初创公司尽调
尽调报告 robotics / hardware Series B 2026-05-18

Agibot

人形机器人——中国规模领先者,披露滞后者

Agibot 是可信的量产级人形机器人竞争者,Longcheer 工厂给出了少见的工业落地证据;但财务披露缺位、地缘约束加重,当前更适合观察而非买入。

封面要素

创立时间 01
Feb 2023 [CO001]
总部 02
Shanghai, China [CO002]
已披露融资 03
83.8 USDm [CO014]
最新公开估值锚点 04
960 USDm pre-money [CV004]
2025 年人形机器人出货量 05
5100 units [CV002]
2026 年 3 月前产量 06
10000 units [CV001]

公司概况

Agibot(AGIBOT Innovation (Shanghai) Technology Co., Ltd., 智元机器人)是一家总部位于上海的具身 AI 与人形机器人公司,成立于 2023 年。公司把机器人硬件、基础模型、部署软件和运营数据集合成全栈平台,覆盖工业操作、物流、酒店、零售和互动服务等场景。公开证据显示,它在中国的生产规模没有明显对手,并在 Longcheer 有一个有说服力的工厂部署;但公司仍是私有、财务不透明的 Series B 企业。

官网
www.agibot.com
成立时间
2023-02-27
创始人
Deng Taihua, Peng Zhihui
创立地点
Shanghai, China
总部
Shanghai, China
产品
通用具身机器人组合,覆盖 A2 人形系列、G1/G2 工业机器人、X1/X2 开发者与互动机器人、灵巧手、Genie Studio 部署栈,以及面向具身 AI 的 Agibot World / 配套模型工具。
客户
寻求劳动力自动化或互动机器人的制造、物流、酒店、零售服务、公共互动和企业运营方。
商业模式
面向企业与商业运营方,组合硬件销售、项目部署包,以及早期机器人即服务 / 部署平台变现。
阶段
Series B
融资情况
2023 年以来完成 10 多轮融资;最近公开披露的一轮是 2025 年 8 月 LG Electronics 和 Mirae Asset 的战略投资,Pandaily 报道估值超过 CNY 7 billion。
[CO001, CO002, CO003, CO005, CO007, CO008, CO014, CV013]

执行摘要

主要优势

  • 行业内披露最高的人形机器人产量:截至 2026 年 3 月已生产 10,000 台
  • Longcheer 工厂部署,是人形机器人领域公开工业验证中最强的证据之一
  • 全栈平台把硬件、数据集、模型和部署工具串在一起,而不是只卖单台机器人
  • BYD、Tencent、JD.com、LG、Mirae 等战略投资者,可带来供应链和渠道杠杆
  • 中国制造规模和标准推进,为国内商业化打开顺风

主要风险

  • 收入、毛利率、烧钱速度和合同经济性仍未披露
  • 截至本报告日期,Longcheer 是唯一公开确认的量产级客户
  • 美国及更广泛的地缘限制,可能压制西方市场扩张和采购意愿
  • 现场可靠性、安全支持和长期维护,还没有在企业级机群规模上跑出证据
  • Unitree 以及资金更厚的西方同行带来的价格压力,可能在工业 ARR 成熟前挤压利润率

未决问题

  • 已部署机群的工业 ARR、毛利率和回本经济性
  • 第二个具名制造业客户,以及更广泛的多客户量产证据
  • 真实累计融资额、投后估值和投资者优先权堆叠
  • 目前已披露 12 个月窗口之外的长期软件 / 安全支持承诺
  • 跨境数据治理、出口管制敞口和客户数据本地化立场

目录

Chapter 01

01公司概况

1.1 身份、使命与公司结构

AGIBOT Innovation (Shanghai) Technology Co., Ltd.——公开品牌为 AGIBOT,中文名为 智元机器人(Zhiyuan Jiqiren)——是一家总部位于上海的机器人与具身 AI 公司,成立于 2023 年 2 月。公司宣称的使命,是靠 AI 与机器人技术的融合创新,打造世界领先的通用具身机器人及其应用生态。AGIBOT 采用垂直整合的硬件与 AI 平台模式:自己设计硬件,用自有数据训练基础模型,并把机器人直接部署给制造、商业和面向消费者场景中的企业客户。 公司围绕三条机器人系列组织产品组合:Yuanzheng(远征)覆盖全尺寸与轮式人形机器人;Lingxi(灵犀)覆盖模块化与表演机器人;Genie(精灵)覆盖工厂地面专用平台。到 2026 年 1 月,AGIBOT 已公开称累计交付 5,000 台机器人;到 2026 年 3 月,产量升至 10,000 台——据 HumanoidIndex 引用的 Omdia 数据,这一生产提速让 AGIBOT 在 2025 年人形机器人出货量上位列全球第一。公司的 A2 Ultra 旗舰机型已获得四项主要国际认证(CR, CE-MD, CE-RED, FCC),可在美国、欧盟和中国商业销售。截至本报告日期,AGIBOT 产品已进入多个国家和地区,线上商店中 X2 娱乐机器人标价 $24,240,A2 Lite 标价 $44,560 USD。 公司治理仍未公开。审阅的公开材料中,AGIBOT 没有发布公司章程、董事会构成或股权结构表。公司处于后期 Series B / 战略投资阶段(私有、未披露画像)。收入、年经常性收入(ARR)和关键财务指标仍未公开。[CO001, CO002, CO003, CO004, CO005, CO006]

快照 KPI 表
指标数值 / 状态日期置信度缺口 / 尽调问题
总部所在地中国上海2026-05无——官方已确认
成立时间February 20232023-02无——官方已确认
阶段B 轮 / 战略轮(未上市,未披露)2025-08具体轮次结构和条款未披露
累计融资(已披露)~$83.8M USD2025疑似存在多轮未披露融资;LG-Mirae 金额未公开
据报道估值$6.4B USD(Tracxn)/ ¥7B+(Pandaily,较早报道)2025–2026来源冲突;两者都不是一手 / 经审计披露
累计生产台数截至 2026 年 3 月为 10,000 台2026-03生产数量,未经商业部署验证
员工数500+(估计)2026HumanoidIndex 估计;无官方披露
收入 / ARR未披露2026私营公司;尽调中要求提供财务数据
国际认证CR、CE-MD、CE-RED、FCC(A2 Ultra 认证)2025仅限 A2 Ultra;其他产品线未确认
主要部署用例工厂(G2)、商业 / 活动(A2 Ultra、X2)2026消费者部署收入尚未确认

估值数据来自第三方数据库和新闻报道,并非经审计披露。融资数据只代表公开确认的轮次;企业登记数据表明还存在其他轮次。置信度反映底层来源的独立性和层级。

[CO001, CO002, CO014, CO021, CO022, CO034]
FO002: 公司快照逻辑

截至 2026 年 5 月,AGIBOT 的身份、产品线、生产飞轮、资本、客户和监管依赖如何连接。

[CO004, CO005, CO034, CO040, CO046, CO047]

1.2 领导层、治理与关键人物依赖

AGIBOT 由两名核心人物共同创立并领导。Deng Taihua(邓泰华)担任 CEO,也是控制性受益所有人;中国工商登记数据(Qichacha)显示,截至 2025 年 7 月,他作为实际控制人持有约 33% 的受益所有权。Peng Zhihui(彭志辉)是联合创始人兼 CTO,也是公司全栈机器人平台的技术架构师;Tracxn 的公司资料专门把他与 Deng 并列列出,并且在 2026 年 3 月 10,000 台里程碑活动中,他以「联合创始人兼 CTO」身份出席。审阅材料尚未独立确认 Peng Zhihui 是否就是中国科技内容创作者「稚晖君」(Zhi Hui Jun),尽管网络上广泛流传;这一点仍是待解问题,不是既定事实。 已确认的其他高管包括高级副总裁 Chuang Wang(王闯),他共同发布了第 10,000 台里程碑;以及 Dr. Yao Maoqing(姚茂庆),合伙人、高级副总裁兼具身业务单元总裁,他在 CES 2026 以官方发言人身份代表 AGIBOT。官方称创始团队由「全球科技龙头核心高管和顶尖 AI 科学家」组成,但审阅的公开材料没有确认个人过往雇主和详细履历。 董事会构成、投资人席位和治理条款均未公开披露。公司自创立以来未宣布任何重大领导层变化。关键人物集中度偏高:两位联合创始人分别占据最高经营与技术岗位,公司也未披露继任人或分布式领导架构。截至本报告日期,公司似乎没有发布任何 IPO、双重上市或 SPAC 的治理里程碑。[CO007, CO008, CO009, CO010, CO011, CO012]

领导层与创始人表
人员角色背景(已确认)创始人状态关键人风险
Deng Taihua(邓泰华)CEO 兼联合创始人实际控制人;据 Qichacha 2025 年 7 月数据,约 33% 受益所有权;曾任职于全球科技公司(细节未确认)创始人极高——CEO 且最大股东;企业文件中唯一高管面孔
Peng Zhihui(彭志辉)联合创始人兼 CTOTracxn 将其列为联合创始人;10,000 台里程碑中被称为 CTO;可能(未确认)是中国知名科技人物“稚晖君”创始人高——唯一具名技术负责人;离职会威胁产品路线图
Chuang Wang(王闯)高级副总裁共同出席第 10,000 台里程碑活动;承担运营领导角色非创始人中——高级运营负责人;若竞争对手激进挖角,存在留任风险
Yao Maoqing(姚茂庆)合伙人、SVP、具身业务单元总裁CES 2026 上被列为 AGIBOT 官方发言人;负责商业化策略非创始人中——主要收入业务单元总裁;对美国和国际扩张重要

所有人的前雇主细节均来自公司说法或独立来源未确认。董事会席位和股权权利等治理结构未公开披露。

[CO007, CO008, CO010, CO011, CO012]

1.3 融资历史、投资人和资本位置

AGIBOT 自 2023 年 2 月创立以来已完成 10 多轮融资。按 Tracxn 口径,已披露资金约 $83.8 million,但这个数字很可能偏低:中国工商登记数据(Qichacha)列出 15 个以上不同股东载体,其中包括多个匿名有限合伙,与更多未披露融资轮相符;估值来源之间也存在实质差异。Pandaily 报道估值「超过 7 billion yuan」(按报道时汇率约 $960 million),而 Tracxn 较新的数据库条目把估值列为 $6.4 billion USD——这一差异更可能反映计量日期不同,而非真实分歧。 投资人基础横跨中国工业与科技公司、国内 VC 和国际投资者。已确认的产业投资人包括 BYD(Qichacha 显示 A+ 轮,2023 年 8 月)、Tencent Holdings(Qichacha 显示 B 轮,2025 年 3 月)、BAIC Group、SAIC Motor、JD.com 和 TCL Technology(据 Yicai Global 报道)。HumanoidIndex 将 CATL 列为重要支持方。Gaocheng Ventures(与 GGV/Gaocheng 相关)被识别为早期投资人。最近披露的一轮融资来自 Mirae Asset-LG Electronics New Growth Fund 的战略投资,于 2025 年 8 月完成;这是 LG Electronics 首次投资具身智能领域。LG-Mirae 轮的金额和隐含估值未披露。 收入和盈利数据完全未公开。审阅来源没有报道债务、信贷额度或老股交易。公司的资本位置意味着它正依靠机构股权烧钱,符合一家已规模化运营但仍处于收入前或早期收入阶段的硬件 AI 公司特征。员工数未获官方披露;HumanoidIndex 报道为 500+ 人。[CO014, CO015, CO016, CO017, CO018, CO019]

利益相关方 / 投资人图谱
利益相关方角色 / 轮次层级 / 战略重要性战略角度尽调问题
Deng Taihua(创始人持股工具)创始股东;实际控制人(约 33%)关键;最大受益所有人创始人控制;激励一致确认归属安排、创始人锁定和领售权
Gaocheng Ventures(高瓴创投)早期投资人(天使 / 种子,2023)高;GGV 关联;技术投资逻辑强深度中国科技生态入口;后续跟投能力投资金额和轮次结构?
BYD(比亚迪)A+ 轮投资人(据 Qichacha 为 2023 年 8 月)高;全球 EV 龙头;硬件供应链匹配汽车制造用例;电池 / 执行器供应链协同是否附带任何商业机器人采购或供应协议?
BAIC Group(北汽)战略投资人高;中国国资背景车企工业部署与制造自动化渠道持股规模和是否有独家部署协议?
SAIC Motor(上汽)战略投资人高;中国最大车企智能制造;向 SAIC 工厂供应机器人已承诺商业部署量?
JD.com(京东)战略投资人高;中国领先电商和物流运营商物流自动化和仓储机器人部署JD 场地中活跃机器人部署数量?
Tencent HoldingsB 轮投资人(据 Qichacha 为 2025 年 3 月)高;中国领先互联网集团平台 / 生态集成;AI 模型部署战略产品集成,还是纯财务投资?
TCL Technology战略投资人中高;消费电子 OEM消费机器人渠道;显示与硬件集成是否有联合开发或制造安排?
LG Electronics / Mirae Asset(New Growth Fund 基金)战略投资(2025 年 8 月;金额未披露)高;首个非中国机构投资人;标志国际验证LG Electronics 进入具身 AI;韩国工业与消费渠道确认投资金额、董事会观察员权利和商业协议
CATL战略投资人(HumanoidIndex 引用)高;全球最大 EV 电池制造商电池与储能供应链;执行器整合投资金额和日期?是否有独家电池供应协议?

大多数轮次的投资金额未披露。企业登记数据反映的是已报告持股结构,但可能未覆盖所有间接或代持股权。CATL 确认依赖二级数据库(HumanoidIndex);未找到直接一手来源。

[CO014, CO015, CO016, CO017, CO018, CO019]

1.4 产品组合、部署与商业证据

AGIBOT 商业成熟度最高的产品是 A2 Ultra,公司称其为行业首个大规模商业部署的全尺寸人形机器人。A2 Ultra 获得四项国际认证(CR, CE-MD, CE-RED, FCC),完成全球首次 24 小时户外行走直播,并以跨省行走 106.286 kilometres 创下 Guinness World Record,展示了户外移动耐力。它已部署在 20 多家领先企业。A2-W 是轮式移动操作变体,把底盘、机械臂和感知系统整合起来,可在一天内完成部署。 G 系列(Genie)面向工厂环境。G1 拥有 26 个自由度,可在工业、商业和家庭场景中运行。G2 是公司的主力工厂部署单元。2026 年 4 月,AGIBOT 宣布多台 G2 机器人已整合进 Longcheer Technology 位于江西南昌的平板电脑产线——公司称这是全球首个具身 AI 在消费电子精密制造中的大规模落地。AGIBOT 报告称,G2 在 8 小时连续运行中完成 2,283 个任务且零错误,双方计划到 2026 年 Q3 将 Longcheer 产线上的 G2 扩至 100 台。 X 系列(Lingxi)包括 X1——一个有 29 个关节的开源硬件和软件平台——以及 X2,一个半尺寸娱乐和商业表演机器人。Genie Studio 平台提供零代码机器人部署编排;AIMA 开放生态在 APC 2026 发布,旨在降低第三方开发者的集成门槛。在 APC 2026(2026 年 4 月),AGIBOT 还发布五个新硬件平台,包括 A3 全尺寸人形机器人、D2 Max 全地形四足机器人和 G2 Air 移动操作机器人。软件侧,公司发布了「一体三智」架构下的八个基础 AI 模型。[CO024, CO025, CO026, CO027, CO030, CO031]

FO003: 快照 KPI

截至 2026 年 5 月,AGIBOT 的成熟度和牵引力 KPI,覆盖生产规模、产品认证、商业定价和已披露投资。

估值数字来自第三方数据库,可能无法反映最新融资事件。台数为公司报告产量,不是经独立审计的出货记录。

[CO025, CO033, CO034, CO027, CO028, CO029]

1.5 里程碑时间线与战略轨迹

AGIBOT 从创立走向全球大规模部署,只用了三年多。公司于 2023 年 2 月在上海成立,数月内完成天使轮和早期 VC 融资,并在 2025 年 1 月交付第 1,000 台具身机器人——公司称这是通用具身机器人量产的行业首例里程碑。2026 年 1 月,AGIBOT 在拉斯维加斯 CES 完成美国市场首秀,宣布累计交付机器人 5,000 台。到 2026 年 3 月,累计产量达到 10,000 台,最后 5,000 台大约用三个月生产完成——相较上一阶段,生产速度大约提速 4×。 地缘政治上,AGIBOT 的国际推进积极且高调:2026 年 3 月,AGIBOT A2 机器人在巴塞罗那 MWC 2026 为西班牙国王 Felipe VI 提供引导;2025 年早些时候,Xi Jinping 在上海考察期间参观了公司的机器人。与此同时,公司面临实质监管逆风:美国参议员 Tom Cotton 和 Chuck Schumer 于 2026 年 3 月 26 日提出并获得两党支持的 American Security Robotics Act,将禁止美国联邦机构购买或运营中国制造的人形机器人。如果法案通过,将直接限制 AGIBOT 可触达的美国政府市场。 2026 年 4–5 月,战略扩张继续推进,包括 Genting Malaysia MOU(2026 年 4 月 17 日)、APC 2026 Shanghai Partner Conference,以及 First Hong Kong Embodied AI Industry Summit(2026 年 5 月 12 日)。公司宣布 2026 年为「部署元年」,显示重心正从生产规模转向商业 ROI。[CO032, CO033, CO034, CO035, CO042, CO043]

里程碑表
日期事件类型金额 / 估值 / 状态参与方 / 合作伙伴影响
2023-02AGIBOT 在上海成立成立公司成立Deng Taihua、Peng Zhihui;创始团队来自全球科技龙头确立公司身份;中国人形机器人竞赛开始计时
2023-04Gaocheng Ventures 天使 / 种子投资融资金额未披露Gaocheng Ventures(GGV 关联)获得首个机构背书;早期验证机器人投资逻辑
2023-08BYD 参与 A+ 轮融资金额未披露BYD、其他投资人关键汽车产业战略支持方;打开汽车部署渠道
2024B 轮融资融资披露总额约 $83.8M(Tracxn,跨全部轮次)Tencent(据登记资料为 2025 年 3 月)、其他投资人拿到规模化资本;投资人基础扩展至互联网巨头
2025-01第 1,000 台通用具身机器人下线规模行业里程碑(公司声称首个)AGIBOT 生产团队展示量产能力;以四位数产量验证供应链
2025习近平在上海参观 AGIBOT 机器人规模背书事件中国国家主席习近平高规格政治背书;强化国家战略重要性
2025-08LG Electronics 和 Mirae Asset 战略投资融资金额未披露LG Electronics、Mirae Asset-LG Electronics New Growth Fund(基金)首个非中国机构投资人;标志国际资本信心
2026-01CES 2026 美国市场首秀;5,000 台机器人出货里程碑产品5,000 台累计出货AGIBOT;美国媒体和贸易社群正式进入美国市场;CES 验证生产规模声明
2026-03-02MWC Barcelona;A2 为西班牙国王 Felipe VI 导览合作国际展示西班牙国王 Felipe VI;AGIBOT 位于 MWC Hall 6王室级品牌能见度;展示国际商业可信度
2026-03第 10,000 台人形机器人下线规模10,000 台累计生产AGIBOT;Peng Zhihui 和 Chuang Wang 出席里程碑活动迄今最快人形机器人爬坡;约 3 个月从 5K 到 10K 显示工厂规模
2026-03-26美国 American Security Robotics Act 提出反向法案提出(尚未通过)参议员 Tom Cotton(R-AK)和 Chuck Schumer(D-NY)美国两党法案将禁止联邦使用中国人形机器人;直接压制 AGIBOT 的美国政府渠道
2026-04-14G2 部署于 Longcheer Technology 平板工厂产品2,283 项任务,8 小时运行零错误AGIBOT G2、Longcheer Technology(江西南昌)全球首个声称在 3C 精密制造中大规模部署的具身 AI
2026-04-17APC 2026 上海;宣布“Deployment Year One”产品五个新硬件平台;八个 AI 模型AGIBOT;全球合作伙伴生态战略重心从生产规模转向商业 ROI;AIMA 生态启动
2026-04-17签署 Genting Malaysia MOU合作MOU(非约束性)AGIBOT、Genting Malaysia Bhd首个东南亚酒店 / 文旅合作伙伴;从工业用例向外多元化
2026-05-12首届香港具身 AI 行业峰会和 APC Hong Kong合作峰会和合作伙伴大会AGIBOT;Hong Kong Chinese Enterprises Association;香港行政长官 John Lee强化大中华枢纽策略;将 AGIBOT 定位为具身 AI 生态锚点

大多数轮次的融资金额未披露。2023 年和 2024 年融资事件日期为根据企业登记元数据推定的近似值。美国法案状态仅反映提出日期;截至报告日期,是否通过尚未确认。

[CO001, CO015, CO016, CO017, CO018, CO032]
FO001: 公司里程碑时间线

2023 年 2 月成立至 2026 年 5 月香港 APC 峰会的关键公司事件,覆盖融资、产品、规模、合作和不利事件。

[CO001, CO032, CO033, CO034, CO042, CO043]

