algorithms. SI001
CI004 Unitree's public GitHub organization reinforces the developer-edition posture around humanoid hardware and software tooling. SI026
CI005 The Go2 official page advertises pricing from US$1,600. SI005
CI006 Unitree's shop page lists Go2 at US$2,800 before freight and customs duties, with higher configurations also shown. SI007
CI007 The G1 official product page advertises a headline price of US$13.5K excluding tax and shipping. SI006
CI008 The G1 shop page lists a US$16,000 ecommerce price. SI008
CI009 The H1 shop page lists US$90,000 before customs duties. SI009
CI010 STEMfinity lists Unitree Z1 robotic arms at US$13,387.50. SI013
CI011 Reviewed B2 pages emphasize industrial capabilities and quote-led workflows, but the extracted page text did not expose a stable self-serve list price. SI010, SI014, SI015, SI016
CI012 The observable public price ladder runs from low-thousands Go2 units to mid-teens G1 pricing, low-teens Z1 arm pricing, and roughly US$90K H1 pricing, with B2 handled more like an enterprise quote product. SI005, SI006, SI007, SI008, SI009, SI010, SI013
CI013 Official shop pages exclude taxes, shipping, freight, or customs duties, so listed prices are anchors rather than landed customer economics. SI007, SI008, SI009
CI014 TechNode reported that Unitree generated more than RMB 1 billion (US$137 million) in revenue in 2024. SI001
CI015 The Robot Report separately said Unitree claimed 1 billion yuan (about US$140 million) in annual revenue. SI002
CI016 Humanoid Index is a secondary-data company profile that aggregates Unitree scale signals, so its headcount and shipment figures should be treated as medium-confidence context rather than filed disclosure. SI003
CI017 CnTechPost reported Wang's 2026 target of 10,000 to 20,000 humanoid shipments based on a Cailian interview. SI017
CI018 The 10,000-20,000 shipment figure is management guidance rather than a filed backlog or booked revenue disclosure. SI017
CI019 TechNode said Unitree completed Series C at a post-money valuation above RMB 12 billion (about US$1.64 billion). SI001
CI020 Forbes also reported a RMB 12 billion (about US$1.7 billion) Series C valuation and verified only a partial subset of investors. SI020
CI021 eHangzhou reported that Unitree expected to submit an IPO application between October and December 2025 and placed post-Series C valuation at 12-15 billion yuan. SI018
CI022 Global Times reported that operating data would be formally disclosed when filing documents were submitted to the exchange. SI019
CI023 Public materials reviewed in-session do not disclose gross margin, cash balance, debt, backlog, CAC, or burn. SI001, SI018, SI019
CI024 Because Unitree sells physical robots and components, its economics are likely more sensitive to manufacturing scale, actuators, batteries, inventory, and field support than a pure software company would be. SI004, SI002, SI010, SI011
CI025 Official and reseller pages imply Unitree can transact consumer and developer SKUs online while industrial SKUs increasingly route through quote-led channels. SI007, SI008, SI009, SI010, SI012, SI013
CI026 Official about and product pages show peripherals, arms, SDKs, and related services as plausible attachment revenue, but no recurring software revenue is publicly disclosed. SI004, SI011, SI026
CI027 Hill Dickinson notes that humanoid deployment creates new liability, privacy, and documentation burdens that can raise service-delivery cost as use cases mature. SI023
CI028 Trade.gov documents continuing U.S. export-control friction around China, a non-trivial cost and compliance overhang for overseas expansion. SI023, SI024
CI029 People.cn said China released a national humanoid standard system in March 2026, a commercialization tailwind that does not by itself solve unit economics. SI022
CI030 If 2024 revenue was roughly US$137-140 million and headcount is roughly 1,000-plus, rough revenue per employee is only about US$137k-140k. SI001, SI002, SI003
CI031 The wide spread between Go2, G1, H1, Z1, and B2 public pricing means Unitree's blended ASP is highly product-mix dependent. SI005, SI006, SI007, SI008, SI009, SI013
CI032 2025 humanoid shipment claims cannot be translated directly into 2024 revenue because the periods differ and Unitree also sells cheaper quadrupeds, components, and services. SI001, SI002, SI003, SI017
CI033 CnTechPost says technical bottlenecks in embodied intelligence still constrain large-scale application, making 2026 shipment guidance execution-sensitive. SI017
CI034 IPO preparation implies Unitree is still financing growth before full public-market transparency is available. SI018, SI019, SI025
CI035 TechNode, Forbes, The Robot Report, and Humanoid Index do not reconcile on the exact Series C amount or a complete investor list. SI001, SI002, SI003, SI020
CI036 B2 reseller and product pages emphasize endurance, payload, and ruggedness rather than transparent checkout pricing, reinforcing an enterprise/industrial sales motion. SI010, SI014, SI015, SI016
CI037 The H1 shop page's customs-duty disclaimer shows that international delivered cost can exceed sticker price materially. SI009
CI038 The U.S. distributor page for Z1 confirms a separate sales channel for the arm, supporting the view that Unitree monetizes peripherals beyond core robot bodies. SI012, SI013
CI039 Public pricing is visible for some SKUs, but realized enterprise discounts, channel margins, and service attach rates remain unknown. SI007, SI008, SI009, SI010, SI013
CI040 A directly accessible filing or prospectus is the clearest next path to resolve margin, cash, and order-book opacity. SI018, SI019, SI025
CI041 The best-supported current financial verdict is that commercial traction exists, but underwriting still depends on future filing-stage disclosure of margins, cash, and backlog. SI001, SI018, SI019, SI023, SI024
CI042 Public evidence today supports hardware sales and visibility, not a clean proof of sustainable gross-margin structure. SI001, SI002, SI007, SI008, SI009, SI023
CE001 Unitree's public product surface spans consumer and industrial quadrupeds, humanoids, dexterous hands, and a robotic arm rather than a single-purpose robot line. SE002, SE011
CE002 The core 2026 lineup visible in reviewed sources includes Go2, B2, G1, H1/H1-2, and Z1. SE002, SE004, SE005, SE006, SE007, SE008
CE003 Go2 is positioned as a retail quadruped for developers and light-duty field use, with pricing starting at $1,600 and built-in OTA, mapping, and secondary-development options. SE006, SE010
CE004 B2 is positioned as an industrial-grade quadruped for inspection and rugged mobility with continuous walking load above 40 kg and unloaded endurance above five hours. SE007
CE005 G1 is sold as a low-cost humanoid developer platform with 23 to 43 joint degrees of freedom, lidar plus depth sensing, and optional dexterous-hand configurations. SE004, SE009
CE006 H1 and H1-2 are full-size humanoids advertised with 3D lidar, depth cameras, and maximum joint torque around 360 N.m on major leg joints. SE005
CE007 Z1 is positioned as an open-program robotic arm for mobile robots, logistics, and research workflows, but its public documentation is much thinner than the legged robots' pages. SE008
CE008 Unitree's about page says the company fully self-researches key robot components including motors, reducers, controllers, lidar, and perception and motion-control algorithms. SE002
CE009 The A1 motor page provides concrete hardware evidence for vertical integration by documenting an integrated motor-controller module with planetary reducer, crossed roller bearing, sensors, RS485, and 1 KHz control. SE024
CE010 Go2 is advertised with Unitree's self-developed 4D lidar L2 offering hemispherical recognition and all-terrain mapping support. SE006
CE011 Unitree's GitHub organization exposes SDK, ROS, URDF, RL, teleoperation, world-model, and lidar-SLAM assets across multiple robot families. SE011
CE012 SDK2 supports current-generation Unitree robots on Ubuntu 20.04 across x86_64 and aarch64 development environments. SE012
CE013 The official ROS2 repository says SDK2 uses CycloneDDS and that ROS2 messages can directly control Go2, B2, and H1-class robots without wrapping the SDK interface. SE014
CE014 Unitree RL Gym documents a Train-to-Play-to-Sim2Sim-to-Sim2Real workflow for Go2, G1, H1, and H1-2. SE013
CE015 The UnifoLM-WMA repository positions Unitree's world-model stack as both a simulation engine and a policy-enhancement layer for general-purpose robot learning. SE015
CE016 The Isaac Lab repository says Unitree's simulator reuses the same DDS topics as the real robot and supports data collection, replay, generation, and model validation. SE017
CE017 The app-templates repository shows Unitree expects third-party applications to be packaged as Docker-backed services for UniStore-style deployment. SE016
CE018 Unitree's official open-source page says its imitation-learning framework covers data collection, algorithm development, model training, and real-machine deployment for G1, Z1, and Dex3 hardware. SE023
CE019 Go2, B2, H1/H1-2, and G1 pages all advertise OTA or upgraded OTA updates, implying a shared software-lifecycle story across the line. SE004, SE005, SE006, SE007
CE020 Unitree's public product pages embed explicit civilian-use and hazardous-modification disclaimers rather than presenting the robots as unrestricted autonomous systems. SE004, SE005, SE006, SE007, SE024
CE021 The G1 page explicitly warns that the global humanoid industry is still in an early exploratory stage and that some sample functions are still under development and testing. SE004
CE022 Xinhua reports that Unitree's public performances generate interim commercial value while broader household and workplace commercialization is still a future objective. SE018
CE023 Unitree's about page says the company is still enriching its product line and improving its production lines across both ToB and ToC legged, humanoid, and arm products. SE002
CE024 A 2026 China Daily government article describes Hangzhou's robotics infrastructure as combining testing, certification, pilot production, and training for companies including Unitree. SE021
CE025 Robotics & Automation News says Unitree founder Wang serves as a deputy director of China's humanoid standardization committee and publicly frames standards as essential for long-sequence work. SE020
CE026 Go2's public page says part of the robot's functionality still requires human operation or secondary development rather than turnkey autonomy. SE006
CE027 B2's public page says its top-speed claim is realized only in special configurations and that practical speed may be limited for safety. SE007
CE028 Independent CES coverage in early 2025 shows Unitree presenting Go2, Go2-W, B2-W, H1, and G1 together as a coherent lineup to a global buyer audience. SE027, SE028
CE029 RoboticsTomorrow describes Go2's CES demo as including deep-RL-driven movements such as triple backflips and handstands, reinforcing that Unitree uses public showcases to market its autonomy stack. SE027, SE028
CE030 Correll Lab found a Unitree Go-1 could navigate rugged terrain but also slipped, entangled in vegetation, and required manual disentangling and rebooting in field use. SE025
CE031 A 2025 cybersecurity paper alleges that the Unitree G1 exposes BLE provisioning and encryption weaknesses that could enable root access and covert surveillance behavior. SE026
CE032 Unitree's about page says the company has submitted more than 200 patent applications and has more than 180 authorized patents. SE002
CE033 Forbes included Unitree among 16 leading humanoid manufacturers in 2025, supporting the view that the company matters in the global humanoid discussion rather than only in quadrupeds. SE022
CE034 Humanoid Index describes Unitree as a mass-production leader with more than 5,500 units shipped in 2025, but that figure is a secondary aggregation rather than an audited company filing. SE022
CE035 Retail and distributor surfaces show a two-track commercialization model in which some robots are sold directly online while higher-complexity deployments still rely on support-heavy partner workflows. SE009, SE010, SE016, SE027
CE036 The G1 store page explicitly markets imitation learning, reinforcement learning, and world-model creation as part of the humanoid's developer proposition. SE009
CE037 The Go2 store page couples low pricing with an EDU upsell and GPT-style marketing language, signaling that Unitree uses a low-friction retail funnel to seed a broader developer pipeline. SE010
CE038 The ROS2 repository exposes concrete developer tasks including low-level control, sport mode state, wireless-controller status, and lidar visualization rather than only high-level marketing demos. SE014
CE039 The Isaac Lab repository warns about officially recommended hardware resources, first-run loading overhead, and specific tested GPU generations, indicating nontrivial compute and integration dependencies. SE017
CE040 Unitree's strongest technical differentiation is the combination of vertically integrated embodied hardware and a broad open developer surface, while the strongest open risks are safety validation, security response, and real-world reliability. SE002, SE011, SE024, SE025, SE026
CU001 The strongest public customer evidence clusters into research labs, education and developer buyers, industrial inspection prospects, public-event organizers, and western distributors or integrators. SU001, SU002, SU013, SU014, SU019, SU020
CU002 Unitree's about page says its robots appeared at the 2021 Spring Festival Gala, the 2022 Winter Games opening ceremony, the 2023 Super Bowl, the 2023 Asian Games and Asian Para Games, and the 2025 Spring Festival Gala. SU002
CU003 Those marquee event appearances are visibility or showcase evidence rather than proof of durable paying customer deployment. SU002, SU010, SU011, SU012, SU017
CU004 Unitree's official surfaces say its robots are used in agriculture, industry, power inspection, survey and exploration, and public rescue. SU001, SU002
CU005 The Go2 page specifically highlights an Asian Games field workflow in which a robot dog carried discus and javelin on the field. SU004
CU006 RoMI Lab says its Unitree Go1 has been part of the lab since Spring 2022 and is a key platform for legged and agile locomotion research. SU014
CU007 A Zenodo-backed RTU MIREA paper documents a Go1 Air integration with ArduPilot controllers and reports 206.7 ms average latency with standard deviation below 53 ms. SU015
CU008 Correll Lab's field deployment report says a Unitree Go-1 can navigate rugged terrain but can slip, tangle in shrubs, and require manual recovery and extra engineering. SU016
CU009 Taken together, RoMI, Zenodo, and Correll provide the strongest public proof that Unitree robots are actually used and iterated on by external research customers. SU014, SU015, SU016
CU010 The Wire China says universities, companies, local police stations, and even the U.S. Army are looking to buy Unitree robotic dogs across America. SU013
CU011 The Wire China names Central Connecticut State University as a U.S. campus operator of a Unitree Go2 robotic dog. SU013
CU012 The Wire China says Topeka Police Department bought three Unitree robotic dogs because Boston Dynamics alternatives were unaffordable. SU013
CU013 The Wire China says U.S. Army buyers earmarked more than $60,000 for four Unitree robotic dogs since 2023 through American distributors. SU013
CU014 The Wire China says Unitree does not have a direct presence in the United States and instead sells through licensed distributors. SU013
CU015 Futurology says it became an authorized U.S. distributor in 2024 and explicitly markets Unitree hardware to government and federally funded institutions. SU021