1.6 图表证据

Chapter 02

02市场分析

2.1 市场边界、替代品与相邻支出

人形与具身 AI 机器人市场,指商业化生产、具有人形或半人形形态的机器人,包括双足设计(Agibot A2、A3、G2 系列)和紧凑型轮式躯干平台;它们用灵巧操作、移动行走和多模态交互,在为人类建造的环境中运行。不纳入本市场的包括:执行单点重复任务的固定臂工业机器人(Fanuc、KUKA、ABB);没有操作能力的自主移动机器人(AMR);以及没有物理具身形态的纯软件 AI 系统。人形形态的竞争优势在于,它能在为人类工人建造的空间中运行——使用现有门、货架、工具和布局——而不需要重建物理环境。 主要目标细分中的现状替代品包括:(1)人类劳动力——制造、物流和服务环境中最主要的替代选择;(2)传统固定臂工业机器人,用于结构化高速任务,这类任务中人形形态没有优势;(3)协作机器人(如 Universal Robots UR 系列),在固定位置运行且没有移动能力;(4)专用任务机器人(洗地机、配送机器人、条码扫描器),解决狭窄、结构化工作流。关键切换触发点,是从结构化单任务自动化转向灵活、多产品、多工位工作流,而后者超出了固定自动化的能力边界。 本分析纳入的支出包括机器人硬件购买或租赁、嵌入式 AI 软件许可、部署集成服务、训练数据和售后维护。不纳入的支出包括设施基础设施重建(传送带、货架重新设计)、劳动力再培训项目,以及相邻的非人形自动化品类。能力成熟后,人形机器人可能扩展进入的相邻市场包括工业外骨骼、协作机器人手臂、AI 驱动配送机器人和无人机物流平台。[CM001, CM002, CM003, CM004]

市场定义:纳入支出、排除支出与替代方案
细分市场 / 类别纳入支出排除支出主要替代方案买方 / 付款方与 Agibot 的相关性
工业制造(3C、汽车、半导体)机器人硬件、AI / 软件许可、集成、维护固定机械臂产线、工厂输送线重新设计固定机械臂工业机器人;协作机器人;人工装配工制造副总裁 / CFO(capex 预算)近期核心细分市场;G2 已部署于 Longcheer、HT-Tech、Intel
物流与履约(电商、包裹处理)机器人硬件 / 租赁、分拣软件、车队管理输送线升级、仓库布局调整AMR;人工拣选员;条码扫描工作流运营副总裁 / CFO(取决于租赁形式,opex 或 capex)Damon 和 Xiaonan 已有早期进展;不规则物品处理具优势
零售服务与酒店 / 文旅(百货、餐厅、交通)机器人硬件 / 租赁、交互软件、内容管理门店动线重设计、数字标牌基础设施人工服务人员;基于自助终端 / 屏幕的自助服务品牌 / 市场 + 运营预算;GM / COOAnta、Haidilao、Guangzhou Metro 试点显示双重 ROI
公共交互(地铁、政府服务中心)机器人租赁或采购、与运营系统集成人员配置、柜台重设计人工坐席;数字自助终端;应用政府运营预算 / 城市采购Guangzhou Metro 部署;Whale Cloud 管线
研究与教育(大学、数据工厂)机器人采购、数据许可、仿真软件实验室基础设施、安全认证人类研究员;遥操作演示设备高校资本预算;科研经费主导 2025 年出货;AgiBot-World 开放数据集

细分边界综合自 Agibot 的七个标准化解决方案(APC 2026 新闻稿)、IDC 关于人形机器人商业化的博客,以及 Goldman Sachs 中国调研(2026 年 1 月)。纳入 / 排除支出反映行业实践,并非 Agibot 披露定价。买方画像根据部署案例研究和分析师细分定义推断。

[CM001, CM002, CM003, CM019, CM025]
FM001: 市场规模金字塔:TAM、SAM 和受限 SOM

从全球 TAM(Goldman Sachs 2035 基准情景)到 Agibot 2026–2028 年估计 SOM 的四层规模金字塔。 除非另有说明,数值单位为十亿美元。

TAM 来自 Goldman Sachs 发布的基准情景(2035 年 $38B)。SAM-1 由 MarketsandMarkets 中国($2.80B) 和美国($4.60B)2030 年点估计相加。SAM-2 是分析师推断的 Agibot 可触达垂直场景估计;并非任何单一分析师发布。 SOM 是根据公开机队和租赁数据作出的粗略推断;Agibot 未披露收入或 SOM 预测。

[CM006, CM007, CM008, CM009, CM010]

2.2 市场规模测算:多重视角、矛盾与受限估计

人形机器人市场有多组独立估计,但彼此相差数个数量级,原因是范围定义不兼容。Goldman Sachs 在基准情景中估计,全球人形机器人市场在 10-15 年内至少达到 $6 billion(来自 2024 年报告),被广泛引用的 2035 年数字为 $38 billion;若硬件成本、安全和接受度障碍被克服,蓝天上限为 $154 billion。MarketsandMarkets(2025 年 4 月)更细:全球硬件市场 2025 年为 $2.92 billion,到 2030 年增至 $15.26 billion,CAGR 为 39.2%;中国市场从 $0.40 billion 增至 $2.80 billion,CAGR 为 47.6%;美国市场从 $0.86 billion 增至 $4.60 billion,CAGR 为 39.9%。 IDC 的出货量预测换了一个视角:到 2030 年全球出货量超过 510,000 台,CAGR 约 95%——如果采用 $20,000-$50,000 的合理 ASP,意味着到 2030 年收入底线为 $10-25 billion;这个结果与 MarketsandMarkets 大体一致,但方法不同。Goldman Sachs 在 2026 年 1 月对中国制造商的调研中,把近期预期修正为 2025 年全球出货 15,000-20,000 台,2026-2027 年目标处于数千至数万台区间,说明市场仍处在早期商业化阶段,而非大规模普及。 Goldman Sachs 的 $38B 硬件收入估计与 ARK Invest 的 $24 trillion 经济价值估计之间有 600× 差距,这不是对硬件市场的分歧,而是 ARK 衡量的是如果人形机器人替代 $40+ trillion 年度全球人类劳动力市场的一部分,可能带来的劳动力替代经济影响。这里同时保留两组估计,因为它们代表市场概念上的上下界。基于当前机队规模、ASP 和新兴租赁收入,Agibot 2026-2028 年受单位经济约束的「SOM 底线」估计为 $150-500 million,但公开数据无法验证。Grand View Research 和多家其他分析机构也发布了覆盖同一机会的人形机器人市场分析,进一步说明该赛道已是一个成熟研究垂直,拥有广泛机构兴趣;但其具体 2035 年数字在本轮研究窗口内无法取得。[CM006, CM007, CM008, CM009, CM010, CM011]

市场规模视角表:公开估计与方法论
发布方发布年份地区市场价值 / 数量CAGR时间范围方法论说明置信度局限
Goldman Sachs Research2024全球$6B(基准)– $154B(蓝天)未披露10–15 年;2035分析师情景建模;制造业劳动力缺口 + 老年护理需求情景区间很宽;$38B 是常被引用的中点;取决于成本 / 采用门槛
MarketsandMarkets2025全球$2.92B (2025) → $15.26B (2030)39.2%2025–2030按组件、传感器和执行器类型自下而上测算硬件收入仅硬件收入;不含软件和服务;2025 年 4 月版本
MarketsandMarkets2026中国$0.40B (2025) → $2.80B (2030)47.6%2025–2030全球报告中的中国特定细分市场中国子市场;不含软件;发布于 2026 年 5 月
MarketsandMarkets2025美国$857.9M (2025) → $4.6B (2030)39.9%2025–2030全球报告中的美国特定细分市场仅美国子市场;来自更宽口径的硬件细分
IDC2026全球2030 年 510,000+ 台~95%(台数 CAGR)至 2030 年出货量预测;IDC 2026 年 5 月商业化博客按台数口径;隐含收入高度取决于 ASP 假设
Goldman Sachs(中国调研)2026全球(中国主导)2025 年出货 15,000–20,000 台2026–2027 年数倍增长近期2026 年 1 月对 8 家中国人形机器人公司的调研;出货量获第三方数据印证中国公司调研;仅覆盖近期;出货目标偏愿景化
ARK Invest2025全球$24T 经济价值(劳动力替代情景)不可比2030+自上而下的劳动力市场替代测算;替代 $40T+ 人类劳动力中的一部分与硬件收入估算口径不匹配;极端情景;不是硬件收入预测

所有收入数据均为美元,除非另有说明。Goldman Sachs 区间覆盖基准底线($6B)到乐观情景($154B);媒体常引述的 $38B 是分析师中位情景。MarketsandMarkets 数据来自公开摘要页(全球报告发布于 2025 年 4 月,中国报告发布于 2026 年 5 月)。IDC 数字引自 IDC 的 Humanoid Robotics Commercialization 2026 博客。ARK Invest 数字指劳动力替代带来的经济价值,不是硬件收入,不能与其他行直接比较。Grand View Research 发布了人形机器人市场报告;研究窗口内未能获取详细数据。

[CM006, CM007, CM008, CM009, CM010, CM011]
FM002: 市场估计区间:相互矛盾的分析师预测

五个不同市场口径的低、基准、高区间,单位为十亿美元;各估计的时间跨度不同(2030 vs 2035),范围定义也不同(硬件收入 vs 经济价值)。 为便于比较,所有数值均为十亿美元;ARK Invest 数字在注释中转换为万亿美元。

所有数值单位为十亿美元。Goldman Sachs 区间覆盖已发布基准情景下限($6B)到蓝天情景($154B); 中位反映媒体报道中常引用的 $38B 估计(Forbes 2025)。MarketsandMarkets 低 / 高区间按已发布点估计上下 ±20–25% 置信区间估算;已发布数值为中点。IDC 区间将 510,000 台预测按 ASP $20K(低)、$30K(中)、$50K(高)换算为隐含收入。 ARK Invest 的 $24 trillion 数字未纳入该图,因为它衡量劳动力替代的经济价值,而非硬件 / 软件收入,会扭曲尺度。 时间跨度不同:Goldman Sachs 目标为 2035 年;其他为 2030 年。

[CM006, CM007, CM009, CM010, CM038]

2.3 买方与用户分层:预算所有者和采用触发点

Agibot 提出了七套标准化生产力解决方案,面向不同买方细分:产线上下料;工业物料搬运;物流分拣与履约;导览和零售辅助;零售服务站;酒店;以及研究或数据采集。这种多细分打法反映了市场当前阶段——没有单一垂直行业达到足以支撑单一细分策略的部署密度。 工业制造买方(EMS 组装商、3C 电子制造商、汽车 OEM 和半导体晶圆厂)是近期确定性最高的细分。预算所有者通常是 VP Manufacturing 或 CFO 级别,采购资金来自资本设备预算。采用触发点,是多型号、小批量生产需求超过固定自动化能力:Longcheer 的平板电脑产线、HT-Tech 芯片搬运(最高每日 1 million 颗芯片)和 Intel 的部署(每日 100,000 颗芯片)展示了 G2 在精密制造中的价值。物流和分拣买方(电商履约、包裹处理)被劳动力替代经济性吸引:Agibot 报告称,在 Damon 达到人类吞吐量的 70%,在 Xiaonan Intelligence 降低 50% 劳动力成本,不过这些仍是早期部署。 零售和酒店买方(百货商店、餐厅、交通运营商)呈现不同买方画像:品牌和客户体验团队与运营团队共同掌握预算,采用触发点不是单纯降本,而是提升客户互动。Agibot 报告称 Anta 门店客流增加 20%,Guangzhou Metro 运营成本下降 13%,说明该细分有双重价值主张。研究与教育机构(大学、数据工厂)是 2025 年主导买方,按 Goldman Sachs 口径贡献了不成比例的出货份额;这些买方性能要求较低,在产量爬坡期间带来早期收入。 Agibot 的全球租赁模式按每台机器人每日 €899 收费,降低了酒店和活动场景的资本开支门槛,让无法承诺购买硬件的买方也能试用。Robot-as-a-Service 模式目前是 17 国部署版图中西方买方的入口。[CM016, CM017, CM018, CM019, CM020, CM021]

买方与用户细分图谱
细分领域买方 / 决策者用户付款方预算负责人采用触发因素Agibot 证据
工业制造制造副总裁;工厂经理;CFO装配工人;产线主管;QA 员工资本设备预算(capex)CFO / 制造副总裁多型号生产压力;柔性换线需求;用工短缺Longcheer(2,283 项任务 / 8 hr,零错误);HT-Tech(1M 芯片/天);Intel(100K 芯片/天)
物流与履约运营副总裁;电商物流总监仓库拣货员;分拣员;班次主管运营支出或租赁型资本开支运营副总裁 / CFO人力吞吐缺口;非规则 SKU 处理;24/7 需求Damon(达到人工吞吐量 70%,二次通过率 99%);Xiaonan(劳动力成本降低 50%)
零售与酒店服务品牌/营销总监;COO;总经理门店员工;礼宾人员;客人营销 + 运营预算COO / 品牌总监客户互动提升;服务一致性;差异化Anta(客流提升 20%);Haidilao(排队互动);Genting Malaysia MOU
公共交互与政府服务城市政府采购部门;交通主管机构乘客;市民;运营人员政府资本/运营预算城市运营负责人 / 采购降本;24 小时服务可用;人员短缺Guangzhou Metro(成本降低 13%,80% 咨询自助服务)
研究与数据采集大学实验室主任;AI 训练数据公司 CTO研究人员;遥操作数据标注员学术资本预算或 VC 支持的研发预算PI / CTO开放数据飞轮;AI 模型训练;性能基准测试AgiBot-World 开放数据集;Goldman Sachs 调研显示研究与数据采集为 2025 年主导买方细分

买方画像综合自 Agibot 的 APC 2026 大会材料、官方新闻稿和 Goldman Sachs 2026 年 1 月中国调研。证据数据点来自公司声称;ROI 指标尚无独立第三方审计。预算负责人分类根据资本设备典型企业采购模式推断。

[CM019, CM020, CM021, CM022, CM023, CM024]
FM003: 买家与细分市场地图

将五类买家细分映射到用户画像、付款方类型、预算权限和主要采用触发因素。用于推动 Agibot 七方案组合的买家优先级和上市顺序。

[CM019, CM020, CM022, CM023, CM032]

2.4 增长驱动、采用约束与中国市场优势

最主要的结构性需求驱动来自全球制造业劳动力短缺,中国、美国和日本最为突出。IFR 的 World Robotics 2025 Report 显示,中国制造业约有 2 million 台工业机器人在运行,是日本运行存量的 4.5 倍,并贡献全球年度机器人装机量的 54%。2026 年 5 月发布的中国第十五个五年规划(2026-2030)把机器人和具身 AI 明确放在国家产业战略中心,瞄准 AI 在实体产业中的应用;国企采购与政策同向,让 Agibot 这类本土厂商获得系统性采购优势。 中国供应链的结构优势正在复利:国内机器人装机中,中国本土供应商份额从 2020 年的 30% 提升到 2024 年的 57%。中国工信部(MIIT)于 2026 年 2 月发布全球首个人形机器人国家标准体系,覆盖安全、类脑计算、部件规格、系统集成、应用场景和伦理。这给中国公司带来监管先发优势——它们的产品为自己参与起草的标准而设计,而西方进入者必须反向工程合规要求。 采用约束实质且多维。资本强度仍是门槛:每台 $10,000-$100,000+ 的硬件成本,需要相对于灵活人力给出正 NPV 论证,尤其是在劳动力充足的市场。Goldman Sachs 指出,生产一致性和多阶段测试流程是 2026-2027 年扩产目标的近期障碍。IFR 明确提醒,人形机器人作为通用工厂助手,在中短期内不会大规模普及——在高速精密任务中,传统工业机器人仍有优势,因为「形态服从功能」规则不利于人形形态。安全和责任问题也是近距离人机环境部署的门槛;工业机器人事故已经体现这一风险,而 MIIT 2026 年标准只能部分应对。 对 Agibot 来说,最不利且重大的约束是美国地缘政治风险。2026 年 3 月,美国参议员 Cotton 和 Schumer 跨党提出 American Security Robotics Act,援引数据传输风险和国家安全担忧,拟禁止联邦政府购买或使用中国公司制造的人形机器人。如果立法生效,将在美国政府采购、乃至潜在的盟国防务和基础设施采购中,在中国来源机器人与西方来源机器人之间形成正式市场分裂——在主权场景中直接限制 Agibot 的美国可触达市场,同时私人部门仍可进入。[CM026, CM027, CM028, CM029, CM030, CM031]

增长驱动因素与采用约束
因素方向时间幅度对 Agibot 的含义尽调追问
全球制造业用工短缺(中国、美国、日本)驱动因素现在 — 2032印证工业细分需求;Agibot G2 直接覆盖 3C/汽车装配缺口按地区量化未满足的制造业用工需求;映射到 Agibot 可服务机器人队列
中国“十五五”规划:机器人被列为国家核心战略驱动因素2026–2030国企采购顺风;降低国内市场销售摩擦明确 Agibot 已中标或正在跟进的具体政策挂钩招标
中国国内供应链成熟(2024 年本土份额 57%)驱动因素现在 — 2030单机成本低于西方同业;交付周期更短;供应商享受政策支持跟踪本土执行器和传感器供给放量后的 COGS 轨迹
已验证 ROI 案例(Longcheer、HT-Tech、Guangzhou Metro)驱动因素2026–2028中高加速同垂直场景复制销售;降低买方风险感知需要独立第三方验证所称指标;试点到合同转化率
AI 模型成熟与具身 AI 数据飞轮驱动因素2026–2030高(长期)AgiBot-World 数据集和 AIMA 模型栈相对纯硬件竞争对手形成差异对照已发布 L1–L5 能力框架,基准测试任务成功率轨迹
单机成本与资本开支门槛(每台 $10K–$100K+)约束现在 — 2029高人工成本场景之外的企业采用放慢;限制西方 SME 市场跟踪 ASP 趋势;低于 $10K 的定价何时打开大规模工业细分?
近距离人机协作中的安全与责任缺口约束现在 — 2030中高监管责任不清会拖慢医疗、食品和部分制造场景部署跟踪 MIIT 2026 标准采纳时间线;要求 Agibot 披露安全事件
美国 American Security Robotics Act(禁止中国机器人用于政府场景)约束2026–2027(若通过)高(政府细分)阻断联邦采购;可能扩展到盟国政府或 ITAR 等效制度跟踪法案进展;评估 Agibot 美国收入敞口与应急预案
IFR 评估:近期大规模采用概率不高;形态服从功能的限制约束现在 — 2030近期 TAM 上限落在结构化柔性任务;不威胁核心工业用例在高吞吐重复任务中,对比人形机器人与固定机械臂能力
生产一致性与多阶段测试挑战(Goldman Sachs 警示)约束2026–2027激进出货目标(10,000+ 台)可能在规模化质量或可靠性上落空向部署伙伴索取现场故障率、MTBF 数据和保修索赔率

驱动因素和约束评估综合自 IFR World Robotics 2025 Report、Goldman Sachs 中国调研(2026 年 1 月)、中国 MIIT 标准发布、美国立法记录和 Agibot APC 2026 材料。幅度和时间为定性评估;未做一手财务建模。“尽调追问”列是面向投资人的前瞻性指引,不代表当前证据。

[CM026, CM027, CM028, CM029, CM031, CM032]
FM004: 采用漏斗:从劳动力市场需求到 Agibot 可获取市场

示意性采用漏斗量化从总可寻址劳动力需求,到技术可自动化、经济可行、中国政策加速,再到 Agibot 当前可获取的逐级门槛; 均表示为 2030 年机会的一部分。

漏斗百分比为示意性估计,综合 Goldman Sachs(制造业劳动力缺口规模)、IFR(中国机器人安装率以及人形 vs 工业机器人比较) 和 MarketsandMarkets(近期可寻址市场规模)。数值表示截至 2026 年中,每个采用门槛可触达的总可寻址劳动力需求近似份额。 SOM 估计根据 Agibot 机队规模和公开部署数据推断;并非公司披露预测。

[CM031, CM033, CM036, CM037]

2.5 图表证据

Chapter 03

03竞争格局

3.1 竞争版图——同行、既有厂商与替代品

截至 2026 年中,人形机器人竞争版图可分为五类。直接同行是那些打造已商业可得或已商业出货的通用双足或轮式人形机器人的公司:Figure AI(Figure 03,美国)、Agility Robotics(Digit,美国)、Apptronik(Apollo,美国)、1X Technologies(NEO,挪威)和 Unitree Robotics(G1/H1,中国)。这些公司在形态、企业用例和先发定位上与 Agibot 竞争,争夺正在做首次人形机器人部署决策的客户。 超大市值公司和战略进入者构成第二层:Tesla(Optimus)在内部部署人形机器人,并提出长期 $20,000 消费者价格目标;Hyundai/Boston Dynamics(Atlas)聚焦研究和展示部署,而非一般商业销售;中国大型集团 BYD(Agibot 投资人)和 CATL 则提供西方竞争对手无法复制的供应链协同。 既有厂商和相邻替代品包括传统工业机器人(Fanuc、KUKA、ABB)、来自 Locus Robotics、Vecna 和 Geek+ 的自主移动机器人,以及 Universal Robots 和 Techman Robot 的协作机器人——它们都能以更低单机成本和已验证安全记录处理制造与物流任务。企业买方评估人形机器人时,近期现实替代方案是继续依赖美国仓库时薪 $15–25/hr 的人力,或在任务要求允许时安装传统固定自动化。 Goldman Sachs 的分析强调了市场早期动态:2025 年全球人形机器人出货量估计为 15,000–20,000 台,中国制造商贡献大部分销量;但同一研究也提示,2026 年可能成为「销量验证与预期重置」之年,制造商出货目标可能显著跑在可实际部署的企业需求前面——这是一项适用于 Agibot 和所有中国规模型同行的结构性风险。[CP001, CP002, CP003, CP004, CP005]