CU016 TVC Robotics markets itself as an official Unitree distributor in the USA and Canada with official warranty and professional support. SU019
CU017 Drones Plus Robotics says it is an official Unitree dealer providing integration, training, and ongoing support for enterprise, education, and research buyers. SU020
CU018 RoboStore markets itself as an official Unitree partner serving educational institutions, industrial inspections, search and rescue, law enforcement, and autonomous applications. SU022
CU019 US Robot Systems markets itself as an official Unitree distributor with U.S. support across humanoids, quadrupeds, robotic arms, and dexterous hands. SU023
CU020 The western sales model appears channel-assisted and support-heavy rather than pure e-commerce because distributor pages emphasize onboarding, warranty, integration, and training. SU013, SU019, SU020, SU021, SU022, SU023
CU021 Official shop pages show that at least some Unitree robots can be bought or priced directly online, lowering procurement friction for smaller developer and lab buyers. SU008, SU009
CU022 The Go2 store page pairs direct pricing with shipping, customs, and EDU contact-sales language, showing a gradient from retail checkout to managed procurement. SU008
CU023 The Go2 official page says certain 4G functions outside mainland China support 43 European countries plus some Asian regions, indicating public export-localization constraints. SU004
CU024 The Wire China says Wang reported that half of Unitree's annual sales now come from overseas. SU013
CU025 Xinhua says Unitree's public performances serve the dual purpose of showcasing technological progress and generating interim commercial value before broader workplace and household adoption. SU010
CU026 Global Times, AIbase, and Xinhua all confirm that 16 Unitree humanoid robots performed at the 2025 Spring Festival Gala. SU010, SU011, SU012
CU027 RoboticsTomorrow says Unitree was featured at the 2023 Super Bowl pre-game show and the 2022 Winter Olympics, reinforcing the breadth of its event-led visibility. SU017
CU028 Repeated event and CES appearances support awareness and top-of-funnel demand generation, but they do not establish customer durability or renewal economics. SU010, SU017, SU018
CU029 CES 2025 coverage shows Unitree presenting a multi-product lineup in Las Vegas, suggesting a deliberate global buyer-acquisition effort rather than China-only visibility. SU017, SU018
CU030 Industrial inspection has abundant marketing support but weak independent customer proof because named third-party operators and quantified outcomes were not found in the reviewed public sources. SU001, SU003, SU005, SU022, SU025
CU031 Unitree's homepage and news pages promote a power intelligent inspection solution for dangerous, urgent, and repetitive tasks. SU001, SU003
CU032 The about page says Unitree's robots have been widely used in agriculture, industry, power inspection, survey and exploration, and public rescue without naming specific operators. SU002
CU033 Research and education demand surfaces recur across RoMI Lab, the Zenodo paper, Correll Lab, Drones Plus Robotics, RoboStore, and Futurology. SU014, SU015, SU016, SU020, SU021, SU022
CU034 No public NRR, GRR, churn cohort, or consolidated customer-count disclosure was found in the reviewed source set. SU001, SU002, SU013, SU024
CU035 The best public repeat-usage proxies are long-lived lab usage since 2022 and recurring marquee events from 2021 to 2025, but neither proxy proves enterprise renewal. SU002, SU014
CU036 Public proof quality is highest for research usage, moderate for channel and procurement evidence, and lowest for industrial inspection and event-based commercial claims. SU013, SU014, SU015, SU016, SU019, SU020, SU021, SU022
CU037 Because public proofs are densest in research, showcase, and distributor-led export surfaces, those segments are the likeliest concentration pockets in the current evidence set. SU010, SU013, SU014, SU019, SU020, SU021
CU038 Western procurement is exposed to political and security friction because The Wire China documents congressional scrutiny and the lack of direct U.S. presence. SU013
CU039 The 2025 G1 security paper adds an additional procurement risk because institutions may hesitate to buy robots that carry unresolved surveillance or command-injection allegations. SU026
CU040 Low price versus western rivals is a recurring reason Unitree converts interest into purchases or trials in research, police, and distributor-led channels. SU008, SU009, SU013
CR001 The American Security Robotics Act introduced in March 2026 would ban the U.S. federal government from procuring and operating unmanned ground systems manufactured by foreign entities of concern, including humanoid robots. SR005, SR006
CR002 IEEE Spectrum says the proposed Chinese-robot ban is part of a broader U.S. tech-sovereignty push and warns that American robot makers still depend on Chinese-made components. SR006
CR003 Trade.gov states that the United States imposes export controls on dual-use and less-sensitive military items under the EAR in response to China's military modernization and related security concerns. SR001, SR002
CR004 The CRS and GAO records show that advanced-semiconductor export controls toward China have expanded and continue to impose compliance burdens and uncertainty around scope and enforcement. SR002, SR003
CR005 USCC frames U.S.-China competition in emerging technologies as strategic and long-term, which supports the view that robotics scrutiny is structural rather than a one-off political flare-up. SR004
CR006 Sponsor rhetoric around the ASRA bill explicitly frames Chinese robots as privacy and national-security threats, raising the chance that enterprise buyers self-restrict before any formal ban passes. SR005, SR006
CR007 Hill Dickinson says humanoid deployment creates unresolved safety, liability, and accountability questions over whether responsibility sits with the manufacturer, operator, or software provider. SR007
CR008 Hill Dickinson also says visual, biometric, and cloud-processed data can conflict with GDPR-style privacy regimes, especially when storage or transmission paths are opaque. SR007
CR009 The arXiv G1 assessment reports a BLE command-injection path to root access via malformed Wi-Fi credentials and says the robot transmitted telemetry every 300 seconds without operator notice. SR011
CR010 The Wire China reports that cybersecurity researchers said Unitree humanoids could send audio and visual data back to servers in China and that policy critics tied those concerns to procurement scrutiny. SR011, SR012
CR011 Unitree's own G1 and H1 pages warn buyers that humanoids remain early-stage products requiring cautious operation and safe distance from people. SR016, SR017
CR012 The G1 page publishes only an 8-month warranty for the base unit and 18 months for G1 EDU, which is short compared with typical multi-year enterprise automation lifecycles. SR016
CR013 G1 publishes only about 2 kg arm load on the base model and optional hands, showing that low purchase price comes with real manipulation and production-use limitations. SR016, SR018
CR014 TechNode says Unitree's humanoids are geared toward developers who customize perception and motion algorithms, which reduces product maturity risk for R&D users but increases packaging and support risk for mainstream enterprises. SR013
CR015 Unitree's GitHub footprint shows a broad, open developer stack spanning SDK, ROS2, simulation, and reinforcement learning, which supports adoption but also lowers switching cost and increases imitation risk. SR020, SR021, SR022, SR023
CR016 The Wire China attributes part of Unitree's cost edge to China's concentrated component supply chains, implying that export-control escalation or domestic bottlenecks could hit both cost and speed advantages. SR012
CR017 The Robot Report says Unitree can be up to 50% cheaper than other systems, but lower upfront cost does not answer questions on MTBF, field returns, or enterprise service economics. SR014
CR018 Humanoid Index and TechNode together imply that Unitree is scaling quickly—5,500+ units shipped in 2025 after more than RMB 1 billion of 2024 revenue—but public quality KPIs remain absent. SR013, SR015
CR019 China's 2026 humanoid standards rollout may improve domestic testing discipline and certification consistency, but it also creates the risk that Chinese and western compliance frameworks diverge further. SR008, SR009, SR010
CR020 Hangzhou's robotics-hub buildout shows that Unitree benefits from concentrated local policy support, pilot-zone infrastructure, and ecosystem clustering. SR024
CR021 Strong local policy support mitigates near-term domestic scaling risk but also increases concentration on Chinese policy continuity and can worsen western geopolitical perceptions. SR020, SR024, SR025
CR022 The Wire China says U.S. buyers currently include universities, police departments, and Army research buyers, which proves interest but not broad industrial ROI or repeatable production workflows. SR012
CR023 The same Wire China reporting says Unitree sells through licensed distributors such as RobotLab rather than a direct U.S. operating presence, leaving support depth and compliance ownership structurally ambiguous. SR012
CR024 RobotLab told The Wire that most customers buy Unitree for research because there is no set use case, making customer-proof risk one of the cleanest thesis-break variables for the story. SR012
CR025 Agility publicly discloses Arc workflow software, on-site service, live monitoring, and named customers, which sharpens Unitree's customer-proof gap rather than reducing it. SR030
CR026 Boston Dynamics publishes an industrial-performance Atlas story while Spot already has broad field deployment, setting a trust and support benchmark that Unitree has not matched publicly. SR031
CR027 Figure says it is shipping a safe, high-quality product and 1X emphasizes passive safety in the home, illustrating how peers frame trust and safety as product attributes rather than side notes. SR032, SR033
CR028 TechNode says Unitree began IPO preparation with CSRC guidance while The Robot Report and Humanoid Index put its private valuation around $1.6B-$1.7B and discuss much higher IPO aspirations. SR013, SR014, SR015
CR029 A move from a $1.6B-$1.7B private valuation range toward a rumored $7B IPO narrative would require public-market buyers to underwrite both sustained scale and policy resilience that are not yet proven in the open record. SR013, SR014, SR015
CR030 Goldman and IFR both argue that adoption still depends on affordability, useful task fit, and commercialization reality, which means shipment ambition can outrun real paying demand. SR028, SR029
CR031 IDC-linked coverage says China dominated 2025 humanoid shipments, which supports Unitree's scale story but also makes it a visible target in geopolitical decoupling debates. SR025, SR027
CR032 Unitree's open developer stack is a moat-building distribution tool, but it also means the company's IP defensibility is less obvious than a closed workflow platform or proprietary AI service layer. SR020, SR021, SR023, SR030
CR033 No major public litigation against Unitree surfaced in the reviewed sources, which lowers immediate legal cash-burn risk but leaves product-liability precedent and insurance posture unresolved. SR007, SR013
CR034 Residual geopolitical exposure remains high even with domestic policy support because the same Chinese origin that helps manufacturing speed is the root cause of procurement-ban and privacy scrutiny abroad. SR005, SR006, SR024
CR035 Residual cyber exposure remains high because the public record shows a specific independent exploit path and telemetry concerns, while Unitree has not published an equivalent third-party security assurance package. SR011, SR012, SR016
CR036 Residual quality exposure remains medium-high because Unitree has real scale and OTA cadence, but short warranties and missing public reliability metrics make field durability hard to underwrite. SR013, SR015, SR016, SR017
CR037 Residual customer-proof exposure remains high because current overseas use cases cluster in research, public safety, and distributor-led purchases rather than large disclosed production deployments. SR012, SR030, SR031
CR038 Residual valuation exposure remains high because a successful IPO would crystallize scrutiny on policy, customer concentration, support obligations, and disclosure quality all at once. SR013, SR014, SR015
CR039 A thesis-break policy event would be passage of procurement restrictions that close off meaningful overseas public-sector or regulated-industry demand before Unitree proves alternative enterprise channels. SR005, SR006, SR012
CR040 A thesis-break cyber event would be a confirmed exploit, data-exfiltration finding, or serious injury incident that forces product recalls, customer pauses, or regulator intervention. SR007, SR011, SR016
CR041 A thesis-break commercial event would be IPO delay or repricing combined with continued absence of named industrial customers and support metrics. SR013, SR014, SR015, SR030
CR042 Mitigations exist—domestic standards, local policy support, OTA updates, broad developer reach, and low pricing—but none eliminates the core externality that Chinese origin creates in western procurement and privacy debates. SR009, SR010, SR016, SR024
CR043 Brookings fellow Kyle Chan told the House Select Committee in April 2026 that U.S.-China AI and robotics competition is strategic and ongoing, reinforcing that scrutiny of Chinese robotics platforms is likely to persist across agencies and committees. SR004, SR034
CR044 Associated Press reported that the Trump administration banned imports of new foreign-made routers on supply-chain and cybersecurity grounds in April 2026, showing that hardware-security restrictions can move quickly from policy concern to import control. SR006, SR035
CV001 TechNode and The Robot Report place Unitree’s 2025 Series C post-money valuation around RMB12B or roughly US$1.6B–US$1.7B while also reporting 2024 revenue above 1 billion yuan. SV001, SV002
CV002 TechNode says Unitree has begun IPO preparation with CITIC Securities as advisor, which makes public-listing optionality part of the current valuation narrative rather than a distant fantasy. SV001
CV003 Humanoid Index summarizes Unitree at a US$1.4B-plus valuation and says management is eyeing a US$7B IPO valuation, which is directionally useful but not the same as the last closed-round anchor. SV003
CV004 Humanoid Index’s US$1.4B-plus summary understates or rounds down the later US$1.6B–US$1.7B Series C anchor reported by TechNode and The Robot Report. SV001, SV002, SV003
CV005 Unitree already sells the G1 at a published US$13.5K price point, giving investors evidence of real product-market commercialization that many richer peers still present mainly as pilots or future potential. SV004
CV006 Goldman’s framework implies private humanoid valuations should be scenario-based because large outcomes depend on clearing use-case, affordability, and public-acceptance hurdles that remain unresolved. SV006
CV007 IDC-backed reporting says the entire global humanoid market produced only about US$440M of revenue in 2025 and that deployments were still dominated by performances, education, data collection, and reception. SV007, SV032
CV008 Using the public floor of roughly US$137M for 2024 revenue, Unitree’s current implied valuation equates to roughly a 12x trailing-revenue multiple before any adjustment for margins, cap table, or dilution. SV001, SV002
CV009 Figure AI is the sector’s outlier valuation case at roughly US$39B, making Unitree’s US$1.6B–US$1.7B mark look tiny by comparison but not necessarily cheap on fundamentals. SV010, SV011