竞争对手画像表
公司类别总融资额目标细分旗舰产品核心差异化核心限制
Figure AI美国双足机器人初创公司$1.9B工业制造;家庭辅助Figure 03板块最高估值($39B);自研内部 AI;BMW 工业试点;投资人基础深BMW 部署条款未披露;未公布定价;企业验证样本有限
Agility Robotics美国双足机器人初创公司$641M+物流;仓储;制造Digit(双足)企业验证最多的人形机器人(Amazon、GXO、Schaeffler);Arc 机队平台加深锁定对 Amazon 的依赖可能限制独立性;设计偏物流场景;消费者路径有限
Apptronik美国双足机器人初创公司$1.45B+制造;物流Apollo承接 NASA Valkyrie 技术积累;RaaS 模式;Mercedes-Benz 部署;估值 $5.5B+定价未披露;公开的规模化部署证据有限;商业化早期
1X Technologies挪威双足机器人初创公司$125M+企业安防;消费者家庭NEO(双足);EVE(轮式)OpenAI 支持;唯一拥有企业商业化机队的西方人形机器人;NEO 目标价 $20KEVE 为轮式;NEO 尚未商业化规模部署;资本基础显著低于同业
Unitree Robotics中国硬件制造商未公开披露研究;价格敏感型企业G1($16K 双足);H1($90K)双足机器人价格最低;声称 2025 年 5,500 台;已递交 IPO 文件;产品线宽自主 AI 有限;美国数据安全顾虑;出口限制风险
Tesla(Optimus)超大市值公司入局者内部资金(Tesla 公司)消费者;内部工厂自动化Optimus Gen 2Tesla 制造规模;Dojo AI;宣称长期消费者目标价 $20K;全球品牌尚未面向第三方商业供货;交付时间线未确认
Boston Dynamics(Atlas)Hyundai 旗下研究平台未披露(Hyundai 支持)工业研究;展示演示Atlas(电动双足)完整杂技级双足能力;品牌声望;Hyundai 制造支持无通用商业供货;仅用于研究和展示
Agibot(A2 Ultra / G2)中国全栈初创公司$83M+(已披露)制造;商业服务;娱乐A2 Ultra(人形);G2(工业)量产规模领先(10K 台);通过 FCC/CE 认证;全系列产品组合;数据飞轮美国监管逆风;西方企业验证有限;披露融资额低于同业
传统工业机器人(Fanuc、KUKA、ABB)自动化既有厂商不适用(成熟行业)制造;汽车;精密装配6 轴机械臂安全性已验证;重复精度高;生态成熟;每台机械臂 <$50K;易集成无法在非结构化环境作业;位置固定;不具备双足移动
AMR(Locus Robotics、Vecna、Geek+ 等)相邻自动化既有厂商不适用(成熟行业)仓储;物流;履约移动底盘平台成本低于人形机器人;规模化已验证;易接入 WMS操作能力有限;不能爬楼梯;形态固定;不具备通用能力

融资数据来自新闻稿、HumanoidIndex 和媒体交叉核对;未经审计。传统机器人和 AMR 行代表现状替代方案,用于补全竞争边界。Tesla Optimus 和 Boston Dynamics Atlas 属于战略级条目;公司内部资金未单独披露。

[CP002, CP003, CP006, CP009, CP011, CP013]
FP001: 竞争定位图

人形机器人公司在两个轴上的竞争定位:生产与部署规模(x 轴,0=无到 10=大众市场已验证)和西方企业信任与验证 (y 轴,0=无到 10=强验证信任),基于截至 2026 年 5 月公开证据。位置是定性序数估计,非独立验证指标。

坐标轴是有证据支撑的序数评分,不是独立验证指标。生产规模反映累计出货或商业部署数量,不包括内部使用原型。西方企业信任度反映观察到的企业客户基础、监管姿态,以及是否独立于中国政府供应链。位置为作者基于公开来源的估计。

[CP001, CP004, CP005]

3.2 直接竞争对手画像与能力对比

Figure AI 是资本最充足的西方纯人形机器人公司,累计融资 $1.9B,包括 2025 年 9 月按 $39B 估值完成的 $1B+ Series C——这是全球该赛道最高估值。Figure 在 2025 年初终止与 OpenAI 的合作,完全转向自研内部 AI 模型;其 Figure 03 机器人现在瞄准通用家庭辅助。Figure 与 BMW 的 Figure 02 工厂试点仍在继续,但规模和经济条款未披露。Figure 的资本规模远超 Agibot 已披露的 $83M+。 Agility Robotics 是西方商业部署记录最经过验证的公司。Digit 已部署在 Amazon 履约中心、GXO Logistics 和 Schaeffler 制造场景,使其成为全球最经企业验证的双足人形机器人。Agility 融资 $641M+,包括 2025 年 6 月的 $400M Series C。其 Arc 机队管理平台与现有 WMS 和 AMR 基础设施集成,加深客户锁定效应。 Apptronik(Austin, TX)瞄准仓库和制造,并引用其源自 NASA Valkyrie 项目的背景。Apptronik 于 2026 年 2 月完成 $520M Series A,使已披露融资总额达到 $1.45B+,估值 $5.5B+。其机器人即服务模式与 Agility 的企业合同打法相似,Mercedes-Benz 是具名部署伙伴。 1X Technologies(挪威)由 OpenAI 支持,于 2024 年 1 月完成 $100M Series B,总融资约 $125M+。其轮式 EVE 机器人已部署于企业安防客户,而其 NEO 双足机器人瞄准 $20K 消费者价位,已开放预订,但截至 2026 年初尚无大规模商业部署。 Unitree Robotics(中国)是 Agibot 最直接的国内竞争对手,也是全行业最尖锐的硬件商品化威胁。Unitree 声称 2025 年人形机器人出货 5,500 台,并在中国提交 $580M IPO 申请,直接挑战 Agibot 的销量领先叙事。其 G1 双足机器人标价 $16,000,是全球公开可得价格最低的双足人形机器人;H1 标价 $90,000,面向研究和开发者细分。G1 拥有 23 个自由度和基础双足行走能力,但相较 Agibot 的全栈具身智能架构,其原生自主 AI 栈有限。 Agibot 通过三个角度差异化,而多数西方竞争对手无法同时匹配:一是统一全系列产品组合,横跨娱乐(X2 紧凑人形,$24.2K)、商业人形(A2 Lite,$44.6K)和工业机器人(G2 工厂平台);二是「一体三智」架构,把运动智能、交互智能和操作智能整合起来;三是国际认证(FCC, CE-MD, CE-RED, CR),多数中国竞争对手的人形产品线尚未取得这些认证。G2 工业平台已在 Longcheer Technology 的消费电子精密制造线上实现 8 小时零错误运行——这类企业运营证据,正是西方买方从试点走向部署承诺所需要的证明。[CP006, CP007, CP008, CP009, CP010, CP011]

功能 / 能力矩阵
能力Agibot(A2/G2)Figure AIAgility DigitApptronik ApolloUnitree G11X NEO
规模化商业部署是(累计 10K 台)仅试点(BMW)是(Amazon、GXO、Schaeffler)商业化早期是(声称 2025 年 5,500 台)仅预订
双足运动是(A2/A3 系列)是(成熟)是(成熟)是(基础,23 DoF)是(设计中)
灵巧操作是(OmniHand)是(多指)是(高负载)部分(可选灵巧手)是(设计中)
自主任务执行(结构化)部分部分部分部分部分(遥操作优先)
国际安全认证是(FCC、CE-MD、CE-RED、CR)UnknownUnknownUnknownUnknownUnknown
已发布标价是($24.2K–$44.6K)是($16K–$90K)是(目标 $20K)
美国 / 欧盟企业信任姿态反向(监管逆风)反向(监管逆风)中性
机队管理软件平台是(Genie Studio)Unknown是(Arc)UnknownUnknown

能力评估来自截至 2026 年 5 月的官方产品页、新闻稿和第三方报道。尚无独立技术基准对这些系统做正面对比。“部分”表示能力已演示,但受任务约束或需要人工监督。“未知”表示公开文档缺失。

[CP019, CP020, CP021, CP022, CP023]
FP002: 功能广度 / 能力图谱

对主要人形机器人竞争对手在八个维度做序数能力评分。评分(强 / 部分 / 弱 / 未知)来自截至 2026 年 5 月的公开产品页面和媒体报道。没有独立基准横向比较这些平台。

评分基于截至 2026 年 5 月的公开产品信息和媒体报道做定性判断。没有独立技术基准横向比较这些平台。自主 AI 评分反映部署或演示中观察到的结构化任务表现,不代表通用自主能力愿景。未知表示缺少公开文档。

[CP008, CP010, CP013, CP015, CP017]

3.3 定价、部署证据与 GTM 分销

Agibot 是少数公布国际标价的人形机器人公司之一。其 A2 Lite 在国际线上商店标价 $44,560,X2 娱乐机器人标价 $24,240——相比普遍不披露企业定价的西方竞争对手,价格比较更可操作。Figure AI、Agility Robotics 和 Apptronik 均不披露单价或合同经济性;行业对企业双足部署的估计为每台 $100K–$150K+,或相当的年度 RaaS 合同价值。Apptronik 的 RaaS 模式已公开说明,但具体经济性未披露。 Unitree G1 的 $16,000 仍是该赛道最具冲击力的价格信号。双足机器人低于 $20K,压低了所有西方竞争对手,也逼近 Agibot 的 X2 消费级价位。如果 Unitree 通过自有 AI 或第三方模型集成实现足够的自主运行,它可能拿下价格敏感的研究和中小企业细分,并迫使中国人形机器人层级爆发硬件价格战。 部署证据方面,Agility Robotics 的企业验证最深:Amazon Digit 项目是唯一由战略投资人作为锚定客户的高销量工业人形机器人部署。Agibot 的 10,000 台累计机队在数量上亮眼,但娱乐、商业展厅和工厂地面应用混合在一起,使西方企业买方更难对照到具体劳动力替代 ROI。Agibot 的 G2 零错误制造声明(Longcheer)是一个早期但重要的、西方企业可对标的证明点;若要在规模上建立企业信任,还需要独立验证。 GTM 和分销上,Agibot 运营全球租赁平台和直接国际线上商店,而西方竞争对手主要通过专门销售团队签企业合同。Agibot 的分销打法——高销量、较低价格、平台生态——反映其中国市场出身;西方竞争对手则聚焦以证据驱动的企业销售,并提供高接触支持和客户管理。[CP024, CP025, CP026, CP027, CP028, CP029]

定价 / 打包方案对比
公司 / 产品每台标价(目录价)订阅 / RaaS 选项定价置信度定价模式竞争含义
Agibot X2(紧凑型人形机器人)$24,240 USD全球租赁平台直接销售 + 租赁已公布国际定价;全球开发者和 SME 可获得
Agibot A2 Lite(全尺寸人形机器人)$44,560 USD企业租赁 / 部署服务直接销售 + 部署与西方企业级估算具竞争力;若能力相当则形成价格压制
Unitree G1(双足)$16,000 USD不提供直接硬件销售全球最低双足价格;对所有人形机器人硬件形成尖锐商品化压力
Unitree H1(双足研究)$90,000 USD不提供硬件直接销售面向研究与开发者细分市场;价格高于 Agibot A2 档位
1X NEO(消费级双足机器人)$20,000 USD$499/month直接销售或订阅与 Tesla Optimus 声称的长期消费级目标直接竞争
Figure AI(企业)未披露未披露BMW 工厂合同BMW 条款保密;没有公开可得的消费级或通用企业报价
Agility Digit(企业)约 $100K–$150K(估算)或 RaaSRaaS 合同(估算)企业合同或 RaaS定价未公开披露;Amazon 交易经济性保密;仅为分析师推断
Apptronik Apollo(企业)约 $100K+(估算)RaaS 模式(已宣布)企业 RaaS已说明 RaaS 模式;具体定价未公布;Mercedes-Benz 条款保密
传统机械臂(Fanuc / KUKA)$30K–$150K维护合同资本设备 + 服务人形机器人在任务等价应用中必须跑赢已验证固定自动化的 IRR

Agibot、Unitree 和 1X 定价来自官方产品页。西方企业竞争对手定价由媒体推断并经分析师估算;没有一手来源披露。任何西方人形机器人公司都未公开详述 RaaS 合同结构。传统机器人定价为近似行业区间。

[CP024, CP025, CP026, CP027, CP016]

3.4 切换成本、锁定效应与供应链接入

企业人形机器人部署通过三种累积机制形成切换成本。第一,系统集成:部署一支人形机器人机队,需要把机器人嵌入仓库管理系统、制造执行系统、安全协议和运营工作流。Agility 的 Arc 平台专门利用这一点,把 Digit 接入现有仓库自动化和 AMR 基础设施;替换 Digit 就必须重新集成更广泛的自动化栈。第二,训练模型定制:每次部署都会产生专有的任务级微调数据。更换供应商意味着失去在特定设施工作流和任务配置上累积的模型表现。第三,在受监管工业环境中,员工培训和重新认证投入并不低,安全认证要求也必须重复完成。 Agibot 的锁定机会大体对应这些机制。它的 Genie Studio 软件平台和 OmniHand 灵巧操作系统,可能成为客户在 Agibot 栈上搭建工作流的工具层。供应链接入提供了西方竞争对手难以复制的结构性成本优势:CATL 支持的电池部件供应、Longcheer 面向消费电子应用的精密制造,以及更广泛的中国供应商生态,共同支撑 Agibot 在可比硬件规格下以低于西方成本结构的价格销售。 当前多栖采购仍可行,因为还没有单一人形机器人平台在任何用例上形成有意义规模的主导地位。部署量增长后,先发客户会累积切换成本;这使 Agility、Apptronik 以及在较小程度上 Agibot 于 2025–2026 年积累的企业参考部署,在战略意义上远超其即时合同收入。[CP030, CP031, CP032, CP033]

3.5 护城河耐久性、商品化风险与反向证据

Agibot 最耐久的竞争护城河,是生产规模驱动的数据飞轮。截至 2026 年 3 月,Agibot 已生产 10,000 台机器人——上一季度生产速度提升 4×——运营着 Tesla 之外最大的真实世界具身 AI 训练机队。每台已部署机器人都会产生专有真实世界数据,没有可比部署规模的竞争对手无法匹配。这在结构上类似 Tesla 从 Autopilot 车队数据获得的 FSD 优势:部署机器人越多,观察到的任务场景越多样,整个机队共享的 AI 模型就越强。如果基于仿真的训练成熟到足以替代真实世界数据需求,这一优势会被削弱。 Agibot 竞争位置上最严重的反向证据来自三个方向。第一,美国拟议的 American Security Robotics Act(2026 年 3 月)将禁止联邦政府购买或运营中国制造的人形机器人;美国参议员明确援引潜在数据向中国传输的国家安全风险,并在法案语境中点名 Agibot 和 Unitree。即使立法只针对政府采购,国防相邻行业、关键基础设施和受监管行业的企业买方也可能自行采用同样逻辑——在没有任何私人部门销售正式限制的情况下,实际上把 Agibot 排除在美国可触达市场的重要部分之外。 第二,相对西方竞争对手的资本劣势很大。Figure AI($1.9B)、Agility($641M)和 Apptronik($1.45B)在 2024–2026 年合计融资超过 $4B——约为 Agibot 已披露 $83M+ 融资的 48×。即便 Agibot 目前制造成本结构和生产规模更优,竞争对手持续砸钱做 R&D,也可能在两到三年内缩小自主能力和功能差距。 第三,也是最迫近的一点,Unitree 和其他中国制造商带来的价格商品化威胁,可能剥掉 Agibot 商业模式中的硬件利润率。Goldman Sachs 特别提示,2026 年可能是预期重置之年,指出头部中国制造商的出货目标可能显著跑在实际企业需求前面。如果供给先于可部署需求冲刺,硬件 ASP 会压缩,Agibot 的定价优势会收窄。中国工信部(MIIT)于 2026 年 2 月发布国家人形机器人标准,为参与制定标准的本土厂商创造潜在监管一致性优势;而西方竞争对手面向全球市场时,既要满足本地标准,也要满足中国规范,合规成本更高。[CP034, CP035, CP036, CP037, CP038, CP039]

护城河持久性 / 竞争风险登记表
护城河或风险类型持久性评估主要威胁风险严重性尽调要求
生产规模数据飞轮(10K 台)数据 / AI 优势中 — 随着仿真训练(NVIDIA Isaac、DeepMind)成熟,真实世界数据溢价可能下降仿真到现实迁移进步,降低对自有真实世界数据的需求量化 Agibot 的任务和环境覆盖范围,并与开源仿真数据集比较
中国制造成本结构和 CATL/Longcheer 供应通道成本与供应链护城河短期高 — 西方竞争对手难以复制;生态带来结构性优势关税、出口管制或地缘政治脱钩冲击中国供应链评估 BOM 对中美贸易政策的敏感性;识别非中国备选供应商
全系列产品组合(娱乐到工业)产品广度中 — 西方竞争对手扩展平台;Unitree 增加形态Figure AI 拓展到家庭场景(Figure 03);Agility 走出物流场景将 Agibot 路线图广度与 Figure AI 家庭发布和 Agility 服务机器人计划比较
国际安全认证(FCC、CE-MD、CE-RED、CR)监管合规护城河短期高;长期中,竞争对手拿到同类认证后优势会下降销量足以覆盖成本后,西方竞争对手取得同等认证低-中核实 CE/FCC 范围是否覆盖工业用例,而不只是消费娱乐产品
美国 American Security Robotics Act(2026 年 3 月提案)监管逆风 / 反向风险高负面风险 — 可能从结构上把 Agibot 排除在美国政府及相邻市场之外法案通过,且企业自我排除范围超出政府采购跟踪法案状态;评估企业买方在政府范围之外的行为性排除
西方竞争对手资本压顶(已融资 $4B+,而 Agibot 披露 $83M+)资本与研发风险高风险 — 竞争对手可在 AI 研发、人才和企业销售基础设施上砸更多钱Figure($1.9B)+ Agility($641M)+ Apptronik($1.45B)持续合计研发投入评估 Agibot 未披露融资总额;比较其 AI 研发投入速度与西方同行
中国硬件竞争带来的价格商品化(Unitree $16K)定价与利润率风险高风险 — Unitree $16K G1 及潜在继续降价,会压缩全行业硬件利润率Unitree IPO 资本支持进一步降价;更多中国新进入者把价格打到 $20K 以下评估 Agibot 当前 ASP 下的毛利率;建模平均价格下降 30–50% 后的盈利能力

持久性和严重性评估基于公开新闻稿、分析师研究和监管文件,均为定性判断。中国人形机器人行业没有独立财务或技术基准可用。风险严重性相对于 Agibot 当前战略位置和已披露信息评估。

[CP034, CP035, CP036, CP037, CP038, CP039]
FP003: 护城河 / 就绪度 KPI

基于截至 2026 年 5 月的公开证据,对 Agibot 与主要对手的关键竞争耐久性指标做对比。

[CP001, CP036, CP037, CP038, CP039]

3.6 图表证据

Chapter 04

04财务情况

4.1 收入来源、定价与变现

AGIBOT 的变现模式组合硬件产品销售、租赁 / 机器人即服务(RaaS)产品,以及正在增长的软件插件和平台服务层。最清晰的定价信号来自公司的全球线上商店(store.agibot.com),公开列出 X2 半尺寸娱乐人形机器人价格为 $24,240 USD,全尺寸 A2 Lite 娱乐人形机器人价格为 $44,560 USD。两款产品都收取 $500–$3,000 运输押金,所有进口关税和税费由客户承担。产品包含一年质保(或 3,000 小时累计运行时间,以先到者为准)。A2 Ultra——AGIBOT 旗舰商业级全尺寸人形机器人——没有公开标价;根据官方产品页,它已部署在 20 多家领先企业,用于展厅、客户服务和品牌背书场景。A2 Ultra 采用 B2B 协商定价,尚未披露。 除了直接硬件销售,AGIBOT 还建立了「全球租赁平台」,公司或其分销伙伴通过该平台提供人形机器人短期商业使用。Humanoids Daily 报道酒店垂直的租赁费率为每日 €899,部署包括 Genting Malaysia(MOU 于 2026 年 4 月 17 日签署),以及通过与 Singtel 的机器人即服务(RaaS)安排进入新加坡。€899/day 提供了具体租赁价格点,但合同结构(最低期限、数量阶梯、是否包含维护)尚未得到独立确认。 软件层变现已在 AGIBOT 产品 FAQ 中明确写出:A2 Lite 和 A2 Ultra 的群控软件、升级技能包和 VR 远程控制套件需要「额外付费」。这些组件的定价没有公开披露,但「技能包商店」和 AIMMASTER 客户端软件的存在,显示硬件之外的附加收入机会正在形成。AIMA 开放栈开发者生态在 APC 2026 发布,瞄准在 AGIBOT 硬件上构建应用的第三方开发者;商业许可或 API 定价尚未公布。G2 平台在工厂客户(Longcheer Technology,并计划扩展到汽车和半导体)处产生的工业部署收入,是第四条收入来源,但项目经济性完全未披露。[CI001, CI002, CI003, CI004, CI005, CI006]