CV010 Figure’s official materials emphasize both workforce and home-use ambition, but they do not disclose revenue, unit economics, or customer concentration, so the US$39B mark is heavily future-state priced. SV008, SV009, SV010
CV011 Agility’s official materials plus Humanoid Index show a commercially deployed logistics position and a verified US$400M Series C, with Humanoid Index summarizing valuation above US$2.1B. SV012, SV013
CV012 1X’s NEO page plus Humanoid Index support a materially lower valuation reference than Unitree, with 1X framed around a US$500M-plus valuation after a US$100M Series B and a home-robot thesis. SV014, SV015
CV013 Apptronik’s official materials, TechCrunch, Crunchbase, Reuters-linked US News coverage, and SEC-linked reporting support a roughly US$5B valuation, an initial US$350M Series A followed by a US$520M extension round, and nearly US$1B total capital raised. SV016, SV017, SV018, SV019, SV033, SV034
CV014 Agibot is a useful China comp on ambition and embodied-AI positioning, but the fetched official evidence does not give a clean current valuation anchor, so it should not be used as a primary pricing comp. SV021
CV015 UBTech is useful as a public Chinese robotics reference, but the fetched official and HKEX pages do not expose a clean machine-readable market-cap figure in text mode, limiting its use as a numeric comp here. SV023
CV016 Boston Dynamics’ fetched official pages are best treated as technology and strategic-reference points rather than direct valuation comps because no current standalone equity value is disclosed in the fetched set. SV024, SV025
CV017 The Wire China shows Unitree wins real overseas purchases because its robots cost a fraction of Western alternatives, which is a commercialization strength that more narrative-heavy peers do not necessarily match. SV026, SV004
CV018 Security and procurement scrutiny around Chinese robots supports a structural discount for Unitree relative to U.S. peers even if Unitree leads on product availability and affordability. SV026, SV027
CV019 Public evidence supports paying more than a generic hardware-startup multiple for Unitree because the company has shipped products, disclosed pricing, overseas demand, and a real revenue floor. SV001, SV002, SV004, SV026
CV020 Public evidence does not support paying a Western-software-style premium for Unitree because gross margin, segment mix, warranty reserves, and recurring software revenue remain undisclosed. SV001, SV002, SV006, SV007
CV021 The gap between Unitree and peers like Figure or Apptronik reflects not just company quality but also domicile, security exposure, and revenue-mix differences between low-cost product sales and future AI-platform narratives. SV010, SV013, SV018, SV026
CV022 China’s standards push and Hangzhou cluster add domestic momentum that can help Unitree grow and improve IPO storytelling, but they do not by themselves erase the foreign-policy discount. SV029, SV030, SV031
CV023 A supportable bull case for Unitree assumes 2028 revenue reaches roughly US$500M–US$800M as industrial adoption expands on top of current developer, research, and quadruped demand. SV001, SV002, SV029
CV024 A supportable base case assumes 2028 revenue reaches roughly US$300M–US$450M with continued volume growth but a demand mix still dominated by developers, research, performances, and early commercial users. SV001, SV007, SV032
CV025 A supportable bear case assumes 2028 revenue reaches only US$180M–US$250M because security scrutiny, ASP compression, and low-margin hardware mix cap monetization quality. SV006, SV026, SV027
CV026 Applying roughly 6–8x, 4–6x, and 3–4x EV/revenue bands to the bull, base, and bear revenue cases yields valuation bands of about US$3.0B–US$6.4B, US$1.2B–US$2.7B, and US$0.6B–US$1.0B respectively. SV006, SV010, SV013, SV018
CV027 At the current roughly US$1.6B–US$1.7B entry, the base case offers only moderate upside while the bear case can lose capital, so public evidence supports discipline rather than aggressive entry. SV001, SV006, SV026
CV028 The US$7B IPO aspiration should be treated as a bull-case narrative marker, not as the base case for underwriting today’s private-market price. SV003, SV029
CV029 The positive thesis rests on real product availability, broad channel reach, China cost advantages, and the unusual fact that Unitree already monetizes both quadruped and humanoid platforms. SV001, SV004, SV005, SV026
CV030 Unitree’s low-price strategy can create upside through scale because it opens developer, education, inspection, and early commercial budgets that are closed to higher-priced peers. SV004, SV026, SV029
CV031 The anti-thesis is that low price may also mean structurally lower gross margin, while today’s measured market still skews to developers, performances, and research rather than durable industrial ARR. SV007, SV017, SV026
CV032 Open international sales also increase political and cyber scrutiny, which can cut off the very overseas channels that currently prove Unitree’s commercial appeal. SV026, SV027
CV033 Apptronik’s US$5B and Figure’s US$39B financings show investors will pay dramatically higher prices for humanoid optionality, but those rounds rely heavily on strategic investors and future automation narratives. SV010, SV017, SV018
CV034 Agility and 1X suggest the more grounded comparable range for pre-scale humanoid companies still sits far below Figure’s outlier mark and straddles Unitree’s current price more closely. SV013, SV015
CV035 Unitree looks cheap relative to Figure and Apptronik but not obviously cheap relative to its own disclosed evidence once China risk and missing margins are taken seriously. SV001, SV003, SV006, SV018, SV026
CV036 Public sources do not disclose Unitree’s revenue split across quadrupeds, humanoids, components, and software or services. SV001, SV002
CV037 Public sources do not disclose Unitree’s gross margin, warranty reserve, or failure-rate profile, leaving investors unable to judge whether low prices are sustainable or dilutive. SV001, SV002, SV026
CV038 Fetched filing sources confirm that private-market filings exist around peer rounds, but they do not expose Unitree’s exact preference stack or liquidation economics. SV019, SV020
CV039 A major expansion of export or procurement restrictions beyond niche public-sector scrutiny would be a thesis-break event because it would compress Western TAM and likely Unitree’s exit multiple. SV026, SV027
CV040 If Unitree’s 2025 shipment claims do not translate into repeat industrial or enterprise contracts by 2027, the company is likely to be re-rated as a low-margin hardware vendor rather than a robotics platform. SV007, SV029
CV041 A materially lower next round or a shelved IPO process would falsify the most optimistic valuation narrative and likely push Unitree toward the bear-case valuation band. SV001, SV003
CV042 The recommendation is to track rather than buy aggressively at the current implied price unless investors receive hard evidence on segment economics, industrial contract quality, and downside protection. SV001, SV006, SV026
CV043 Agibot and UBTech remain relevant China references, but weak supportability in the fetched set means they should be treated as qualitative context rather than primary price anchors. SV021, SV023
CV044 Unitree’s dual exposure to quadrupeds and humanoids makes it less binary than pure-humanoid peers, but it also complicates comparable selection because much of the peer set is humanoid-only. SV001, SV005, SV025
CV045 Taken together, the fetched evidence supports a fair-to-stretched valuation stance: not an obvious bubble like Figure’s outlier case, but not an obvious bargain absent better economics disclosure. SV001, SV006, SV018, SV026
来源
编号出版方标题引文
SO001 Unitree Robotics 宇树科技—全球四足机器人行业开创者
SO002 Unitree Robotics About Unitree - Unitree Robotics Unitree Robotics is the world's first company to start public retail of high-performance quadruped robots.
SO003 KrASIA Unitree Robotics develops personal robot dogs that can jog with you The majority of the company's revenue comes from robot sales, Wang said, while a minor percentage is obtained from public performances at private events.
SO004 Xinhua Unitree Robotics shares vision for China's robot revolution Founded in east China's tech hub Hangzhou in 2016, this world-renowned consumer robotics company has been consistently enhancing the operational capabilities of robots to perform practical tasks for humans.
SO005 TIME TIME100 Most Influential Companies 2025: Unitree Robotics At this year’s Spring Festival Gala...16 humanoid robots...were manufactured by Unitree.
SO006 Unitree Robotics Humanoid robot G1_Humanoid Robot Functions_Humanoid Robot Price
SO007 Unitree Robotics Universal humanoid robot H1_Bipedal Robot_Humanoid Intelligent Robot Company
SO008 Unitree Robotics Robot Dog Go2_Quadruped_Robot Dog Company
SO009 Unitree Robotics Robot Dog B2_Inspection Robot_B2 Robot Dog Advantage Usage
SO010 Unitree Robotics Robotic Arm Z1_Mobile Robot On-board Robotic Arm
SO011 GitHub / Unitree Robotics Unitree Robotics
SO012 GitHub / Unitree Robotics GitHub - unitreerobotics/unitree_sdk2: Unitree robot sdk version 2. https://support.unitree.com/home/zh/developer
SO013 GitHub / Unitree Robotics GitHub - unitreerobotics/unitree_rl_gym
SO014 GitHub / Unitree Robotics GitHub - unitreerobotics/unitree_ros2
SO015 TechNode Unitree Robotics begins IPO prep, valued at $1.6 billion after Series C funding · TechNode Unitree recorded over one billion yuan ($137 million) in revenue in 2024.
SO016 Forbes China’s Robot Maker Unitree Valued At $1.7 Billion In Series C Round Unitree Robotics...completed series C funding this month that values the company at 12 billion yuan ($1.7 billion).
SO017 The Robot Report (via Wayback) Unitree becomes a legged robot unicorn with Series C funding - The Robot Report Some estimates put the round at $97.6 million.
SO018 Global Times Dancing robots at Spring Festival gala raise attention in US media regarding China's technological advancements
SO019 Robotics & Automation News Unitree showcases ‘advanced high-mobility robots’ at CES
SO020 RoboticsTomorrow Unitree Showcases Advanced High-Mobility Robots at CES 2025 | RoboticsTomorrow
SO021 The Wire China While Politicians Fret, Unitree is Selling Robots Across America - The Wire China Lawmakers have ratcheted up scrutiny on the company, alleging ties to the Chinese military and raising alarms about its data collection practices.
SO022 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More | Humanoid Index
SO023 CnEVPost / CnTechPost Unitree founder expects up to 20,000 humanoid robot shipments this year - CnTechPost Unitree targeting 10,000 to 20,000 units.
SO024 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors The robot functions as a trojan horse, continuously exfiltrating multi-modal sensor and service-state telemetry.
SO025 People.cn / Xinhua China releases national standard system for humanoid robotics and embodied AI
SO026 China Daily / eHangzhou Hangzhou builds 'global home port' for robotics industry
SM001 Unitree Robotics Unitree Robotics homepage
SM002 Unitree Robotics About Unitree
SM003 Unitree Robotics Unitree G1 humanoid robot specifications and price Price (Tax and Shipping cost excluded): US $13.5K.
SM004 Unitree Robotics Unitree Go2 quadruped product page
SM005 Unitree Robotics Unitree B2 inspection quadruped product page
SM006 Unitree Store Unitree G1 store listing
SM007 Unitree Store Unitree Go2 store listing
SM008 GitHub Unitree Robotics GitHub organization
SM009 GitHub unitree_sdk2 repository
SM010 TechNode Unitree Robotics begins IPO prep, valued at $1.6 billion after Series C funding
SM011 The Robot Report Unitree becomes a legged robot unicorn with Series C funding
SM012 Goldman Sachs Research Humanoid robot: The AI accelerant Goldman Sachs Research estimates that the global market for humanoid robots may reach a market size of at least US$6bn in 10-15 years... up to US$154bn by 2035E in a blue-sky scenario.
SM013 International Federation of Robotics New IFR position paper on humanoid robots published
SM014 CGTN IDC report: China leads the global humanoid robot rise in 2025
SM015 IDC Worldwide Quadruped Robot Market Shares, 2024
SM016 IDC Worldwide Humanoid Robotics Market Analysis 2026
SM017 Robotics & Automation News Why China’s new humanoid robot standards could change the industry
SM018 People’s Daily / Xinhua China releases national standard system for humanoid robotics and embodied AI
SM019 CGTN / Xinhua China releases national standards for humanoid robots and embodied AI
SM020 China Daily Hangzhou builds global home port for robotics industry
SM021 Forbes Humanoid Robots: Here Are The 16 Leading Manufacturers
SM022 Xinhua Unitree Robotics shares vision for China’s robot revolution
SM023 Global Times Dancing robots at Spring Festival gala raise attention in US media
SM024 The Wire China While Politicians Fret, Unitree is Selling Robots Across America
SM025 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More
SM026 RoMI Lab Robot Motor Intelligence Lab — Unitree Go1
SM027 Zenodo Enabling Navigation and Mission-based Control on a Low-cost Unitree Go1 Air Quadrupedal Robot
SM028 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors
SP001 Unitree Robotics Humanoid robot G1_Humanoid Robot Functions_Humanoid Robot Price
SP002 Unitree Robotics Universal humanoid robot H1_Bipedal Robot_Humanoid Intelligent Robot Company
SP003 Unitree Robotics Unitree G1
SP004 Unitree Robotics Unitree Go2
SP005 Unitree Robotics on GitHub Unitree Robotics
SP006 Unitree Robotics on GitHub GitHub - unitreerobotics/unitree_sdk2
SP007 Unitree Robotics on GitHub GitHub - unitreerobotics/unitree_rl_gym
SP008 Unitree Robotics on GitHub GitHub - unitreerobotics/unitree_ros2
SP009 TechNode Unitree Robotics begins IPO prep, valued at $1.6 billion after Series C funding · TechNode
SP010 The Robot Report Unitree becomes a legged robot unicorn with Series C funding - The Robot Report
SP011 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More | Humanoid Index
SP012 Figure Figure
SP013 Figure Company | Figure
SP014 Humanoid Index Figure AI: Funding, Valuation, Robot Specs & More | Humanoid Index
SP015 Tech Market Briefs Figure AI IPO 2026: $39B Valuation, Risks & Bull Case | TMB
SP016 1X Technologies NEO Home Robot
SP017 1X Technologies Introducing NEO Gamma
SP018 Humanoid Index 1X Technologies: Funding, Valuation, Robot Specs & More | Humanoid Index
SP019 Agility Robotics Humanoid Solutions | Agility
SP020 Humanoid Index Agility Robotics: Funding, Valuation, Robot Specs & More | Humanoid Index
SP021 Boston Dynamics Atlas Humanoid Robot | Boston Dynamics
SP022 Boston Dynamics Spot | Boston Dynamics
SP023 AGIBOT AGIBOT Official Website
SP024 AGIBOT AGIBOT A2 Ultra
SP025 AGIBOT AGIBOT X2
SP026 AGIBOT AGIBOT A2 Lite
SP027 DEEP Robotics DEEP Robotics - Global Quadruped Robot Leader
SP028 UBTECH Robotics UBTECH: Humanoid Robot | AI Education Robot | Commercial Robot Solutions | UBTECH Robotics
SP029 Fourier New_Milestone_of_Humanoid_Robotics
SP030 Forbes Humanoid Robots: Here Are The 16 Leading Manufacturers
SP031 The Wire China While Politicians Fret, Unitree is Selling Robots Across America - The Wire China
SP032 Goldman Sachs Research Humanoid robot: The AI accelerant
SP033 International Federation of Robotics New IFR position paper on humanoid robots published
SP034 Congresswoman Elise Stefanik Stefanik, Cotton Introduce Bipartisan Bill to Propel America's Robotics Superiority, Protect U.S. National Security
SP035 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors
SI001 TechNode Unitree Robotics begins IPO prep, valued at $1.6 billion after Series C funding · TechNode Unitree recorded over one billion yuan ($137 million) in revenue in 2024.