收入流表
收入流机制计价单位当前数值 / 状态质量尽调要求
娱乐 / 商业表演硬件(X2、A2 Lite、A2 Ultra)通过线上商店和 B2B 渠道一次性买断式硬件销售每台X2:$24,240;A2 Lite:$44,560;A2 Ultra:未披露 B2B 价格中 — X2 和 A2 Lite 标价已公开确认;A2 Ultra 定价未披露;实际 ASP 未知确认 A2 Ultra 标价和实际 ASP;获取按 SKU 和地域组合拆分的销量
工业制造部署(Longcheer 的 G2)G2 移动操作机器人在工厂产线按项目部署每次部署 / 每台多台已部署于 Longcheer(平板产线);计划 2026 年 Q3 部署 100 台;合同条款未披露低 — 部署已确认,但定价、抽成率和合同结构完全未披露获取 G2 部署定价(单台购买 vs. 服务费)、合同期限和 ROI 分成条款
租赁 / 机器人即服务(酒店、商业活动)通过全球租赁平台提供短中期租赁;通过 Singtel 和区域伙伴面向企业提供 RaaS 模式每天 / 每台机器人每月€899/天(据报道用于酒店);Singtel 企业 RaaS 条款未披露低 — 仅一个价格点见诸报道;平台规模、MRR 和合同条款未验证确认租赁机队总规模、MRR、客户集中度和合同续约率
软件附加项(技能包、群控软件、VR 套件、AIMMASTER)随硬件销售或租赁附加的软件 / 配件购买每个许可证 / 每份订阅根据官方 FAQ,已确认「需要额外付费」;具体价格未披露低 — 收入流原则上已确认;定价、附着率和贡献未披露获取软件 SKU 定价、每台硬件附着率和软件毛利率
开发者生态 / 数据服务(AIMA、Genie Studio、开源数据集)面向第三方开发者和企业 AI 集成商收取平台许可、API 访问、开发者订阅或数据服务费用每位开发者 / 每次 API 调用 / 每份订阅AIMA 生态已在 APC 2026 发布;截至 2026 年 5 月尚未公布商业条款尚未货币化 — 生态已确认,但未公布定价或收入明确 AIMA 货币化时间表、定价模型和目标可服务开发者群体

收入和定价数据来自官方 AGIBOT 商店页面、产品 FAQ、公司新闻稿及第三方媒体。AGIBOT 未披露总收入、按收入流拆分收入或实际定价。质量评级反映证据强度,不代表收入规模。A2 Ultra 和工业部署定价完全是未披露的 B2B 条款。

[CI001, CI002, CI003, CI004, CI005, CI006]
定价 / 货币化表
产品 / 服务标价 / 费率包含内容折扣 / 未知项来源
X2(半尺寸娱乐人形机器人)$24,240 USD机器人本体、充电器、遥控器、麦克风、电池;基础技能集批量折扣结构未知;群控软件和高级导航包需额外付费官方商店:store.agibot.com/products/x2
A2 Lite(全尺寸娱乐人形机器人)$44,560 USD机器人本体、电池、充电器、配件充电器、无线急停、控制器、运输车技能包、群控软件、VR 套件:需额外付费;保修 1 年 / 3,000 小时官方商店:store.agibot.com/products/a2-lite
A2 Ultra(全尺寸商业人形机器人)未披露(B2B 协商)已在 20+ 家企业部署,用于展厅、服务、表演;4 项认证(CR、CE-MD、CE-RED、FCC)定价完全未披露;IP 定制最低订单:5 台官方产品页:store.agibot.com/products/a2-ultra
租赁 / 酒店 RaaS€899/天(据报道)机器人使用权,可能含基础运营支持;维护条款未确认Humanoids Daily 报道酒店垂直场景费率;企业 RaaS(Singtel)条款未披露;批量或多周阶梯价未知Humanoids Daily(10,000 台报道)
G2 工业部署未披露(按项目)集成、通过 Genie Sim 3.0 调试、现场培训;机型换线时 95% 设备复用无公开定价;已有 4 个月集成周期和 36 小时仿真到现实调试记录PRNewswire;CnTechPost;Interesting Engineering

标价来自官方 AGIBOT 商店页面,因此只代表折扣前标价。实际成交价、批量折扣、渠道利润率和企业合同定价均完全未披露。€899/天租赁费率只是媒体报道的单个数据点,合同结构未确认。工业 G2 定价没有公开资料。

[CI001, CI002, CI003, CI004, CI005, CI006]
FI001: 收入模型桥

定性展示 AGIBOT 的客户互动如何通过硬件销售、工业部署和租赁 / RaaS 渠道转化为收入及潜在毛利。收入节点以下的所有美元流均为估计或代理指标;AGIBOT 未披露收入、毛利率或 COGS。

所有财务量级均为定性或估计。AGIBOT 未披露任何渠道的收入、COGS 或毛利率。节点描述反映有证据支撑的运营事实(定价、部署证据),而不是已确认的财务流。毛利节点有意不量化——公开信息不足以支撑具体估计。

[CI001, CI002, CI003, CI005, CI006, CI039]

4.2 成本结构、资本强度与制造经济性

AGIBOT 的成本结构很像一家早期垂直整合机器人公司:研发和 AI 模型训练吃掉大量资本开支,制造端投入很重,每一次工业部署还要叠加服务交付成本。公司按车规级生产原则制造,每台 A2 Lite 设置 92 个综合检验点,A2 Ultra 还要经受最高 2,000 小时行走测试(包括 360 小时连续运行且无异常)。这套质量控制匹配高端硬件制造的经济模型,也意味着单机质量保证成本不低,但公司没有公开披露。热插拔电池系统、UWB 群体定位(支持 100 台机器人同步表演)、带 3D 触觉感知的灵巧手,都让 AGIBOT 的 BOM 复杂度高于更简单的人形机器人设计。 部署经济性最有信息量的代理指标来自 Longcheer Technology 的 G2 工业项目。AGIBOT 称,借助 Genie Sim 3.0 数字孪生压缩调试,G2 现场安装只需 36 小时,而传统调试通常要数月——这显著改善部署 COGS。产品型号切换时,设备复用率达到 95%;新产线配置的再训练不超过 4 小时。这些指标说明,G2 的经常性服务成本结构低于第一代工业机器人,但每台已部署 G2 的绝对美元经济性仍未披露。Longcheer 项目从启动到投产用了 4 个月,这可以作为单客户集成成本的合理代理,投资者应在尽调中量化。 制造爬坡的经济性可以从产速数据看出来:AGIBOT 从成立(February 2023)到 1,000 台大约用了 24 个月,early 2026 累计达到 5,000 台,随后又在大约 3 个月内从 5,000 台跃升至 10,000 台,吞吐速度加快 4×。这轮爬坡意味着公司在供应链工业化上投入不小,覆盖零部件、电池系统、执行器和检测产能;其投资方名单包括与汽车供应链相关的 BYD、Tencent、JD.com 以及 Longcheer Technology 本身(上市代码:603341.SH),也与此相符。HumanoidIndex 还提到,电池系统供应有 CATL 支持。Swancor Advanced Materials 收购(CNY 2.1B 财团交易,~$290M)显示 AGIBOT 想向上游高性能材料垂直整合,这是一笔重资本承诺,也会进一步放大融资依赖。[CI009, CI010, CI011, CI012, CI013, CI014]

单位经济效益表
指标数值 / 估计置信度重要性尽调要求
A2 Lite 单台硬件 COGS未披露;按车规级生产和 92 项 QC 检查点估算为 $15,000–$30,000(类比同行基准)低(估算)决定最大公开定价 SKU($44,560 标价)的毛利率;若当前销量下为负毛利,说明硬件低于成本销售要求按 SKU、按批次提供 COGS 瀑布(BOM、制造人工、QC、运费、质保准备金)
X2 单台硬件 COGS未披露;按当前销量估算为 $8,000–$18,000(半尺寸、DOF 更少)低(估算)$24,240 标价;若 COGS > $20,000,毛利率为负或极薄要求管理账提供单台 COGS 和毛利率数据
产量每扩大 10× 后的制造成本改善未披露;行业类比显示,单位产量每扩大 10×,BOM 成本可下降 20–35%低(据行业基准推断)决定硬件业务走向毛利为正的路径,以及 RaaS 经济性何时转正要求提供各生产批次学习曲线数据;比较 1,000、5,000、10,000 台时的 BOM
G2 工业部署毛利率未披露;Longcheer 合同条款未公开;€899/天租赁参考数据暗示满负荷利用时毛利可能为正低(无可用数据)工业部署是投资逻辑最关键的收入流;利润率质量决定长期业务价值获取 G2 部署定价、每个客户集成成本、每台每月持续服务成本
客户获取成本(CAC)未披露;企业人形机器人销售周期按行业类比估计为 6–18 个月;CAC 未披露低(按行业常态估算)回本周期决定资本需求;早期工业部署阶段 CAC 高可以预期,但必须量化要求提供销售周期数据、平均合同价值、按细分市场(娱乐 vs 工业)拆分的 CAC 和回本周期
毛利率(合并口径)公司完全未披露;未披露收入或毛利润不可得核心承销输入;没有 COGS、渠道折扣和收入组合数据,不能仅凭标价推断要求提供 FY2024 和 FY2025 经审计或管理层审阅的利润表或 P&L 摘要

本表所有财务数值要么是基于行业基准和类比的估算,要么只是标注为未披露。AGIBOT 未公布收入、毛利率、COGS 或任何利润表数据。COGS 估算来自同行硬件成本基准,未经验证。本表用于组织尽调问题,并非断言具体财务结果。

[CI009, CI010, CI011, CI012, CI039]
FI002: 单元经济性桥

以 A2 Lite 这个公开定价 SKU 为代表,从单台收入经估计成本层到单台毛利率做定性流转。所有成本节点均来自行业基准估计;AGIBOT 未披露 COGS 或利润率数据。

所有成本和利润率估计均来自已发布的人形机器人硬件成本基准,以及可比单一来源证据的类比(行业分析师估计;未针对 AGIBOT 验证)。COGS 区间刻意设宽,以反映高度不确定。任何投资决定都应先要求实际经审计 COGS 数据,再接受该估计为可靠。

[CI001, CI009, CI010, CI039]
FI004: 资本强度 / 现金流图

定性映射 AGIBOT 在四个运营阶段的资本消耗:从研发和制造扩产,到商业部署,再到通过收购 Swancor 做上游垂直整合。量级为定性;公司未披露财务数据。

所有强度评估均为定性。AGIBOT 未披露研发、制造资本开支和营运资本的财务量级。Swancor 收购成本是本图唯一硬财务数字(来源于 Swancor A 股披露)。相对强度排序来自运营里程碑的逻辑推断,而非已确认财务报表。

[CI011, CI013, CI019, CI028, CI034]

4.3 产量里程碑、部署指标与商业牵引

AGIBOT 最具体的财务牵引信号是产量。公司在 March 2026 生产出第 10,000 台人形机器人;Omdia(HumanoidIndex 引用)将 AGIBOT 列为 2025 年全球人形机器人出货量第一,约 5,100–5,200 台。这个排名存在争议:Unitree Robotics 声称其 2025 年人形机器人出货 5,500 台,并已申请 $580 million A 股 IPO——这一竞争性说法尚未得到独立验证,两家公司出货口径的方法差异也没有解决。尽调中应把 AGIBOT 的产量视为收入潜力的代理指标,而不是已确认收入,因为公司没有公开披露已商用部署设备、演示 / 展示设备和预商用库存之间的构成。 最清晰的部署证明来自 G2 工业项目。多台 G2 机器人已确认在 Jiangxi Nanchang 的 Longcheer Technology 平板产线上运行,工位为 MMIT。实测表现:8 小时连续运行完成 2,283 个任务,零错误;每小时吞吐 310 件产品;单次节拍 18–20 秒。这种在产线条件下跑通 24/7 双班制的能力,是 AGIBOT 迄今发布的最有说服力的商业证明,但 Longcheer 当前部署台数和商业条款均未披露。公司计划到 Q3 2026 扩展至 100 台,并延伸到汽车和半导体制造。 在娱乐和商业表演侧,A2 Ultra 产品页确认已部署于「20 多家头部企业」。A3 人形机器人支持 100 台机器人 UWB 同步群体表演,已公开演示,并将在 mid-2026 进入商用。根据 Humanoids Daily,全球租赁 / RaaS 平台截至 March 2026 覆盖 17 个国家,指向一个正在运转的商业运营模型。IDC 数据显示,2025 年全球超过 85% 的人形机器人部署集中在娱乐、教育和导览场景,这与 AGIBOT 的早期商业基础一致。公司在 April 2026 的 APC 2026 上宣布转向工业制造部署,并称其为「部署元年」;这正是投资逻辑的关键转换,但尚未产出已披露的财务结果。[CI015, CI016, CI017, CI018, CI019, CI020]

公开财务缺口表
缺失指标对承销的重要性缺口持续的影响尽调路径
收入(总额;按收入流;按地域)确认产量代理指标是否转化为确认收入;决定 ARR 运行率和增长轨迹阻断 — 没有收入轨迹,投资逻辑无法验证;LTV、回本周期或增长率模型都无法建立要求管理账(FY2024 和 FY2025),并按渠道(硬件 vs 租赁 vs 工业部署)拆分收入
毛利率和每个 SKU 的 COGS判断硬件业务是否已越过成本线;若当前销量下为负毛利,走向盈利所需资本会大得多阻断 — 规模化后仍为负毛利,会显著重塑融资规模和时间表要求按产品代际提供 COGS 瀑布;BOM、制造人工、QC、质保;拆分硬件 vs 服务利润率
现金余额、烧钱速度和现金跑道决定下一轮融资紧迫性;考虑 Swancor 收购,资本消耗可能远高于已披露 VC阻断 — 没有烧钱速度和现金跑道,无法评估资本充足性和稀释风险要求截至 2026 年 Q1 的资金头寸、月度现金消耗和前瞻资本计划
工业部署合同经济性(G2 Longcheer 和管线)G2 工业部署是投资逻辑最关键的高价值收入流;定价和利润率完全未知重大 — 若 G2 经济性在早期销量下边际或低于 COGS,工业收入爬坡将需要额外补贴资本要求 G2 部署合同模板(定价、期限、服务费、排他性);Longcheer 合同 ACV;按阶段拆分客户管线

本表反映截至报告日期(2026 年 5 月)的信息缺口。四个缺口都被视为投资承销的重大或阻断项。仅靠公开来源无法解决任何一个缺口;都需要正式数据室访问。公司概况章节记录融资时间线;本表专注于前瞻资本充足性和收入质量缺口。

[CI031, CI039, CI040]
FI003: 财务估计区间

关键财务估计的有来源支撑或由证据推导的数值区间。由于 AGIBOT 几乎完全没有披露财务数据,区间反映不确定性。所有估计均为暂定,任何投资决定前都应由正式数据室确认数据替换。

已确认标价(A2 Lite、X2)来自官方来源,数字精确。其他估计均有重大不确定性。COGS 区间有意设宽——没有公开依据可在缺少经审计数据的情况下收窄。估值和收购数字来自第三方媒体和 Swancor 公司披露。AGIBOT 未披露收入、毛利率、烧钱速度或现金头寸。

[CI001, CI002, CI006, CI023, CI024, CI027]

4.4 资本充足性、融资依赖与 IPO 信号

AGIBOT 自 February 2023 以来已完成十余轮融资;完整时间线已在公司概况章节列出,这里只作为评估资本充足性的背景。按 Tracxn 和 HumanoidIndex,已披露风险投资约为 $83.8 million。但这个数字几乎肯定低估了实际融资总额:中国民营企业披露通常不完整,QCC 工商登记显示公司有 40+ 名股东,覆盖多轮未披露规模的融资。QCC 数据中的关键轮次包括 BYD 参与 A+ 轮(August 2023)、Tencent 的 Series B(March 2025)、JD.com / JoyGen 的 B+ 轮(May 2025),以及 LG Electronics / Mirae Asset New Growth Fund 战略轮(August 2025)。AGIBOT 拒绝披露 LG Electronics / Mirae Asset 轮次规模或投后估值;Pandaily 报道称,其投前估值超过 7 billion yuan(按 July 2025 汇率约 $960 million)。 影响最大的资本事件是 Swancor Advanced Materials 收购。July 8, 2025,Swancor Advanced Materials(深圳 A 股上市)披露,由 AGIBOT 及其核心团队组成的财团同意支付最高 CNY 2.1 billion(~$290 million USD)取得公司控制权。这个金额约为 AGIBOT 已披露 VC 总额的 3.5×,代表其押注上游高性能材料垂直整合。AGIBOT 澄清,12 个月内没有改变 Swancor 主业的确定计划,并否认借壳上市;但市场参与者怀疑这是一条上市路径,这一假设仍然存在。该收购大幅提高 AGIBOT 的融资需求,也压缩了可用于机器人研发和部署的资本。 IPO 信号:Reuters(经 US News,March 2026)报道,AGIBOT 和 Unitree 是准备在 2026 年于中国上市的中国人形机器人公司之一。公司尚未宣布正式招股书、路演日期或承销商任命。任何 A 股或同等上市流程都必须披露收入和毛利率;这一披露时点可能是公开投资者首次看到已确认财务指标的机会。在此之前,收入、烧钱速度和现金跑道完全未披露,构成财务承销的阻断级缺口。[CI023, CI024, CI025, CI026, CI027, CI028]

资本充足性表
项目数值 / 估计置信度备注
已披露风险投资(所有轮次)约 $83.8 million USD(据 Tracxn 和 HumanoidIndex)中(第三方数据库汇总;可能低估真实总额)中国民营轮次常只披露部分信息;QCC 显示 40+ 名股东。Series B:Tencent(2025 年 3 月);B+:JD.com/JoyGen(2025 年 5 月);战略轮:LG Electronics / Mirae Asset New Growth Fund(2025 年 8 月)
投前估值(最近报道)超过 70 亿元人民币(按 2025 年 7 月汇率约 $960 million USD),据 Pandaily 报道低(第三方媒体报道;公司未确认)估值未经 AGIBOT 确认;LG/Mirae 轮次规模和投后估值未披露
Swancor Advanced Materials 收购成本最高 CNY 2.1 billion(约 $290 million USD),由 AGIBOT 与核心团队牵头的财团(2025 年 7 月 8 日宣布)中(Swancor 作为上市公司公告披露)约为已披露 VC 总额的 3.5×;上游材料整合逻辑;AGIBOT 否认借壳上市计划
月度烧钱速度 / 现金跑道完全未披露不可得未披露收入、P&L 或烧钱速度;没有数据室访问无法估算
IPO / 上市计划据 Reuters(经 US News,2026 年 3 月)报道,公司准备 2026 年在中国上市;未提交招股书,也未确认日期低(媒体报道;公司未确认正式上市时间表)A 股上市将要求公开披露收入;时间未知;Swancor 收购引发借壳上市猜测,公司已否认

资本充足性数据来自第三方数据库(Tracxn、HumanoidIndex)、中国企业注册文件(QCC)、媒体报道(Pandaily、Yicai Global、Reuters)以及 Swancor Advanced Materials 作为 A 股上市公司的公开披露。AGIBOT 未披露手头现金、月度烧钱速度或现金跑道。烧钱速度和现金跑道单元格保留为「未披露」,是为了标示尽调优先级,不代表数值为零。

[CI023, CI024, CI025, CI026, CI027, CI028]

4.5 财务结论与尽调阻断项

AGIBOT 讲出了一个可信的产能规模故事——从成立起约 26 个月生产 10,000 台、按 Omdia 为 2025 年全球人形机器人出货第一、已确认工业部署且有可量化性能指标——但截至 May 2026,支撑这些运营里程碑的财务实质对外部投资者几乎完全不透明。 没有已披露财务数据,就无法评估收入质量。当前收入结构看起来主要由娱乐和商业表演硬件销售驱动,产品包括 X2、A2 Lite、A2 Ultra,标价 $24,240–$44,560;同时有刚起步的工业部署(Longcheer 的 G2)和租赁 / RaaS 平台。硬件销售收入天然低于订阅或 RaaS 等经常性收入的质量:它不可重复、履约依赖资本开支,还带有保修责任。€899/day 的租赁 / RaaS 平台和 Singtel 企业 RaaS 安排说明公司正在形成一层经常性收入,但其规模、合同结构以及对总收入的贡献都未知。A2 Ultra 在 B2B 商业渠道的定价未披露,是关键缺口:如果该产品规模化单价达到 $60,000–$150,000+(与国际可比商业人形机器人定价一致),娱乐 / 展厅垂直领域的收入集中度可能很高。 资本结构风险较高。Swancor 收购(~$290 million)把已披露 VC 总额数倍的资金投向单一上游整合赌注。再加上公司已确认的 10,000 台产能(意味着持续 BOM 和制造现金流),AGIBOT 的实际资本消耗很可能是已披露 $83.8 million VC 的数倍。毛利率、烧钱速度和现金跑道未知,但公司的运营规模、制造强度以及新增的 Swancor 收购都说明融资依赖高,近期资本规划非常关键。 拟议中的美国 American Security Robotics Act 将禁止美国政府采购或使用中国制造的人形机器人,这对 AGIBOT 声称的美国市场扩张战略构成重大不利因素,可能封闭相当一部分可触达需求。Hyperscale Data / Omnipresent Robotics 的 Michigan 工厂合作(100,000 sq. ft.)以及 Singtel / NCS 新加坡合作确认了国际部署野心,但若没有已确认的合同经济性,仍不足以证明收入重要性。 核心尽调阻断项包括:(1)收入和年经常性收入(ARR)——完全未披露;(2)毛利率和单机 COGS——完全未披露;(3)烧钱速度和现金跑道——完全未披露;(4)Swancor 收购财务结构——对资产负债表的影响未量化;(5)工业部署合同条款——价格、规模和 ROI 未公开。若不能通过正式数据室解决 1–3 项,任何投资决定都无法负责任地承销。[CI031, CI032, CI035, CI039, CI040]