SI002 The Robot Report (via Wayback) Unitree becomes a legged robot unicorn with Series C funding - The Robot Report Unitree also sells the Z1 robotic arms and the L2 4D lidar scanners, as well as components and services.
SI003 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More | Humanoid Index
SI004 Unitree Robotics About Unitree - Unitree Robotics
SI005 Unitree Robotics Robot Dog Go2_Quadruped_Robot Dog Company
SI006 Unitree Robotics Humanoid robot G1_Humanoid Robot Functions_Humanoid Robot Price
SI007 Unitree Robotics Shop Unitree Go2
SI008 Unitree Robotics Shop Unitree G1
SI009 Unitree Robotics Shop Unitree H1
SI010 Unitree Robotics Shop Unitree B2
SI011 Unitree Robotics Robotic Arm Z1_Mobile Robot On-board Robotic Arm
SI012 shop-unitree.us Unitree Z1
SI013 STEMfinity Unitree Z1 Robotic Arms
SI014 RobotShop Unitree B2 Robotic Dog
SI015 Top3DShop Unitree Robotics B2
SI016 RobotXShop B2
SI017 CnEVPost / CnTechPost Unitree founder expects up to 20,000 humanoid robot shipments this year - CnTechPost Unitree targeting 10,000 to 20,000 units.
SI018 eHangzhou / China Daily Robot maker Unitree Robotics to file IPO application in Q4
SI019 Global Times Unitree expected to submit IPO filing documents between October and December 2025: company
SI020 Forbes China’s Robot Maker Unitree Valued At $1.7 Billion In Series C Round
SI021 KrASIA Unitree Robotics develops personal robot dogs that can jog with you
SI022 People.cn / Xinhua China releases national standard system for humanoid robotics and embodied AI
SI023 Hill Dickinson Humanoid robots and the law - preparing for a new era of risk
SI024 Trade.gov China - U.S. Export Controls
SI025 CSRC disclosure platform 405
SI026 GitHub / Unitree Robotics Unitree Robotics
SI027 Unitree Robotics 宇树科技—全球四足机器人行业开创者
SI028 Unitree Robotics Universal humanoid robot H1_Bipedal Robot_Humanoid Intelligent Robot Company
SE001 Unitree Robotics Unitree Robotics homepage
SE002 Unitree Robotics About Unitree
SE003 Unitree Robotics News Center
SE004 Unitree Robotics Humanoid robot G1
SE005 Unitree Robotics Universal humanoid robot H1
SE006 Unitree Robotics Robot Dog Go2
SE007 Unitree Robotics Robot Dog B2
SE008 Unitree Robotics Robotic Arm Z1
SE009 Unitree Robotics Store Unitree G1
SE010 Unitree Robotics Store Unitree Go2
SE011 GitHub Unitree Robotics organization
SE012 GitHub unitree_sdk2
SE013 GitHub unitree_rl_gym
SE014 GitHub unitree_ros2
SE015 GitHub unifolm-world-model-action
SE016 GitHub unitree-app-templates
SE017 GitHub unitree_sim_isaaclab
SE018 Xinhua Unitree Robotics shares vision for China's robot revolution
SE019 Global Times Dancing robots at Spring Festival gala raise attention in US media regarding China's technological advancements
SE020 Robotics & Automation News Why China's new humanoid robot standards could change the industry
SE021 China Daily Government Service Hangzhou builds 'global home port' for robotics industry
SE022 Forbes Humanoid Robots: Here Are The 16 Leading Manufacturers
SE023 Unitree Robotics Official Open Source
SE024 Unitree Robotics A1 Motor - Stable and Efficient with Hardcore Strength
SE025 Correll Lab Testing the Field Capabilities of the Unitree Go-1
SE026 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors
SE027 RoboticsTomorrow Unitree Showcases Advanced High-Mobility Robots at CES 2025
SE028 Robotics & Automation News Unitree showcases advanced high-mobility robots at CES
SE029 Robotics Observer From Robot Dogs to Humanoids: The Unlikely Rise of Unitree Robotics
SU001 Unitree Robotics Unitree Robotics homepage
SU002 Unitree Robotics About Unitree
SU003 Unitree Robotics News Center
SU004 Unitree Robotics Robot Dog Go2
SU005 Unitree Robotics Robot Dog B2
SU006 Unitree Robotics Humanoid robot G1
SU007 Unitree Robotics Universal humanoid robot H1
SU008 Unitree Robotics Store Unitree Go2
SU009 Unitree Robotics Store Unitree G1
SU010 Xinhua Unitree Robotics shares vision for China's robot revolution
SU011 Global Times Dancing robots at Spring Festival gala raise attention in US media regarding China's technological advancements
SU012 AIbase Unitree Technology's Humanoid Robot Debuts at CCTV Spring Festival Gala with Performance 'Yang Bot'
SU013 The Wire China While Politicians Fret, Unitree is Selling Robots Across America
SU014 RoMI Lab Unitree Go1
SU015 Zenodo Enabling Navigation and Mission-based Control on a Low-cost Unitree Go1 Air Quadrupedal Robot
SU016 Correll Lab Testing the Field Capabilities of the Unitree Go-1
SU017 RoboticsTomorrow Unitree Showcases Advanced High-Mobility Robots at CES 2025
SU018 Robotics & Automation News Unitree showcases advanced high-mobility robots at CES
SU019 TVC Robotics Unitree Robotics Official Distributor | TVC Robotics USA & Canada
SU020 Drones Plus Robotics Unitree Dealer: Buy Unitree Robots with Business Solutions Support
SU021 Futurology Tech Futurology became authorized U.S. distributor for Unitree Robotics
SU022 RoboStore RoboStore | Official Partner of Unitree
SU023 US Robot Systems USRS | Official Distributor of Unitree
SU024 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More
SU025 robot.tv Unitree Robots Guide: G1, H1, H2, Go2, B2 Comparison
SU026 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors
SU027 GrabaRobot Unitree Robot Price: Complete Guide to Every Unitree Model in 2026
SR001 International Trade Administration China - U.S. Export Controls
SR002 Congressional Research Service U.S. Export Controls and China: Advanced Semiconductors