Chapter 05

05产品与技术

5.1 产品线与客户工作流

Agibot 将商业产品组合拆成三条品牌机器人系列,对应不同部署场景和客户工作流阶段。Yuanzheng (Expedition) 系列覆盖用于交互和企业服务的全尺寸双足人形机器人:A2 Ultra 是已获认证的旗舰机型(CR、CE-MD、CE-RED、FCC),A2 Lite 是 Agibot 商店标价 $44,560 的低成本商业版本,A2-W 则是面向数据采集和工厂导航优化的轮式底盘版本。Genie (精灵) 系列面向工业制造:G1 机器人提供原生数据采集能力,配备毫秒级延迟 VR 遥操作、8 个高分辨率摄像头和双臂六轴力感知;G2 是量产级工业人形机器人,现已部署在消费电子制造中。Lingxi (灵犀) 系列服务科研、教育和娱乐:X1 是面向学术研究者的开源 29 关节人形机器人,X2 和 X2 Ultra 是紧凑半身人形机器人(25–30 DOF),用于展览、互动零售和研究,起价约 $24,240。APC 2026 上,Agibot 又推出 4 个平台:A3(173 cm、55 kg,娱乐级,续航 10 小时,10 秒换电)、G2 Air(紧凑 7-DOF 单臂移动操作臂,面向人在回路作业)、D2 Max 全地形四足机器人,以及 OmniHand 3 Ultra-T(22+3 DOF 腱驱灵巧手)。D1 四足系列负责复杂户外环境巡检。产品架构遵循客户工作流逻辑:运营方先选择适合物理环境的机体形态,再匹配相应智能栈,并通过 Genie Studio 部署,尽量减少现场集成工程。[CE001, CE002, CE003, CE004, CE005, CE006]

Agibot 产品模块与资产矩阵
产品 / SKU系列目标用户状态 / 成熟度关键差异化尽调缺口
A2 UltraYuanzheng企业服务、交互商业化 — 已获 CR/CE/FCC 认证全尺寸双足,4 项国际认证企业部署的独立可靠性数据
A2 LiteYuanzheng企业服务(低成本档)商业化 — $44,560 标价相比 A2 Ultra 入门价更低与 A2 Ultra 的规格差异未完全公布
A2-WYuanzheng数据采集、工厂导航商业化轮式底盘,UniGrasp/Uni6DPose,275T 算力部署客户名单未披露
G2Genie工业制造(24/7 运行)生产 — 已部署于 Longcheer310 UPH、零错误 8 h 运行、36 h 现场集成仅一个公开记录的生产场地
G2 AirGenie轻量人机协作已于 APC 2026 发布 — GA 时间表不清7 DOF、3 kg 负载、宽度小于 800 mm、零半径转向发货日期、定价和完整规格尚未发布
G1Genie机器人数据采集机队商业化8 个摄像头、6 轴力感知、毫秒级延迟 VR 遥操作机队规模和客户部署未披露
X1Lingxi学术 / 研究工程师开源硬件29 个关节,完整 BOM/STEP/SolidWorks 开源X1 上的模型性能基准未发布
X2 / X2 UltraLingxi娱乐、教育、零售商业化 — $24,240 基础标价25–30 DOF、多模态交互、LiDAR 自动导航(Ultra)面向公众部署的独立安全认证未确认
A3Yuanzheng(下一代)娱乐、表演、教育已于 APC 2026 发布 — 计划 2026 年 4 月上市173 cm、55 kg、0.218 kW/kg、10 h 电池、10 s 更换、UWB 群体同步生产供货和定价未确认
D1 / D2 Max四足机器人巡检、户外现场作业D2 Max 达到 L3 — 已于 APC 2026 发布全地形 L3 自主四足机器人技术规格和客户部署未披露

状态与成熟度基于官方产品页、商店列表和 APC 2026 新闻稿。价格为截至 May 2026 的商店 USD 标价,可能不反映折扣后的合同价格。G2 部署数据来自公司新闻稿,并由 CnTechPost 和 Interesting Engineering 独立佐证。APC 2026 平台的 GA 时间表尚未获得官方渠道确认。

[CE001, CE002, CE004, CE007, CE009, CE010]
工作流与用例图
用户任务 / 场景现有工作流痛点Agibot 方案可衡量收益(公司声称)局限
消费电子 MMIT 测试人工上下料,按单一产品定制的刚性自动化G2 人形机器人,借助 Genie Sim 3.0 做 sim-to-real 迁移310 UPH,99.9%+ 成功率,36 h 现场集成,4 h 换型单一客户现场;缺少独立审计
面向机器人训练的工厂现场数据采集遥操作数据管线慢且割裂G1 配备毫秒级延迟 VR 遥操作,并自动同步云端1,000+ frames/device/day;本地积压 <2 h数据质量审计由内部完成;第三方 QA 未确认
展会与导览讲解静态展示;缺少自然交互A2 系列支持多模态交互和自主导航公司称客户已在 8 个应用领域采用使用率、正常运行时间和 NPS 未披露
娱乐表演与教育机器人按固定脚本执行,动作表现力不足X2/A3 具备人形步态、舞蹈、语音同步和情绪表达A3 展示 100 台机器人 UWB 同步群体表演截至报告日,A3 尚未商业化发售
复杂地形工业巡检危险或难以到达区域仍需要人工进入D1/D2 Max 四足机器人,支持 L3 自主导航公司称 D2 Max 为全球首款全地形 L3 自主四足机器人尚无独立认证或测试数据
算法开发与具身 AI 研究硬件成本高,数据管线封闭X1 开源硬件 + GO-1 开放模型 + AgiBot World 数据集1M+ 轨迹,IROS 2025 最佳论文奖入围GO-1 微调需要 ~70 GB GPU 显存(batch 16,A100)

收益指标除标注为独立佐证外,均为公司口径。Longcheer 指标由 CnTechPost 和 Interesting Engineering 佐证。其他场景主张来自官方产品页以及 CES/APC 2026 新闻稿。

[CE005, CE006, CE008, CE027, CE028]
FE001: Agibot 全栈产品架构

从机器人硬件到 AI 模型、数据平台和机队编排的五层架构。

层级构成来自官方产品页面、Genie Studio 文档、AimRT 发布说明和 APC 2026 新闻稿;部分子组件可能仅供内部使用,尚未完全公开。

[CE011, CE016, CE018, CE021, CE019]

5.2 技术架构与运营模型

Agibot 的「One Robotic Body, Three Intelligences」框架把智能拆为运动、操作和交互三层,每一层由一个或多个基于自研数据训练的基础模型支撑。运动智能由 Behavioral Foundation Model (BFM) 驱动,它从短视频演示中学习类人运动模式;Generative Control Foundation Model (GCFM) 则根据文本或音频提示实时即兴生成动作。操作智能依赖 GO-2 (VILLA),用 Action Chain-of-Thought 推理把高层任务指令翻译成精细运动命令,并由 Genie Sim 3.0 仿真平台补足,支撑快速 sim-to-real 迁移。交互智能由 WITA Omni 处理,这是一个机器人原生多模态模型,把视觉、语音和手势统一起来,而不是拼接独立语言和视觉模块。这些模型训练在 AgiBot World 数据集上;截至 March 2025,该数据集包含 1,003,672 条操作轨迹(~43.8 TB),由 100 台机器人在 5 个目标领域采集,仓库还获得 IROS 2025 Best Paper Award Finalist 提名。底层机器人中间件是 AimRT,一个开源框架,提供实时通信(1 kHz)、ROS 2 / Jazzy 互操作、zenoh 和 gRPC 后端,并保持活跃月度发布节奏(April 2026 为 v1.7.0)。SOP (Scalable Online Post-training) 框架闭合了部署循环,支持分布式车队级 VLA 模型更新:多台机器人并行执行任务,经验在云端汇总,更新后的策略在数小时内推回车队;SOP 训练后,机器人可在无人干预下自主运行超过 36 小时。Genie Studio 是集成开发者平台:它编排数据采集(最高 1,000 frames/device/day)、跨 GO-1 / RDT / Pi0 / OpenVLA 的模型训练与微调、仿真评估(8,000+ 个 3D 资产,sim-to-real 误差低于 5%),并支持云到设备一键部署,相比单 GPU 基线带来 2–3× 推理吞吐提升。[CE011, CE012, CE013, CE014, CE015, CE016]

技术与运营架构
层级 / 组件作用关键依赖风险
AimRT 中间件(v1.7.0)实时模块间通信、ROS 2 互操作、插件路由开源;由 Agibot 维护,社区参与贡献单一厂商主导的开源;社区采用广度未知
BFM / GCFM(运动模型)类人运动生成,以及由文本 / 音频提示驱动的即兴动作训练数据来自 AgiBot World 和自有动捕在演示之外,对新环境的泛化尚未验证
GO-2(VILLA)操作模型用 Action Chain-of-Thought 执行精细动作任务AgiBot World 操作轨迹;GPU 推理 ≥7 GB标准化任务操作基准未发布
WITA Omni(交互模型)统一视觉、语音、手势多模态管线端侧推理;可选云端兜底移动 / 边缘约束下的延迟未公开跑分
Genie Studio(数据 + 训练 + 评估 + 部署)端到端 VLA 开发生命周期平台自研云;兼容开放模型(RDT、Pi0、OpenVLA、GR00T)客户基于 Genie Studio API 开发,会面临供应商锁定风险
SOP 框架(机队后训练)基于真实世界经验的分布式在线 VLA 更新需要机队联网和中心化云训练基础设施大规模下的机队协调开销和收敛时间尚未验证

层级作用来自官方 Genie Studio 文档、AimRT 发布说明、APC 2026 新闻稿和 SOP 研究博客。依赖与风险评估为作者判断。性能数据除独立佐证外均为公司声称。

[CE011, CE012, CE013, CE014, CE018, CE019]
FE002: 客户部署工作流——Agibot 具身 AI 运营流程

制造客户如何借助 Agibot 全栈,从项目范围界定走到自主生产上线。

流程代表官方新闻稿所描述的 Longcheer 部署模式;具体步骤时长为 Agibot 声称,未经独立审计。

[CE022, CE024, CE026, CE027, CE019, CE020]
FE003: 关键依赖图——Agibot 技术与生态

支撑 Agibot 部署能力的关键外部依赖、合作伙伴集成和生态组件。

合作伙伴关系来源于 Aparobot 部署文章(APC 2026 报道)、官方新闻稿和 AimRT/GitHub 文档。Intel 和 TI 芯片依赖由 Aparobot 报道,未经硬件拆解独立确认。

[CE018, CE035, CE038, CE039]

5.3 部署、集成、可靠性与支持

G2 在 Longcheer Technology 位于 Jiangxi Nanchang 的平板工厂投产,是最有文档支撑的可靠性证明。借助 Genie Sim 3.0 数字孪生,Agibot 把现场调试从传统数月压缩到 36 小时,从项目启动到完成整线集成用了 4 个月。在 24/7 双班制实产运行中,G2 在每个 MMIT 任务约 19–20 秒节拍下实现 310 件 / 小时(UPH)的稳定吞吐,并有独立报道显示其在单个 8 小时窗口内完成 2,283 个任务且零错误。混线生产换型不需要定制工装;新产品型号再训练少于 4 小时,设备复用率达到 95%。Agibot 计划到 Q3 2026 将 Longcheer 部署扩展至 100 台,并延伸到汽车和半导体领域。工业部署之外,Genie Studio Agent 平台引入零代码部署工作流:运营方用拖拽式节点编辑器,把预置的感知、导航、VLA 和 RL 工具链组件组装成机器人应用,先在仿真中验证,再部署到实体硬件。软件支持受 Defined Support Period V4 文件约束:Agibot 保证自产品发布日期起至少 1 年安全更新;按 January 2026 发布日期,A2、G2 和 X2 Ultra 的更新支持截至 January 2027。Aimmaster 控制 App、PAD 软件和机器人固件也在同一窗口内。这个 1 年最低期限明显短于企业 SLA 预期,对资产生命周期更长的工业客户构成尽调缺口。[CE025, CE026, CE027, CE028, CE029, CE030]

路线图与部署里程碑
日期 / 阶段功能 / 里程碑状态影响来源
2024-12-30AgiBot World Alpha 数据集发布(92,214 条轨迹)已完成建立开放学术数据集;成为 GO-1 预训练基础GitHub OpenDriveLab
2025-03-01AgiBot World Beta 发布(1,003,672 条轨迹,~43.8 TB)已完成公共具身 AI 数据集中规模靠前;入围 IROS 2025 最佳论文奖GitHub / HuggingFace
2025-09-19GO-1 和 GO-1 Air 基础模型开源已完成降低外部微调门槛;建设开发者生态GitHub / HuggingFace
2026-01(CES)美国市场首秀;展示完整机器人产品组合;引用 5,000 台数据已完成进入国际商业化;CR/CE/FCC 认证覆盖 US/EU/CNPR Newswire
2026-03-30第 10,000 台机器人下线;G2 部署至 Longcheer已完成证明生产规模;首个 3C 精密制造具身 AI 部署CnTechPost / PR Newswire
2026-04-17(APC)A3、G2 Air、D2 Max、OmniHand 3 Ultra-T 发布;AIMA 生态启动;L1-L5 能力框架发布已发布 — 上市日期不一拓宽产品组合;把能力基准叙事制度化HumanoidsDaily / Agibot 官方
2026-Q3(目标)Longcheer G2 机队扩至 100 台;进入汽车和半导体行业计划中 — 公司口径收入和部署规模里程碑;行业多元化PR Newswire / CnTechPost
2027-01-01(政策)A2、G2、X2 Ultra 首发批次定义安全更新支持结束政策已发布 — 可能延期部署 Jan 2026 批次的客户一年后面临支持断崖Agibot Defined Support Period V4 PDF(支持期文件)

April 2026 之后的路线图里程碑为公司口径目标,尚未独立验证。数据集轨迹数量来自截至 May 2026 的官方 GitHub 仓库 README。安全支持结束日期按已发布政策;Agibot 称可能延长。

[CE015, CE016, CE026, CE031, CE032, CE033]
FE004: 产品成熟度与能力图谱

评估 Agibot 主要产品线在五个能力维度上的成熟度。

成熟度评估是作者基于截至 2026 年 5 月公开证据的判断。AGIBOT L-scale 等级由公司定义(APC 2026 发布),未经独立审计。部署证据评级反映可用证据的深度和独立性,而非商业收入。

[CE002, CE003, CE006, CE009, CE016, CE036]

5.4 差异化与开发者生态

Agibot 的核心技术差异化来自数据飞轮、开放开发者生态,以及硬件、模型、部署工具链的垂直整合。AgiBot World 数据集包含来自 100 台机器人、横跨 5 个领域的 1M+ 条轨迹,并配套开源 GO-1 基础模型(September 2025 发布)和 HuggingFace 上的训练脚本,是已公开的最大具身 AI 数据集之一,也让 Agibot 成为学术界和开发者的参照对象。GO-1 Air 轻量版本(不含 Latent Planner)降低了边缘硬件推理门槛。Agibot 的开源中间件 AimRT 为外部工程师提供有文档的集成路径,并以月度发布节奏推进到 v1.7.0,显示上游维护真实存在。APC 2026 上,Agibot 正式提出一个类似 SAE 自动驾驶分级的 L1–L5 能力框架,并声称当前系统在运动和任务智能上达到 L3(场景级自主);这个基准虽由公司自定义,但给了其结构化的差异化叙事。「Powered by AGIBOT」计划让系统集成商可以在 Agibot 量产底座上定制硬件和软件,降低技术门槛(例如意大利 SIR Spa 用于制造 / 汽车,美国 Michigan 的 Hyperscale Data 用于北美工业部署)。Agibot 的 OmniHand 3 Ultra-T 灵巧手(22+3 DOF、500 g、3D 触觉感知)以及 A3 的 UWB 厘米级群体定位,可支持 100 台机器人同步表演,这些硬件特性多数同行尚未匹配。Peng Zhihui 担任中国 MIIT HEIS 人形机器人标准化委员会副主任,也让 Agibot 直接影响正在形成的国家监管框架。[CE015, CE016, CE017, CE018, CE034, CE035]

5.5 信任、安全、隐私与合规

Agibot 已搭起基础合规设施,但在工业规模部署所需的产品安全和网络安全深度上仍有缺口。认证方面,A2 Ultra 持有 CR(中国)、CE-MD(欧盟机械指令)、CE-RED(欧盟无线电设备指令)和 FCC(美国)认证,可在三大主要市场商业销售。G1 和 G2 配备急停按钮、碰撞保护、底盘避障,以及故障检测后的即时报警和停止;G2 还在双臂嵌入六轴力感知,用于受控施力。中国 MIIT HEIS 框架(February 2026 发布)由 Agibot 通过 Peng Zhihui 的委员会参与共同制定,要求限力、急停机制、热管理,以及软件故障下的最低风险状态行为。关键问题是,Peng Zhihui 承认,人类优于传统自动化的操作任务中,近 80% 与触觉感知强相关;这一能力在全行业都尚未充分标准化,技术也不成熟,Agibot 自身产品同样如此。数据隐私方面,Agibot 网站隐私政策(April 30, 2026 生效)指定 AGIBOT PTE. LTD.(新加坡)为多数国际市场的数据控制者,日本 / 韩国(Agibot K.K.)和 Hong Kong SAR 则由不同实体负责。该政策遵循合法、正当、必要和诚信原则,但没有发布漏洞披露计划,也没有 SOC 2 / ISO 27001 认证。1 年安全更新底线是企业买家的尽调缺口。[CE003, CE040, CE041, CE042, CE032, CE033]

信任、质量与合规控制
控制 / 认证状态范围缺口 / 尽调问题
CR(中国强制认证)已获得 — A2 Ultra中国大陆商业销售G2 或 X2 系列未确认
CE-MD(EU 机械指令)已获得 — A2 UltraEU 商业销售G2 工业机器人 CE-MD 状态未确认
CE-RED(EU 无线电设备指令)已获得 — A2 UltraEU 商业销售(无线电 / 无线)覆盖机型范围未披露
FCC(美国射频)已获得 — A2 Ultra美国商业销售G2 FCC 状态未确认;若缺失,美国进口存在风险
安全更新支持(V4 政策)自发布起至少 1 年;A2/G2/X2 Ultra 至 Jan 2027A2、G2、X2 Ultra、Aimmaster、App、PAD 软件1 年最低期限低于企业工业生命周期预期(通常 3-5 年)
MIIT HEIS 人形机器人标准(2026 版)参与共建 — Peng Zhihui 任委员会副主任限力、急停、热管理、最低风险状态标准发布于 Feb 2026;合规测试 / 认证路径尚未定义
隐私政策(Apr 30, 2026 生效)已发布 — AGIBOT PTE. LTD.(新加坡)为数据控制者日本 / 韩国 / 香港特别行政区以外所有市场的网站用户漏洞披露计划或 SOC 2 / ISO 27001 均未确认
硬件安全(G1/G2)已部署 — 碰撞保护、急停、故障告警已部署的 G1 和 G2 设备已部署机型没有第三方安全审计报告

A2 Ultra 认证状态来自 Agibot 官方产品页和 CES 2026 新闻稿。G2 认证状态为从缺推断(未找到官方确认)。安全更新日期来自 Agibot Defined Support Period V4 PDF(Jan 2026 发布)。合规缺口评估为作者判断。

[CE003, CE032, CE033, CE035, CE040, CE041]

5.6 展示材料

Chapter 06

06客户情况

6.1 客户分层与收入架构

商业发布阶段,AGIBOT 面向 4 类核心客户:工业制造(以电子装配和汽车供应链为主)、物流与仓储(抓取放置、分拣)、酒店与零售服务(酒店运营、豪华水疗、清洁),以及企业 IT 或数据中心运营。工业制造是最成熟的分层,由 Longcheer Technology 生产部署锚定。制造业买方通常是工厂运营或制造工程团队;物流买方是设施或运营负责人;酒店业买方则是度假村 / 酒店总经理或采购部门。 AGIBOT 同时运行两种收入模型。第一种是机器人直接购买或车队租赁,企业一次性采购 G2 或 A2 设备,或签订多年期租赁安排。第二种也是战略上更重要的一种,是机器人即服务(Robot-as-a-Service,RaaS):March 2026 公司在 Mobile World Congress 通过 store.agibot.com 推出该服务,价格从每台机器人每天 €899 起,未披露最低合同期限。截至 March 2026,RaaS 模型覆盖 17 个国家和地区,旨在降低国际客户采购门槛,让不能承诺大额前置硬件支出的客户也能试用。商店还列出 C5 清洁机型等可直接购买机器人,价格 $32,900,说明其商业化路径同时覆盖自助式数字渠道和企业直销渠道。 发布初期的地域集中度明显偏向中国生产部署;东南亚(Genting Malaysia)、欧洲(SIR Spa Group,Italy)和新加坡(Singtel RaaS 试点)代表早期国际市场测试。March 2026 提出的 American Security Robotics Act 拟禁止联邦机构使用中国制造机器人,美国政府采购因此成为结构性受阻渠道。[CU001, CU002, CU003, CU004, CU005, CU006]

客户分层表
客户分层买方 / 付款方主要用例地域重点收入模式部署状态(May 2026)证据质量
工业制造工厂运营 / 采购电子装配、上下料、混线生产中国(主要)直接采购 / 机队租赁量产部署(Longcheer)高 — 联合新闻稿 + 客户官网
物流 / 仓储设施 / 运营负责人拾放、分拣、数据中心运营美国(Hyperscale MI)、中国RaaS / 直接采购试点 / 已签协议(Hyperscale Data)低 — 仅单一行业媒体来源
酒店与服务酒店总经理 / 度假村采购住客服务、主题公园运营、水疗协助马来西亚(Genting)、意大利(SIR Spa)RaaS(€899/day)合作备忘录(Genting Malaysia、SIR Spa)低 — 仅 MOU;无部署时间表
电信 / 企业科技IT / 服务运营网络维护、巡检服务新加坡(Singtel)RaaS 试用试点(Singtel)低 — 单一行业媒体引用
美国政府 / 联邦联邦机构采购通用自主任务执行美国采购受拟议法案阻断结构性受阻(American Security Robotics Act)反向 — 立法推进中

分层状态反映截至 May 2026 的公开证据。部署状态基于已宣布的合作和新闻稿;尚未公开确认的私有部署不纳入。

[CU001, CU003, CU004, CU006, CU035]
FU001: 客户旅程图

映射 AGIBOT 客户从初始认知到企业生产部署,以及跨细分市场扩展 RaaS 的采用路径。

[CU012, CU015, CU024, CU033]

6.2 采用轨迹与部署规模

按单位数量看,AGIBOT 的采用轨迹是人形机器人行业中最陡的一批。Humanoid Index 引用的 Omdia 数据显示,AGIBOT 在 2025 年按出货量全球第一,交付约 5,100 台;同期没有传统厂商或新进入者达到这一数字。到 March 2026,累计产量已跨过 10,000 台门槛,公司在 April 2026 的年度伙伴大会 (APC 2026) 上重点强调这一里程碑。公司提出「部署元年」定位,意味着战略从产品开发转向商业规模化;但公开信息尚未区分终端客户出货、内部测试设备、库存或经销商存货。 Longcheer 部署提供了最具体的采用证据。集成在 36 小时内完成,部署约 4 个月从试点爬升到 24/7 生产排程——如果这种周期可复制,将支撑企业级快速规模采用。AGIBOT 计划到 Q3 2026 将 Longcheer 车队扩至 100 台 G2。每小时 310 台产品吞吐和 99.9% 动作成功率,是截至 mid-2026 全球任何人形机器人生产部署中披露的最高生产率指标。APC 2026 上,来自 17 个国家的全球客户参加了伙伴大会,说明国际参与度在上升——但参会不等于商业合同转化。 AGIBOT 在 2025 年的 5,100 台约占 IDC 预计 2030 年 510,000 台人形机器人规模的 1%(~95% CAGR)。这说明总可用市场(TAM)仍有显著空间,同时也伴随很高的执行风险。Goldman Sachs 估计,全球人形机器人市场将在 10–15 年内达到 $6 billion,乐观情景到 2035 年可达 $154 billion。[CU008, CU009, CU010, CU011, CU012, CU013]

客户增长 / 采用轨迹表
指标数值日期来源置信度含义缺失分母
出货量(2025)~5,100Year-end 2025Omdia 经 Humanoid Index按出货量全球 #1活跃部署设备 vs 库存
累计产量(累计)~10,000March 2026humanoidsdaily.com达到 10k 里程碑;规模化轨迹明确创收设备 vs 总产量
Longcheer 产线吞吐量310 UPHApril 2026AGIBOT/Longcheer PRNewswire商业可行的生产率底线产能上限和已部署产线数
Longcheer 计划机队规模100 台 G2Q3 2026 目标AGIBOT/Longcheer PRNewswire当前机队的 3–4×,意味着 ACV 扩大合同金额和付款节奏
RaaS 覆盖地域17 个国家 / 地区March 2026finance.sina.com.cn(MWC 来源)全球商业渠道已启动付费 RaaS 客户 vs 免费试用

所有数字均由公司披露或来自二级分析机构;目前没有公开的、经独立审计的收入或设备部署数据。

[CU008, CU009, CU010, CU011, CU014]
FU002: 采用 / 部署漏斗

示意性部署漏斗,从已出货单元到已确认生产部署,显示总量与已验证商业使用之间的证据缺口。

「具名商业伙伴」(20+)为 AGIBOT 在 APC 2026 的公司声称数字,未经独立验证。其他所有数字均来自 Omdia(经 Humanoid Index)和公开公告。漏斗展示从产量到已确认商业使用时证据质量逐级下降;实际商业管线可能大于公开披露。

[CU005, CU009, CU014]

6.3 具名客户证明——生产与试点

截至 May 2026,Longcheer Technology 是 AGIBOT 唯一已确认的生产级客户。Longcheer 总部在中国,是 Lenovo ODM 和电子制造合作伙伴,已在其 Shenzhen 和 Dongguan 的 MMIT 消费电子精密装配线上部署 AGIBOT G2 机器人。部署指标——310 UPH 吞吐、99.9% 动作成功率、19–20 秒节拍、100% 准时交付、每天运行 16 小时且其中 8 小时自主运行——由 AGIBOT 和 Longcheer 在 PRNewswire 公告中联合发布,并得到 Longcheer 官网佐证;后者将自己描述为 AI 赋能的智能制造伙伴。Longcheer 在 April 2026 获得 Lenovo 的 2026 Perfect Quality Award,说明性能提升可能强化该部署的留存理由。 Longcheer 之外,AGIBOT 的 APC 2026 伙伴大会宣布了一系列 MOU 和 RaaS 合作。Genting Malaysia 签署了 MOU(New Straits Times,April 2026 确认),计划在 Genting Highlands 的酒店和主题公园运营中部署人形机器人。Genting Malaysia 官网描述了其在全球度假村运营中的活跃扩张策略。Hyperscale Data 是美国 Michigan 一座 100,000-square-foot 数据中心设施的运营方,已与 AGIBOT 签署合作协议。第三方来源还通过 RaaS 项目提到 SIR Spa Group(Italy)和 Singtel(Singapore)为其他早期合作伙伴。截至 May 2026,这些非 Longcheer 部署都没有确认进入生产阶段。AGIBOT 声称 APC 2026 有「20 多家头部企业」参与,但没有披露完整名单。 AGIBOT 官网发布的视频证言和客户 Q&A(APC 2026 报道)显示,来自不同行业的国际参会者正面评价了大会,但公开记录中没有任何非 Longcheer 客户披露具体部署指标或合同价值。BYD 和 JD.com 以战略投资者身份出现,可能有部署兴趣,但尚未确认是活跃生产客户。[CU016, CU017, CU018, CU019, CU020, CU021]

具名客户证据表
客户分层部署 / 用例状态(May 2026)关键结果指标证据来源局限
Longcheer Technology消费电子制造深圳和东莞 MMIT 精密装配线量产部署 — 24/7 运行310 UPH;99.9% 动作成功率;计划 Q3 2026 达 100 台联合 PRNewswire;Longcheer.com;AGIBOT detail/60单一现场 / 产线类型;合同金额未披露
Genting Malaysia Bhd酒店 / 主题公园Genting Highlands 度假村运营(酒店、主题公园)MOU — 无约束力无生产指标New Straits Times;Genting Malaysia 公司官网仅 MOU;部署时间表未披露
Hyperscale Data物流 / 数据中心美国 Michigan — 100k sqft 设施运营已签协议 — 试点阶段未披露指标aparobot.com year-one-of-deployment 文章无独立佐证;无指标
SIR Spa Group奢华服务 / 美容意大利水疗服务协助MOU / 试点未披露指标aparobot.com year-one-of-deployment 文章单一来源;无独立佐证
Singtel电信 / 企业服务新加坡 — 服务运营 RaaS 试用试点 / RaaS 试用未披露指标aparobot.com year-one-of-deployment;TechInAsia(付费墙)付费墙来源;无已确认生产指标

仅列出公开公告中具名的客户。AGIBOT 在 APC 2026 称拥有 20+ 家企业合作伙伴,但尚未发布完整名单。

[CU016, CU017, CU018, CU019, CU020, CU024]
FU003: 客户证据矩阵

从证据质量、结果具体性、留存可见度、生产成熟度四个维度,评估五个具名客户的客户证据质量。

[CU016, CU021, CU026, CU031, CU032]

6.4 留存、耐久性与扩张

AGIBOT 没有披露任何净留存率(NRR)、总留存率(GRR)、流失率、合同续约率或客户满意度指标。市场上也未发现 AGIBOT 产品的 G2 评论、Gartner Peer Insights 评分或第三方客户满意度平台数据。因此,所有留存信号都只能从结构性指标推断,而不是从已披露数据实测得出。 最强的结构性留存信号是 Longcheer 部署的扩张轨迹:从试点扩到 100 台车队,意味着先落地再扩张的模型,即初始部署表现带动有机车队增长。Longcheer 从 4 个月试点转入生产,也说明 AGIBOT 的现场工程支持足以促成试点转化,这是客户耐久性的早期关键指标。不过,RaaS 模型的定价(€899/day,未披露最低期限)留下了低切换成本终止的可能,这会意味着早期商业客户队列流失率高于平均水平。 Genting Malaysia MOU 是无约束力的意向书,没有披露合同价值、时间表或 SLA 承诺。人形机器人行业的 MOU 经常无法在 12 个月内转化为已签署生产合同,因此 Genting 和类似关系不适合作为耐久性经常性收入的证据。AGIBOT 当前商业阶段的整体留存状态最适合概括为「结构性不确定,但方向偏正面」:Longcheer 扩张计划令人鼓舞,但单一标杆客户无法证明可复制模式。[CU028, CU029, CU030, CU031, CU032, CU033]

留存 / 重复使用 / 满意度表
指标数值 / 状态客户分层置信度尽调问题
净收入留存率(NRR)未公开披露全部企业客户向 AGIBOT IR 索取披露;与 SaaS/RaaS 同业对标
总收入留存率(GRR)未公开披露全部企业客户从合同续约推断;要求第一方确认
合同续约率未公开披露制造业(Longcheer)与 Longcheer 采购团队做结构化客户访谈
客户 NPS / 满意度评分未找到 G2、Gartner Peer Insights 或第三方评分全部细分市场搜索评价平台;向 APC 2026 参会方索取客户访谈机会

公开渠道未披露留存、流失或满意度数据。所有数值均表示:截至 2026 年 5 月,经穷尽搜索仍未找到公开可用指标。

[CU028, CU039]
FU004: 留存 / 复购队列

说明 AGIBOT 当前商业状态的关键客户侧指标;留存队列数据不可得——所有 NRR/GRR/流失值均为 null,等待一手披露。

NRR 和 GRR 显示为 null,因为 AGIBOT 尚未公开披露这些指标。出货量和客户数来自第三方报道。没有时间序列留存数据,就无法搭建队列图;此处按 plannedFigures 契约提供 KPI 替代图,并附 figureType 说明。

[CU009, CU011, CU004, CU028, CU039]

6.5 集中度风险与不利因素

AGIBOT 的商业订单簿高度集中。截至 May 2026,Longcheer Technology 是唯一生产级客户;按任何合理口径,商业生产收入的客户集中度都是 100%。如果 Longcheer 因运营、竞争或成本原因暂停或终止部署,AGIBOT 将失去唯一已披露的商业生产参照。这种集中度对一家刚开始商业发布的早期深科技公司并不罕见,但对尽调很重要,必须用客户多元化路线图和管线转化证据来回应。 Senators Cotton 和 Schumer 在 March 2026 提出的 American Security Robotics Act 拟禁止美国联邦机构采购或使用中国制造机器人。该法案不影响商业企业采购或直接面向企业的 RaaS 订阅,对 AGIBOT 现有客户订单簿的短期影响有限。但它释放了更广泛的美国监管和采购风险信号,影响 AGIBOT 长期进入美国政府市场的战略。 人形机器人部署中的法律责任仍是未解决的系统性风险。Hill Dickinson 的 2026 年分析指出,产品责任、职业健康与安全义务、运营方责任,都是企业客户在工厂地面部署人形机器人时的重大顾虑。这些责任缺口可能拉长采购周期,并要求 AGIBOT 提供明确赔偿或保险产品。中国 MIIT 预计在 2026 年推出的人形机器人国家安全标准,也可能增加认证要求,影响新客户部署前置时间。AGIBOT 的 APC 2026 伙伴生态模型通过全球行业伙伴渠道销售标准化生产力解决方案;如果 MOU 到已签生产合同的转化率偏低,该模型会产生渠道依赖风险。[CU035, CU036, CU037, CU038, CU039, CU040]

扩张与集中度风险表
因素当前状态影响程度尽调路径
Longcheer 先落地再扩张已确认从试点 → 量产 → 2026 年 Q3 达到 100 台规模高上行空间 —— 验证扩张模型确认 100 台规模下的单位经济性;获取合同扩张条款
单一客户量产集中度Longcheer = 已知唯一量产客户(100% 集中度)高风险 —— 收入完全依赖单一客户要求提供管线数据:90/180 天内从试用转为量产的数量
美国政府采购封锁American Security Robotics Act 提案已提出(2026 年 3 月)—— 尚未生效中等风险 —— 阻断美国联邦采购渠道跟踪立法进展;量化联邦 TAM 占总可用市场的比例
APC 2026 MOU 转化率已签署多份 MOU;截至 2026 年 5 月,确认转为量产的数量为零中等风险 —— 若合作方未能签下商业订单,将形成渠道依赖要求 AGIBOT 提供 60/90/180 天 MOU 至合同的转化指标

集中度分析基于公开可得客户证据。公开渠道看不到的销售管线和转化数据,可能实质改变风险判断。

[CU032, CU033, CU035, CU038]

6.6 展示材料

Chapter 07

07风险

7.1 监管与法律风险

American Security Robotics Act (ASRA) 由参议员 Tom Cotton (R-AR) 和参议员 Chuck Schumer (D-NY) 于 26 March 2026 提出,是 Agibot 国际扩张面临的最重大近期监管风险。这项两党法案以数据外流和远程控制潜力带来的国家安全担忧为由,拟禁止美国联邦政府采购或运营中国公司制造的人形机器人。Reuters 和 US News 确认,法案明确点名包括 Agibot 和 Unitree 在内的中国人形机器人创业公司。众议员 Elise Stefanik 同时在美国众议院提出配套法案。即便 ASRA 不按现有形式成为法律,它的提出也显示立法趋势可能扩展到州政府、国防承包商、受监管行业和大型企业采购。任何接入美国政府相邻工作流的 Agibot 机器人——物流、基础设施、医疗——都会面临更高监管审查。 中国法律基础设施构成第二层、且结构不同的法律风险。《国家情报法》(2017) 要求所有中国实体在被要求时协助国家情报工作。《个人信息保护法》(PIPL, 2021) 和《网络安全法》(2017) 则设置数据本地化义务、跨境数据传输强制安全评估以及政府访问权。Agibot 的网站隐私政策于 30 April 2026 生效,披露会从机器人运营中收集设备标识符、使用数据和技术信息。对美国、欧盟、日本和韩国的企业客户而言,在运营环境中部署 Agibot 机器人会引出一个问题:产线数据——可能包括专有装配流程、质量控制数据和工厂布局——是否可能被中国国家机关访问。Agibot 没有公开披露如何同时满足中国法律和客户数据主权要求,对数据流进行分段、本地化或限制。 全球范围内,人形机器人的产品责任仍未定型。Hill Dickinson 在 2026 年发布的法律分析提出核心问题:当自主式人形机器人造成损害时,责任可能落在制造商、运营方、软件提供商,或几方共同承担。现有产品责任制度——欧盟产品责任指令、美国州级产品责任法和中国《产品质量法》——并不是为会实时决策的自主具身 AI 系统设计的。美国 OSHA 承认,目前没有适用于机器人行业的具体 OSHA 标准。ISO 10218-1:2011 是前代工业机器人安全标准,适用于固定式工业机器人,不适用于在开放环境中运行的双足人形机器人。Agibot 机器人部署在 3C 电子制造、汽车和服务环境,使用场景天然要求与人近距离接触。能力和安全法律框架之间的缺口,可能在各司法辖区制定具体人形机器人监管时,引发追溯性合规义务、保险不确定性和客户采用阻力。 中国 MIIT 在 February 2026 发布《人形机器人与具身智能标准体系》;Agibot 联合创始人兼 CTO Peng Zhihui 以 HEIS 技术委员会副主任身份参与编写。Agibot 参与标准制定,对中国市场合规是正面信号,但也带来风险:由国内主导玩家塑造的标准,可能嵌入有利于 Agibot 国内竞争的技术要求,同时对外国竞争者形成事实上的非关税壁垒,从而引发 WTO 挑战和关键市场的对等限制。EU AI Act (Regulation 2024/1689) 将于 August 2026 全面适用,把用于安全关键或高风险场景的 AI 系统归类为「高风险 AI」,可能要求部署在欧盟工业环境中的 Agibot 机器人完成符合性评估、CE 标记和注册。 [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发生概率严重性缓释措施剩余敞口尽调路径
American Security Robotics Act(ASRA)—— 禁止联邦采购中国制造的人形机器人美国(联邦)Cotton 和 Schumer 于 2026 年 3 月提出法案;Stefanik 在众议院提出配套法案;委员会审议仍在进行极高只要所有权和来源仍属中国,Agibot 没有公司层面的缓释手段;目前进入美国政府市场的空间有限若 ASRA 通过,任何美国政府或国防邻近客户的采购都会受阻;企业客户也可能提前自我限制跟踪 ASRA 立法进展;确认 Agibot 是否聘请美国监管或政府事务顾问;评估对美国收入的依赖
中国实体依据《国家情报法》承担的数据访问义务中国(对中国注册子公司有域外适用)义务已生效;没有退出机制;不能通过与海外客户签约豁免确定极高若不在中国境外重组公司架构,就没有结构性缓释手段;数据架构隔离可以缩小范围,但无法消除法律义务受监管行业的企业客户(国防、制药、政府供应链)可能拒绝部署;这会实质阻碍进入美国和欧盟政府邻近市场要求出具关于 PIPL / 《国家情报法》是否适用于机器人采集运营数据的法律意见;获取客户对数据流架构的法律审查
EU AI Act(Regulation 2024/1689)—— 工业场景人形机器人的高风险 AI 系统合规欧盟(2026 年 8 月起适用于所有成员国)法规自 2026 年 8 月起适用;高风险 AI 需满足合格评定、CE 标志、技术文档和人工监督义务可以启动 CE 标志项目;截至 2026 年 5 月,Agibot 未披露任何欧盟合格评定接洽国际收入若要扩入欧盟工业市场,需要 Agibot 尚未确认搭建的合规基础设施确认 Agibot 是否已启动欧盟合格评定;要求提供欧盟监管路线图及任何公告机构接洽情况
人形机器人造成人身伤害的产品责任 —— 制造商、运营方、软件提供商如何分担美国、欧盟、中国及全球主要市场法律上尚未定型;现有产品责任框架并非为自主具身 AI 设计;尚无标志性判例与集成商和最终运营方在合同中分配责任,可部分转移风险;赔偿条款属于标准配置一旦 Agibot 机器人发生首起重大工作场所事故,可能引发监管调查和集体诉讼敞口审查 Agibot 客户合同中的责任分配、赔偿上限和保险要求;确认 Agibot 的产品责任险覆盖范围
MIIT 人形机器人标准体系(2026 版)合规认证要求中国(国内;影响「一带一路」和全球南方市场)标准体系于 2026 年 2 月发布;实施时间表和认证机构指定仍在推进Agibot CTO 是 HEIS 副主任;国内合规构成竞争优势认证要求可能偏离 ISO/IEC 标准,给国际业务带来双重合规负担确认 Agibot 在 HEIS 框架下的认证状态;评估认证机构获批时间表
欧盟部署中机器人采集的生物识别和运营数据跨境传输与 GDPR 合规欧盟(GDPR)和英国(UK GDPR)GDPR 已生效;生物识别数据保护要求更高;向中国跨境传输数据需要充分性认定或标准合同条款(SCC)可部分采用隐私设计原则和数据本地化选项;Agibot Privacy Policy 涵盖部分条款若机器人采集数据在缺乏充分法律依据下传至中国,欧盟企业客户会暴露于 GDPR 合规风险审查 Agibot Privacy Policy 是否满足 GDPR;确认面向中国数据流是否已建立标准合同条款