SR003 U.S. Government Accountability Office GAO-25-107386, Export Controls: Commerce Implemented Advanced Semiconductor Rules and Took Steps to Address Compliance Challenges
SR004 U.S.-China Economic and Security Review Commission U.S.-China Competition in Emerging Technologies
SR005 Congresswoman Elise Stefanik Stefanik, Cotton Introduce Bipartisan Bill to Propel America's Robotics Superiority, Protect U.S. National Security
SR006 IEEE Spectrum US Ban on Chinese Robots Could Reshape Supply Chains
SR007 Hill Dickinson Humanoid robots and the law - preparing for a new era of risk
SR008 Robotics & Automation News Why China’s new humanoid robot standards could change the industry
SR009 People.cn / Xinhua China releases national standard system for humanoid robotics and embodied AI
SR010 CGTN China releases national standards for humanoid robots and embodied AI
SR011 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors
SR012 The Wire China While Politicians Fret, Unitree is Selling Robots Across America - The Wire China
SR013 TechNode Unitree Robotics begins IPO prep, valued at $1.6 billion after Series C funding · TechNode
SR014 The Robot Report Unitree becomes a legged robot unicorn with Series C funding - The Robot Report
SR015 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More | Humanoid Index
SR016 Unitree Robotics Humanoid robot G1_Humanoid Robot Functions_Humanoid Robot Price
SR017 Unitree Robotics Universal humanoid robot H1_Bipedal Robot_Humanoid Intelligent Robot Company
SR018 Unitree Robotics Unitree G1
SR019 Unitree Robotics Unitree Go2
SR020 Unitree Robotics on GitHub Unitree Robotics
SR021 Unitree Robotics on GitHub GitHub - unitreerobotics/unitree_sdk2
SR022 Unitree Robotics on GitHub GitHub - unitreerobotics/unitree_rl_gym
SR023 Unitree Robotics on GitHub GitHub - unitreerobotics/unitree_ros2
SR024 China Daily Government Service Hangzhou builds 'global home port' for robotics industry
SR025 CGTN IDC report: China leads the global humanoid robot rise in 2025
SR026 IDC Worldwide Quadruped Robot Market Shares, 2024: China Firms Emerge Leaders as Embodied Intelligence Unlocks Diverse Industry Applications
SR027 IDC Worldwide Humanoid Robotics Market Analysis 2026: Breakthroughs, Commercialization, and Emerging Landscape
SR028 Goldman Sachs Research Humanoid robot: The AI accelerant
SR029 International Federation of Robotics New IFR position paper on humanoid robots published
SR030 Agility Robotics Humanoid Solutions | Agility
SR031 Boston Dynamics Atlas Humanoid Robot | Boston Dynamics
SR032 Figure Company | Figure
SR033 1X Technologies Introducing NEO Gamma
SR034 U.S. House Select Committee on the CCP Testimony of Kyle I. Chan, Ph.D. before the U.S. House Select Committee on Strategic Competition Between the Chinese Communist Party and the United States
SR035 Associated Press Trump administration bans import of new foreign-made routers, citing supply chain and security risks
SR036 Congressional Research Service U.S. Export Controls and China: Advanced Semiconductors
SR037 IEEE Spectrum US Ban on Chinese Robots Could Reshape Supply Chains
SV001 TechNode Unitree Robotics begins IPO prep, valued at $1.6 billion after Series C funding
SV002 The Robot Report Unitree becomes a legged robot unicorn with Series C funding
SV003 Humanoid Index Unitree Robotics: Funding, Valuation, Robot Specs & More
SV004 Unitree Robotics Unitree G1 humanoid robot specifications and price
SV005 Unitree Robotics About Unitree
SV006 Goldman Sachs Research Humanoid robot: The AI accelerant Should the hurdles of product design, use case, technology, affordability, and wide public acceptance be completely overcome, our analysts envision a market of up to US$154bn by 2035E in a blue-sky scenario.
SV007 CGTN IDC report: China leads the global humanoid robot rise in 2025
SV008 Figure Figure homepage
SV009 Figure Figure company page
SV010 Tech Market Briefs Figure AI IPO 2026: $39B Valuation, Risks & Bull Case
SV011 Humanoid Index Figure AI: Funding, Valuation, Robot Specs & More
SV012 Agility Robotics Humanoid Solutions
SV013 Humanoid Index Agility Robotics: Funding, Valuation, Robot Specs & More
SV014 1X Technologies NEO Home Robot
SV015 Humanoid Index 1X Technologies: Funding, Valuation, Robot Specs & More
SV016 Apptronik Apollo
SV017 Apptronik Apptronik closes over $935 million Series A with new $520 million extension round
SV018 US News / Reuters Humanoid Startup Apptronik Raises $520 Million With Backing From Google and Mercedes-Benz
SV019 U.S. SEC EDGAR EDGAR search results for Apptronik Form D filings
SV020 U.S. SEC EDGAR EDGAR search results for Figure AI Form D filings
SV021 AGIBOT AGIBOT homepage
SV023 Hong Kong Exchanges and Clearing HKEX quote page for UBTech Robotics (9880)
SV024 Boston Dynamics Atlas humanoid robot
SV025 Boston Dynamics Spot robot
SV026 The Wire China While Politicians Fret, Unitree is Selling Robots Across America
SV027 arXiv Cybersecurity AI: Humanoid Robots as Attack Vectors
SV028 Forbes Humanoid Robots: Here Are The 16 Leading Manufacturers
SV029 Xinhua Unitree Robotics shares vision for China’s robot revolution
SV030 CGTN / Xinhua China releases national standards for humanoid robots and embodied AI
SV031 China Daily Hangzhou builds global home port for robotics industry
SV032 IDC Worldwide Humanoid Robotics Market Analysis 2026
SV033 TechCrunch Apptronik, which makes humanoid robots, raises $350M as category heats up
SV034 Crunchbase News Amid Record Robotics Funding, Apptronik Raises $520M Series A Extension