各行按严重性排序。ASRA 和《国家情报法》是 Agibot 国际商业可服务市场中最重大的两项监管和法律风险。截至 2026 年 5 月,公开来源未发现针对 Agibot 的诉讼。产品责任风险仍处潜伏期,会在首起重大部署事故发生时显性化。

[CR001, CR002, CR003, CR004, CR005, CR006]
FR001: 风险热力图 — Agibot 主要风险的可能性与严重程度
[CR001, CR005, CR006, CR009, CR010, CR014]

7.2 技术与产品风险

Agibot 从 5,000 台到 10,000 台只用了 3 个月,生产加速在运营上令人印象深刻,但也引入外部资料难以评估的规模化质量风险。Longcheer Nanchang 平板工厂 8 小时零错误运行(完成 2,283 个任务)是有说服力的数据点,但它只代表单一设施、单一部署,在 3C 装配线上执行定义明确的一部分抓取放置任务。当每个客户工业化到 100+ 台,并跨入汽车、半导体和服务环境后,环境条件、任务复杂度、边缘案例和交互场景都会出现差异,单一试点无法预判。公开资料中没有 Agibot 量产规模下的平均故障间隔时间(MTBF)、现场退货率或质量逃逸率数据。 Agibot 商业机器人的官方安全支持期——A2、G2 / G2 Air 和 X2 Ultra——从 January 2026 发布日起只有 12 个月(截至 January 2027)。软件产品(Aimmaster、App、PAD)同样只有 12 个月安全更新窗口。企业客户把人形机器人接入长周期工业流程时,资本设备通常预期运行 5–10+ 年;12 个月安全补丁窗口明显不够。它制造了合同和责任缺口:January 2027 之后,在生产环境中运行过期固件的机器人会成为网络安全攻击目标,运营方独自承担未缓释风险。Agibot 没有公开声明长期支持义务、延长服务协议,或定义期限之外的安全更新治理。 Agibot 把 Genie Sim 3.0 数字孪生定位为快速部署的关键抓手,可把传统数月现场调试压缩到 36 小时。这在受管理部署中确实是竞争优势,但也引入新的单点故障:如果 Genie Sim 基础设施宕机、重新定价或无法使用,现有部署管线就会被打断。MIIT 标准也指出,触觉感知仍是关键瓶颈;Peng Zhihui 本人在 MIIT 委员会承认,高价值操作任务中近 80% 依赖尚未标准化的触觉反馈。没有标准化触觉感知,机器人面对新颖或未表征物体时,错误率可能显著高于受控演示中的报告水平。 Humanoid Index 确认的 IDC 口径称 2025 年出货 5,200 台,但竞争对手 Unitree 对此提出异议,声称其人形机器人出货 5,500 台,并申请 USD 580 million IPO。这场出货量争议说明,行业产量数字没有经过审计,也无法独立验证。如果 Agibot 的产量主张被夸大,或指向已制造但未交付设备,支撑投资逻辑的商业证据就会减弱。这个验证缺口在任何阶段都是重大尽调风险。 [CR010, CR011, CR012, CR013, CR014, CR015]

运营 / 质量 / 安全风险登记表
故障模式发生概率严重性缓释成熟度剩余敞口未解决缺口
上市后 12 个月安全更新支持到期(2027 年 1 月停止支持)确定(已承诺政策)低 —— 公开渠道未见延长支持选项或长期维护协议2027 年 1 月之后仍在运行的已部署机器人将带着未修补安全漏洞;企业部署周期通常为 5–10 年未公布延长安全支持、企业支持协议或安全更新延期计划的路径
产量超过 10,000 台后制造质量下滑部分 —— Genie Sim 数字孪生加速部署校准;未披露规模化后的公开质量 KPI规模化后现场故障率上升,会在关键商业化阶段带来质保敞口、客户流失和声誉损害未公布 MTBF、现场退货率或质量逃逸数据;单一 Longcheer 部署不足以支撑规模化可靠性主张
美国出口管制扩大并瞄准中国来源系统,AI 推理芯片供应中断低 —— 未披露已确认的非中国 AI 芯片来源;BIS 规则自 2022 年以来持续扩大若首选 AI 加速器供应受限,生产可能停摆或性能下降;未披露替代芯片策略确认 Agibot 使用哪些 AI 推理芯片;评估 EAR 分类;要求提供供应链冗余文件
触觉感知缺口导致 3C 装配以外的精密任务出现操作错误低 —— Agibot CTO 承认这是行业级瓶颈;尚无标准化解决方案汽车、半导体和手术邻近环境中的任务失败率,可能显著高于 Longcheer 数据暗示的水平要求提供不同物体类型和非结构化环境下的任务成功率数据;确认触觉传感器校准协议
Genie Sim 数字孪生基础设施不可用,打断新部署集成中 —— 云基础设施可能有冗余,但公开渠道未说明新客户集成延迟;现有部署或可运行,但重新训练和换线需要仿真访问确认 Genie Sim 基础设施的 SLA、冗余架构和数据中心司法辖区
部署中发生造成人类操作员身体伤害的产品安全事故极高中 —— MIIT 安全标准框架提供基线要求;人形机器人部署继承工业安全风险在具名客户处发生单起严重事故,就会带来集体诉讼风险、监管调查和客户退出确认 Agibot 产品责任险覆盖范围;审查客户安全协议和人机界面设计文件

各行按严重性排序。12 个月安全支持窗口有明确日期,是最可量化的运营风险。制造质量和供应链风险具有结构性,并会随产量扩大而上升。触觉感知缺口由 Agibot CTO 自己披露为行业级限制。

[CR010, CR011, CR012, CR013, CR014, CR016]
FR002: 风险传导图 — Agibot 风险如何流向收入与估值
[CR001, CR006, CR019, CR024, CR025, CR030]

7.3 运营与供应链风险

Agibot 从手工式生产爬坡到 10,000 台,用了 4 年开发。下一阶段要扩到 50,000–100,000 台,以服务其客户扩张目标隐含的工业部署规模,难度会高出一个量级。人形机器人制造涉及数百个精密机械部件、自研执行器组件、力敏关节编码器和集成神经处理单元,所有这些都必须达到工业部署所需的车规级可靠性规格。在当前产量下,单机经济性几乎肯定深度为负;要通过制造降本曲线达到商业可行利润率,需要投入专用工装、供应链精简和良率提升,但这些尚未得到公开确认。 中国供应链集中度形成结构性脆弱性。关键部件类别——锂电池包(CATL 是战略投资者和国内供应方)、AI 加速芯片(受美国先进半导体出口管制限制)以及精密执行器部件——都由中国供应商主导。美国对先进半导体制造设备和芯片的出口管制(依据 Bureau of Industry and Security 执行的 Export Administration Regulations)自 2022 年以来持续收紧。若限制扩展到 Agibot 嵌入式系统使用的 AI 推理芯片,或中国半导体代工厂面临额外限制,而 Agibot 又没有多元化供应链替代方案,就可能遭遇生产中断。公司没有公开披露供应链冗余或非中国零部件来源。 运营部署模型依赖 Genie Sim 为产线做数字孪生校准,把机器人接入实产线,因此需要持续技术支持基础设施:现场集成工程师、远程诊断连接、固件更新机制和现场服务网络。IFR May 2026 报告指出,中国现已占全球工业机器人安装量的 54%,凸显中国本土机器人服务网络竞争强度。中国之外,Agibot 已在 CES 2026 宣布市场进入,并拿到 LG Electronics 作为韩国战略投资者,但没有披露美国、欧洲或其他国际市场的现场服务或支持基础设施。没有本地服务基础设施的国际部署,会让客户暴露在更长平均修复时间(MTTR)之下,并提高部署失败、损害 Agibot 标杆案例质量的风险。 [CR018, CR019, CR020, CR021, CR022, CR023]

7.4 合作伙伴、依赖与集中度风险

Agibot 在中国最高层的能见度很高——包括 2026 年习近平主席上海考察期间进行演示——这强烈指向战略背书和国内市场准入。它同时也是地缘政治集中风险。只要公司创始人的政治地位恶化、中央政府优先级从人形机器人转移,或中美技术紧张升级,支撑快速审批、资本获取和政府采购的优惠政策环境都可能削弱。这种政治背书也会加深西方市场对“嵌入国家体系的机器人”的认知,可能推动立法限制在 ASRA 框架之外加速扩大。 Longcheer Technology 是目前文档最充分的商业客户,也是 Agibot “Deployment Year One” 叙事的锚点案例。与 Longcheer 的关系包括全球首个工业规模 3C 具身 AI 部署,以及公开宣布的计划:到 2026 年 Q3 在 Longcheer 场景扩展至 100 台。这个客户加合作伙伴结构是积极信号,但也意味着主要外部部署证据由单一客户支撑。若 Longcheer 推迟扩张、降低投入,或自身业务遇到困难,Agibot 会在最需要给下一波企业客户提供可引用证明的时点,失去最醒目的商业参考案例。 Agibot 的投资人名册——来自中国企业登记数据披露——包括 BYD、SAIC、BAIC Group、JD.com、Tencent、LG Electronics 和 Mirae Asset,以及风险投资机构(Hillhouse、Sequoia China/HongShan、Bluerun、Matrix Partners China)。带国资色彩的投资人(BAIC Group 多数股权国有;SAIC 国有;BYD 获得显著国家补贴)能带来客户入口和监管支持,但也引入治理问题。拥有生产运营场景的投资人客户,可能拿到独立企业客户在正常商业条款下无法复制的优惠部署条件和定价。美国 CFIUS 审查机制会审查具有潜在国家安全影响的外国投资和收购;Agibot 的投资人结构包含国企资本和 LG(韩国)资本,若进入美国市场或设立合资公司,会引发审查关注。 [CR024, CR025, CR026, CR027, CR028, CR029]

合作伙伴 / 依赖风险登记表
依赖项对手方角色集中度失效情景严重性缓释措施剩余敞口
中国国家层面政治背书(习近平考察)中华人民共和国 / 中国共产党市场准入、政府采购优待、监管优先待遇、IPO 便利极高 —— 没有可替代的同等级国内政治背书来源中国共产党产业优先级转向;公司领导层发生政治风险事件;美中紧张升级,抬高该背书在西方市场的声誉成本极高公司定位为商业企业,弱化直接依赖国家的观感;多元化投资人基础提供部分缓冲政治风险没有可用缓释手段;美中贸易冲突任何升级,都会抬高习近平背书在西方市场的商业成本
Longcheer Technology(锚定客户)Longcheer Technology(603341.SH —— 上海上市)唯一具名的量产规模商业客户;主要部署参考案例高 —— 公开证据中唯一量产部署;未披露其他具名量产客户Longcheer 推迟 2026 年 Q3 100 台扩张;Longcheer 自身出现财务或运营困难;Longcheer 转向竞争方案到 2026 年 Q3 扩张至 100 台将降低集中度;规划中的汽车和半导体部署提供多元化路径截至 2026 年 5 月,全部商业部署证明都压在一个客户的一处设施上;客户背书质量强,但集中度极端
国企投资者兼战略客户投资者(BYD、SAIC、BAIC)BYD、SAIC Motor(国有)、BAIC Group(国有)战略投资人,同时可能成为优先客户和分销渠道高 —— 三家中国头部汽车 OEM;制度性偏好国内龙头供应商政策重心离开人形机器人;投资人兼客户关系让治理变复杂;独立交易定价不能代表更广泛市场多个国企投资者降低单一投资人依赖;汽车客户潜力拓宽部署垂直场景投资人兼客户条款可能无法复制到独立企业客户;投资人与客户角色之间存在潜在治理冲突
CATL 电池供应(投资人且可能为供应商)Contemporary Amperex Technology Co.(CATL 公司)为机器人电源系统供应电池;战略投资人与供应链利益一致中 —— CATL 在中国电池供应中占主导;投资人兼供应商身份带来一致激励CATL 供应中断;定价争议;CATL 电池技术在人形机器人电源需求上失去竞争力CATL 在中国电池供应中的主导地位和战略投资人身份,带来供应连续性若 CATL 定价缺乏竞争力,投资人兼供应商关系可能阻碍竞争性电池采购
LG Electronics(韩国战略投资人,潜在分销合作伙伴)LG Electronics Inc.首个韩国战略投资人;可能打开韩国工业和消费电子市场;Mirae Asset 共同投资中低 —— 单一战略投资人;未披露商业分销协议LG 出售持股或降低人形机器人优先级;LG 投资引发 CFIUS 对美国业务的审查LG 投资释放国际企业认可信号;Mirae Asset 共同投资强化韩国金融生态入口该投资未伴随披露商业合作或分销协议;截至 2026 年 5 月,战略价值尚未确认

各行按严重性排序。习近平政治背书被列为极高,因为它同时推动国内市场准入并限制西方市场准入,且没有可用缓释措施。Longcheer 客户集中度是近期最可处理的风险——化解它需要 2026 年 Q3 成功扩张,并出现更多具名客户。

[CR024, CR025, CR026, CR027, CR028, CR020]
FR003: 依赖关系图 — Agibot 关键伙伴、投资者与监管方
[CR020, CR024, CR026, CR027, CR029]

7.5 财务与资本风险

Agibot 最近披露的估值超过 70 亿元人民币(按当前汇率约 $970M)。Goldman Sachs Research 估计,全球人形机器人市场在基准情景下 10–15 年内可能达到 $6B,蓝天情景到 2035 年可达 $154B。IDC 预测,到 2030 年全球人形机器人出货量将超过 510,000 台。这些预测支撑了一个终局很大的市场,但也凸显当前状态——Agibot 已生产约 10,000 台、平均售价未经确认——与支撑近十亿美元估值所需收入规模之间的距离。 Agibot 尚未披露收入、毛利率、经营亏损或烧钱速度。Humanoid Index 统计的多轮已披露融资总额约为 $83M,但 Tencent、BYD、JD.com 和 LG Electronics 的战略融资可能未按完整金额披露。人形机器人制造资本开支很重——模具、零部件采购、质量保证、现场服务、质保准备金——这意味着即便已生产 10,000 台,公司几乎肯定仍处在深度负自由现金流。Goldman Sachs 分析指出,2026 年是“销量验证与预期重置”的关键一年,投资人关注雄心勃勃的出货目标和供应链公司估值能否持续。 据报道,Agibot 正准备在中国 STAR Market A 股上市,监管机构为中国证监会(CSRC)。上市会带来信息披露义务、锁定期动态、市场情绪敏感性和公开市场估值纪律,这些都是私募投资人此前无需面对的约束。若 IPO 认购热度高于可交付的收入证据,上市后 Agibot 可能必须完成激进的商业部署里程碑,才能避免估值压缩。这会诱发过度承诺部署时间表和客户收入确认的激励,构成财务和声誉风险。反过来,若 IPO 市场恶化(因为中国科技市场整体环境或中美紧张升级),Agibot 可能需要推迟公开流动性事件,并在制造最耗资本的阶段继续依赖私募资本。 [CR030, CR031, CR032, CR033, CR034, CR035]

财务与资本风险登记表
风险发生概率严重性量化缓释措施剩余敞口尽调路径
7 billion yuan 估值与早期商业收入之间的缺口确定(估值已披露;收入未披露)Goldman Sachs 基准情景预测 10–15 年后全球市场为 USD 6B;Agibot 估值 USD 970M,意味着市场成熟时全球份额需 >15%大市场上行情景和中国先发优势部分支撑估值;但没有收入基线可验证只有拿下相当大的全球市场份额,估值才站得住;现阶段商业收入证据太薄,无法验证要求披露 2025 年和 2026 年 Q1 收入;获取单台 ASP 数据;建模收入轨迹能否支撑估值
资本密集型制造爬坡期烧钱速度和现金跑道未知确定(烧钱速度未对外披露)烧钱速度未披露;规模化制造通常需要大量资本开支和库存融资已披露融资 USD 83M+(由于战略轮未披露,可能低估总额);凭借投资人网络,后续战略融资仍可获得没有财务披露就无法评估偿付能力或现金跑道;产量每翻一倍,资本密集度风险都会放大要求提供审计财务报表、经营现金流、未来 18 个月现金预测,以及当前产量下的单位经济性
科创板 IPO 时点与市场情绪风险CSRC 审核周期通常为 6–18 个月;美中紧张可能推迟 IPO 或重定价若 IPO 延迟,多个战略投资人可提供私募资本桥接;国内资本市场总体偏好国家级技术龙头IPO 延迟会迫使公司继续依赖私人资本;过度依赖 IPO 作为流动性事件,会在上市前制造短期业绩压力跟踪 CSRC IPO 排队;评估现有投资人锁定期和共同出售权条款;确认 IPO 是否取决于收入里程碑
年产能扩至 50,000–100,000 台的资本需求100,000+ 台硬件制造需要工厂投资、工装和营运资本,显著超过已披露融资国企投资人基础和国内政策支持,提供开发性融资和工业用地补贴入口除当前已披露轮次外,制造爬坡融资尚未确认;资本密集度将要求进一步股权或债务融资要求提供制造资本计划;确认任何已承诺信贷额度、开发银行融资或工业园区补贴

各行按严重性排序。所有财务风险本质上都被收入、利润率和烧钱数据缺失放大。没有私人财务披露,估值 / 收入缺口仍无法解决。IPO 准备可能加速披露义务。

[CR030, CR031, CR032, CR033, CR034]

7.6 人员、执行与战略风险

Agibot 由创始人 Edward Deng(CEO、董事长)和 Peng Zhihui(联合创始人、总裁、CTO)领导。Peng Zhihui 曾是 Huawei “天才少年”工程师,并主导 Kirin 芯片项目,是公司最公开可见的技术人物,常出现在媒体和投资人沟通中。他的前 Huawei 人才身份会强化西方监管框架中的国家安全担忧。Deng 和 Peng 同时支撑公司的技术可信度、投资人关系和政府地位。目前没有公开披露的继任计划、副手管理层结构或 CTO 备份角色。如此集中的关键人风险符合早期中国科技公司的特征,本身不构成否决项,但需要专项尽调。 人形机器人竞争格局包括 Unitree($580M IPO 申报,声称 2025 年出货 5,500 台)、Figure AI(已融资 $1.9B)、Agility Robotics(Amazon 支持,Digit 已部署于仓库)、Boston Dynamics(Hyundai 支持)和 Physical Intelligence。Agibot 作为中国量产领先者的战略位置,如果西方采购限制、数据主权担忧和本土竞争对手部署进展同时加速,未必能转化为国际市场份额。IFR 2026 年 5 月报告指出,中国“十五五”规划(2026–2030)把机器人置于国家工业现代化核心位置,这将加剧国内竞争,并推动竞争对手拿到政府补贴的 R&D。 Clingendael Institute 描述的 China Standards 2035 倡议,以及 MIIT HEIS 委员会框架所体现的方向,目标是把中国技术规格嵌入全球供应链。Agibot 积极参与 HEIS 委员会(Peng Zhihui 任副主任),对国内市场定位来说是理性选择,但也会强化西方市场的一种观感:中国人形机器人标准是为扶持国内冠军而设计,进而加速全球监管分化。若美国、欧盟和盟友国家采纳与中国标准根本不兼容的人形机器人安全和数据治理标准,可服务市场会被切碎,Agibot 也必须为不同监管制度开发不同产品配置。 [CR036, CR037, CR038, CR039, CR040]

人才 / 执行风险登记表
角色或职能依赖或缺口发生概率严重性缓释措施尽调路径
Peng Zhihui(联合创始人、总裁、CTO)核心技术架构师、技术公众代表、HEIS 副主任;前 Huawei「天才少年」工程师;在投资人和政府面前承载技术可信度极高联合创始人身份带来长期利益一致;Huawei 技术履历难以替代;公开机构角色提供部分连续性要求提供继任计划和技术领导梯队;确认 CTO 备份人选和 VP Engineering 架构;评估关键人物保险;注意 Huawei 背景在西方市场的观感风险
Edward Deng(创始人、董事长、CEO)公司愿景、投资人关系、政府沟通、融资、IPO 准备极高创始人兼 CEO 与长期使命绑定,降低商业化早期的离任风险要求提供董事会构成、治理结构和授权框架;确认 CEO 备份人选及 COO 层级运营管理
规模化制造与质量工程执行从 10,000 台扩到 100,000+ 台,需要公开披露尚未证明的工业制造能力投资人兼客户合作关系(BYD、SAIC)提供制造生态入口;未披露专职 VP Manufacturing 或 Chief Operations Officer确认制造负责人、工厂计划,以及超出现有产量后的产能爬坡合作结构
国际市场进入执行(销售、集成、现场服务)未披露中国以外的国际销售负责人、区域团队或认证集成商网络CES 2026 市场首秀和 LG 战略投资释放国际化意图;执行基础设施尚未确认要求提供国际组织架构、员工数目标和集成合作伙伴计划文件
AI 模型与研究人才留存Unitree、Figure AI、Physical Intelligence、Google DeepMind 以及中国大型科技公司都在争夺具身 AI 人才Agibot 的规模和政府背书支撑有竞争力的薪酬与使命认同;未披露员工数或流失数据要求提供员工数变化、技术团队结构,以及与国内外可比竞争对手的薪酬基准

各行按严重性排序。Peng Zhihui 集技术领导、政府影响力和面向投资人的可信度于一身,创始人主导的技术依赖集中,是最重大的人员风险。国际执行缺口是近期最可处理的缺口;国际收入要对业务产生实质贡献,必须先补上它。

[CR036, CR037, CR038, CR022, CR023]
缓释与否决标准表
风险可监控触发项阈值或事件行动含义
ASRA 或类似联邦采购禁令生效美国国会行动;Federal Register 公告;DoD/GSA 采购指引ASRA 在两院通过或经签署成法;California、Texas 或 New York 通过同等州级限制美国联邦和政府邻近客户收入的投资逻辑被打破;加快欧盟和盟国市场策略
Longcheer 部署到 2026 年 Q4 仍未扩至 100 台Agibot 新闻稿;Longcheer 投资人沟通;第三方部署报道到 2026 年 12 月仍未宣布 100 台扩张里程碑投资逻辑承压:主要商业参考案例未能规模化;要求更新管线并提供替代参考案例
出货量主张被独立审计下修或否定IDC、Omdia 或分析师重述 Agibot 出货数据;科创板 IPO 招股书披露公开重述将 2025 年出货量下调至 3,000 台以下;CSRC 招股书显示交付量显著更低投资逻辑被打破:商业证据基础弱于公司主张;估值支撑恶化
已部署机器人发生安全事件或人身安全事故媒体报道;OSHA 300 日志(美国);客户公开声明;MIIT 事故披露任何公开报道的 Agibot 机器人商业部署人身伤害事故投资逻辑承压:产品责任显性化;保险重新定价;客户暂停采用;监管调查
关键创始人离任(Peng Zhihui 或 Edward Deng)LinkedIn;公司新闻稿;投资人沟通;若 IPO 正在推进,则查看科创板文件任一联合创始人在 IPO 完成和商业规模部署前离任投资逻辑被打破:集中的技术和关系资本流失;要求评估接任领导层质量
未宣布 2027 年 1 月之后的延长安全支持Agibot 官方支持政策更新;客户沟通;产品路线图公告到 2026 年 10 月仍未宣布延长支持期或 LTS 计划投资逻辑承压:企业客户采用摩擦上升;合同续签风险增加;确认企业支持路线图

否决标准围绕六类最高后果风险设定:立法、客户、商业证据、安全、人员和产品支持。六项中有三项(ASRA、Longcheer 扩张、延长安全支持)在 2026 年内有可衡量的限时触发点。

[CR001, CR025, CR015, CR010, CR036, CR042]
Chapter 08

08估值

8.1 投资建议与投资逻辑概要

Agibot 获得 TRACK(观察)投资建议,综合评分 6.8/10,置信度中等,风险评级高。这个建议反映出一种真实且有证据支撑的张力:公司在生产规模上取得了异乎寻常的成果,但财务、地缘政治和商业层面仍有一组未解决的不确定性,阻止我们给出更高确信度的立场。 投资逻辑有五根相互支撑的支柱。第一,截至 2026 年 Q1,Agibot 是全球出货规模最高的人形机器人制造商:公司 2026 年 3 月生产出第 10,000 台人形机器人,2025 年出货约 5,100 台,并拿到 Omdia 全球人形机器人出货量第一。西方或亚洲竞争对手都没有达到这一规模。第二,G2 在江西南昌 Longcheer Technology 平板工厂的工业部署——G2 机器人连续 8 小时完成 2,283 项任务、零错误——是截至 2026 年 5 月该领域最有说服力的工业规模人形机器人证据。第三,Agibot 的投资人基础包括 BYD、Tencent、JD.com、LG Electronics、Mirae Asset、Gaorong Capital(High-Flyer/Hillhouse)、Sequoia China 和 TCL Ventures;这组联盟提供了西方竞争对手拿不到的深度供应链、分销和资本市场优势。第四,Goldman Sachs《Humanoid Robot: The AI Accelerant》报告和 IDC 2026 年人形机器人商业化分析都把制造规模和工业部署证据列为私募机器人估值的两大核心价值驱动因素;Agibot 在两项上都领先。第五,投前估值约 $960M(CNY 7B+,2025 年 8 月)较 Agility Robotics C 轮隐含估值折价约 45%,较 Figure AI 最近一轮折价约 70%,为能承受中国敞口的机构投资人提供了相对价值切入点。 反向逻辑同样重要。收入、毛利率和烧钱速度完全未披露:Agibot 是私营公司,尚未发布任何财务披露。主导收入流不是高毛利的工业经常性合同,而是娱乐和服务人形机器人的硬件销售(A2 Lite 售价 $44,560,X2 售价 $24,240);转向工业 RaaS 是投资逻辑关键,但商业上尚未在规模化场景验证。Swancor Advanced Materials 收购(CNY 2.1B,约 $290M)引入了一项规模大、复杂且非核心的资本投入,其战略理由——向高性能材料做垂直整合——合理但未经验证,整合也会占用大量管理层带宽。拟议中的 American Security Robotics Act(2026 年 3 月)将禁止美国政府机构采购中国来源机器人,并预示北美和欧洲市场更广泛的企业采购犹豫。最后,Unitree G1 约 $16,000 USD 的价格在开发者和研究细分市场形成结构性价格压力,而 Figure AI 的 Helix 通用 AI 模型和全球汽车客户进展带来能力叙事风险。 若满足以下条件,建议将从 TRACK 上调至 INVEST:(1) 已确认工业 ARR 超过 $50M(或披露单位经济模型后达到等量规模);(2) Longcheer 之外第二个具名工业制造客户实现规模化(>20 台部署);(3) 明确 Swancor 整合里程碑和资本配置;(4) US Security Robotics Act 没有出现影响企业细分市场的不利进展;(5) 已确认融资轮披露投后估值,验证或重定价 CNY 7B+ 投前锚点。投资建议逻辑图(FV001)展示从证据支柱和风险向量到投资结果的因果链。[CV001, CV002, CV003, CV004, CV005, CV006]

建议摘要表
参数取值证据依据 / 备注
总体建议观察具备量产规模证明;工业 ARR 未披露;地缘政治逆风显著
综合评分6.8 / 10基于市场机会、量产证明、技术护城河、投资人阵容、财务透明度、地缘政治风险和估值支撑加权
置信度关键未知项:收入、毛利率、工业 ARR、Swancor 集成、US Security Robotics Act 结果
风险评级财务不透明、地缘政治导致的市场准入受限风险、Swancor 资本承诺、工业部署仍只有单一客户
估值立场合理至偏高(中国市场天花板;相对西方同业有折价)融资前估值约 $960M,低于 Figure AI($3.2B)和 Agility Robotics($1.75B);但未披露财务数据时, 该估值只能说明产量,不是收入倍数
建议行动观察;确认工业年经常性收入(ARR)+ 第二个制造业客户后加仓投资条件:披露工业年经常性收入(ARR) >$50M、第二个具名规模化客户、Swancor 整合清晰、US Act 未通过(见 TV006)
目标持有期4–6 年(2029–2032)IPO 或 M&A 退出窗口,与中国科创板就绪时间或战略买方时间线匹配
主要退出路径中国 A 股 IPO(主要);工业战略并购(次要);借 LG 桥接韩国 / 日本买方CATL、BYD、SAIC 或 State Grid 最可能成为战略买方;LG Electronics 现有持股带来韩国退出可选性

投资建议和综合评分是分析师基于八章证据作出的判断。没有 Agibot 财务披露支撑该评分;所有估值估计来自可比公司分析和情景建模。 置信度和风险评级是相对 Series B / 战略轮阶段可比硬件 AI 公司的评估。6.8/10 综合评分对应观察阈值的高端。

[CV001, CV002, CV013, CV032]
投资逻辑 / 反向逻辑表
投资逻辑证据基础反向逻辑什么会改变判断
截至 2026 年 Q1,Agibot 是全球产量最高的人形机器人制造商第 10,000 台于 2026 年 3 月下线;Omdia 将 Agibot 列为 2025 年人形机器人出货量第 1,约 5,100 台;最后 3 个月产能提速 4 倍产量 ≠ 商业收入;已部署与演示 / 展示用机型的出货结构未披露;竞争对手 Unitree 声称 2025 年出货 5,500 台(存在争议)在下一轮融资材料中披露工业 vs 娱乐出货结构,以及已确认部署机数量
G2 在 Longcheer 的零错误工业部署,是该赛道最有说服力的工厂验证8 小时连续运行完成 2,283 项任务、零错误;吞吐量 310 件 / 小时;节拍 18–20s;CnTechPost 与 Interesting Engineering 均有佐证单一客户、单一工厂,合同经济性未确认;100 台规模目标(2026 年 Q3)未经验证宣布第二个工业制造客户,并确认部署台数和部署经济性
战略投资者联盟带来西方同业拿不到的供应链、资本市场和分销优势BYD(汽车供应链)、Tencent(软件 / AI)、JD.com(物流分销)、LG Electronics + Mirae Asset(韩国资本和工业入口)、 CATL 支持的电池供应、Sequoia China投资者集中暴露于中国市场;西方机构 LP 基础有限;CFIUS 式审查限制西方 M&A 可选性LG Electronics / Mirae Asset 持股和国际分销协议,在美国路径受阻时提供通向韩国 / 欧洲战略退出的桥
Goldman Sachs 和 IDC 指出,Agibot 的两个关键价值驱动因素——制造规模和工业部署验证——也是人形机器人估值的核心指标Goldman Sachs《Humanoid Robot: The AI Accelerant》将数据飞轮、制造规模、部署验证列为核心价值驱动因素; IDC 2026 商业化报告确认拐点动态;Agibot 在两个产能规模维度都领先Goldman Sachs 和 IDC 的评估并非专为 Agibot 撰写;西方商业部署成熟后,分析师框架可能调整在下一版中确认 Goldman Sachs 和 IDC 分析框架仍有效;监控西方同业是否缩小产量差距
融资前估值约 $960M,相比西方人形机器人同业存在显著相对价值折价Figure AI 约 $3.2B;Agility Robotics $1.75B;Apptronik 约 $1.4B+;1X Series C 隐含 $3–5B;Agibot 约 $960M, 尽管产量领先,仍比西方同业低 45–70%折价反映中国市场溢价天花板、地缘政治风险和财务不透明;相对便宜并不证明内在价值在披露融资轮中确认投后估值;披露任何工业年经常性收入(ARR)指标,为内在价值分析提供锚点

所有投资逻辑和反向逻辑均由分析师基于第 1–8 章的公开证据构建。尚未取得 Agibot 管理层对反向逻辑项的回应。 可能改变判断的条件具体、可验证,并对应 TV006 中描述的尽调路径。

[CV003, CV004, CV005, CV008, CV009, CV015]
FV001: 推荐逻辑 — 从证据到投资结论

这条因果链从 Agibot 四个关键证据支柱(产量领先、工业部署验证、战略投资者联盟、相对估值折价)出发,穿过三条风险线(财务不透明、地缘政治风险、Swancor 资本承诺),落到 6.8/10 的观察建议:中等信心、高风险;同时给出上调至有条件买入的触发路径。

[CV001, CV003, CV006, CV013, CV032, CV040]

8.2 融资背景、估值框架与可比公司组

Agibot 自 2023 年 2 月成立以来已完成十余轮融资。按 Tracxn 和 Humanoid Index 口径,已披露风险资本约 $83.8M,但这个数字几乎肯定低估了总资本:中国私营公司披露常常不完整,QCC 企业登记显示 50 多名股东,横跨多个未披露规模的融资轮次。按轮次确认的关键投资人包括:Gaorong Capital 和 Bailai Ventures(天使轮,2023 年)、BYD(A+,2023 年 8 月)、Sequoia China 等(pre-B/A,2024 年)、Tencent Ventures(B 轮,2025 年 3 月)、JD.com / JoyGen(B+,2025 年 5 月)、LG Electronics / Mirae Asset New Growth Fund(战略轮,2025 年 8 月)。LG Electronics / Mirae Asset 轮被称为“战略”融资,也是该基金首笔全球投资;融资规模未披露。Pandaily 报道该轮投前估值超过 CNY 7B(按 2025 年 7 月汇率约 $960M),是目前公开可得的最佳企业价值锚点。若计入未披露轮次规模,隐含投后估值可能在 $1.0B–$1.3B 区间。 Swancor Advanced Materials 收购是 VC 融资之外最重大的资本事件。2025 年 7 月 8 日,由 Agibot 及其创始团队牵头的财团同意支付最高 CNY 2.1B(约 $290M),取得这家 A 股上市材料公司(深交所)的控制权。这笔收购显示公司有意向上游碳纤维和高性能复合材料做垂直整合;这个多元化方向合理,但资本开支很重,也让总有效资本投放远超已披露的 $83.8M VC 数字。投资人应建模合并主体,而不只是 VC 资助的机器人业务。 Agibot 当前阶段适合用场景加权的里程碑模型估值,锚点应是产量和部署证据,而不是收入倍数(收入未披露)。在基准情景下,假设到 2027 年底累计部署 20,000 台,其中 30% 纳入付费工业合同,单台每年 CNY 300,000–500,000(由 Longcheer 经济模型推导),工业年收入可能达到 $250M–$450M;按当前隐含估值,对应前瞻 EV/Revenue 2–5x,符合早期工业软件和自动化公司水平。乐观情景中,50%+ 产量转向高毛利工业 RaaS,并出现更多制造业客户证据,可支撑 2027 年前瞻倍数 6–10x 下的 $2B–$3B 企业价值。悲观情景下,硬件销售仍占主导、利润率很薄、地缘政治壁垒阻止西方规模化,意味着估值可能下修至低于 CNY 5B($700M)的降轮风险。 可比公司组分为三类。人形机器人私募轮次中,Figure AI 是最相关的全球同行:2024 年 2 月融资 $675M,投后约 $2.6B,后续轮次隐含估值约 $3.2B;Figure 的 BMW 工厂试点和 Helix 通用 AI 模型,是最强的西方竞争对标。Agility Robotics 2025 年 3 月完成 $400M C 轮,投后约 $1.75B,有 GXO 生产部署支撑。Apptronik 从 Samsung 等投资人处融资 $350M,估值未披露,但估计为 $1.4B–$1.6B。按分析师估计,1X Technologies 未确认的 C 轮隐含 $3B–$5B 区间。公开自动化同行中,Teradyne(协作机器人部门,约 $4B EV)和 Zebra Technologies(约 $15B EV)按 5–8x 前瞻收入交易,为 Agibot 达到 $300M+ ARR 后提供退出倍数参照。Boston Dynamics 先例(Hyundai 2021 年从 SoftBank 收购,约 $1.1B)为一家受地缘政治交易审查的中国人形机器人公司设置了战略并购底价。[CV013, CV014, CV015, CV016, CV017, CV018]

可比估值表
公司 / 参照阶段 / 轮次估值 / EV产量或部署指标与 Agibot 的相关性局限性
Figure AI(美国)Series D(2025)投后约 $3.2BBMW 工厂试点;Helix AI 模型;累计融资约 $675M;通用人形机器人主要西方同业;同阶段(收入前规模化);AI 模型路径可比美国注册,无地缘政治天花板;融资规模大得多,削弱可比性
Agility Robotics(美国)Series C(2025 年 3 月)投后约 $1.75BGXO 生产部署;RoboFab 年产能 10,000 台;$400M Series C最直接的商业同业;物流场景对比 Agibot 的制造业重心单一具名客户(GXO);商业范围仅美国;物流细分赛道对比 Agibot 跨垂直模型
Apptronik(美国)Series A+(2024,Samsung 领投)约 $1.4B–$1.6B(估计)累计融资 $350M;Apollo 人形机器人;NASA 和汽车试点Samsung 战略投资者与 LG 有重叠(Mirae-LG 基金结构相似);制造业投资逻辑估值未官方披露;仍处早期试点阶段,没有量产规模验证
1X Technologies(挪威)Series B+(2025)$3B–$5B(隐含,未确认的 Series C)约 390 台 EVE 商用部署;NEO 消费级人形机器人开发中OpenAI 支持;企业安防部署验证;EQT Ventures 领投Series C 未确认;NEO 偏消费级的逻辑与 Agibot 的工业 / 制造重心不同
Boston Dynamics(2021 年被 Hyundai 收购)M&A 先例$1.1B 收购Spot 机器人已部署在工业现场;Atlas 仍是开发阶段双足机器人收入前先进机器人公司的战略并购底价参照2021 年交易;行业已大幅演进;Hyundai 收购结构无法类比受 CFIUS 约束的中国 M&A
Teradyne / Universal Robots(上市同业)上市可比公司EV 约 $4B(Teradyne);UR 部门贡献收入约 $1.5BUR 峰值年销量 30,000+ 台协作机器人;远期收入倍数 5–8x退出倍数参照:如果 Agibot 达到 $300M+ ARR,按 5–10x 可实现 $1.5B–$3B EV协作机器人,不是人形机器人;ASP 低得多;市场动态不同

可比集合反映截至 2026 年 5 月公开已知的私募轮估值和上市公司企业价值。Figure AI、Apptronik 和 1X 的估值部分或全部为估计; 任何人形机器人私营可比公司都没有已确认的股权结构表或优先股堆叠数据。所有比较分析在融资轮披露确认后都需修订。

[CV015, CV016, CV017, CV018, CV019, CV020]
FV002: 估值敏感性 — 企业价值 vs 工业 ARR 情景

条形图展示八个参考情景下 Agibot 隐含企业价值:从悲观的下轮降估底价 $700M,到当前入场隐含估值;再到 Boston Dynamics 和 Agility Robotics M&A 先例给出的可比底线、2028 基准情景 $2.0B,以及 2028 乐观情景 $4.0B。图中凸显估值分散度很大,最终结果首先由工业 ARR 轨迹决定。

[CV013, CV021, CV022, CV023]

8.3 乐观 / 基准 / 悲观情景与回报分析

三个情景决定 Agibot 在 2026–2030 投资期的估值路径。情景概率为近似值,基于截至 2026 年 5 月公开证据权重:乐观约 25%,基准 50%,悲观 25%。 乐观情景(概率:约 25%)假设 Agibot 在 2027–2028 年成功把生产规模转化为规模化工业 ARR。关键假设:(a) G2 工业项目从 Longcheer 证明点扩展到 10+ 家具名制造客户,每家部署 50–200 台;(b) Swancor 收购带来垂直整合的执行器和复合材料成本优势,使单台 COGS 下降 20–30%;(c) 中国国家电网 / State Grid Corporation 采购项目(QCC 备案数据提到的 CNY 68B 具身智能设备招标)授予 Agibot 重要合同;(d) 美国立法限制西方销售,但非限制性市场(ASEAN、India、Middle East、EU)贡献 20%+ 收入。在这些假设下,到 2028 年累计工业 ARR 达到 $400M–$600M,以 6–10x 前瞻收入支撑 $3B–$5B 企业价值。按隐含约 $1.0B–$1.3B 投后入场价,2025/2026 年进入的投资人在乐观情景下 5 年回报约为 2.5–5x。 基准情景(概率:约 50%)假设 Agibot 稳步扩大工业部署,但到 2027 年仍主要是硬件销售公司。关键假设:(a) 到 2027 年底,G2 扩展到 3–5 家工业客户,工业部署总量达到 500–800 台;(b) A2 和 A3 商业 / 娱乐硬件销售维持当前定价,产生 $80M–$120M 年硬件收入;(c) Swancor 整合在 2027 年完成,且无重大运营扰动;(d) 美国或欧盟立法没有阻断亚洲和非结盟市场准入。在这些假设下,2028 年隐含企业价值为 $1.5B–$2.5B。以约 $1.0B–$1.3B 入场价计算,基准情景 5 年回报约为 1.2–2.0x——对能承受中国敞口的机器人基础设施投资人来说不算高,但为正。 悲观情景(概率:约 25%)包括:(a) US Security Robotics Act 通过,并对全球企业买家造成寒蝉效应,把 Agibot 限定在中国市场收入;(b) 工业部署在规模化后暴露质量问题(执行器失效、非结构化环境中的自主能力缺口);(c) Swancor 收购造成资产负债表压力和管理层分心;(d) Unitree 以更低价格实现可比工业部署证明,压缩 Agibot 定价权。在悲观情景下,到 2027 年 Agibot 被迫进行低于 CNY 5B($700M)的降轮,或以受压估值寻找中国国有战略买家。按当前入场价,悲观情景回报为 0.5–0.7x,优先股堆叠可保护后轮机构投资人,普通股则会遭受重创。 以约 $1.1B 投后估值入场,概率加权预期价值到 2028 年约为 $1.5B–$2.0B,对应未来轮次稀释前 1.4–1.8x 预期回报。相对于该行业资本密集度和二元风险画像,这个结果边际上有吸引力,但若没有显著的乐观情景支撑,达不到 3x 回报目标。[CV024, CV025, CV026, CV027, CV028, CV029]

乐观 / 基准 / 悲观情景表
情景关键假设隐含 2028 EV以约 $1.1B 入场计算的回报下行触发因素 / 风险概率信号
乐观G2 扩展至 10+ 个工业客户,每个 50–200 台;Swancor 带来 20–30% COGS 降幅;State Grid CNY 68B 招标获得有意义合同; ASEAN/EU 收入占总收入 20%+$3B–$5B5 年约 2.5–5xState Grid / 第二个工业客户未落地;Swancor 整合延误~25%
基准G2 扩展至 3–5 个工业客户,总部署 500–800 台;A2/A3 硬件销售 $80M–$120M/年;Swancor 到 2027 年完成整合; 美国 / 欧盟法律不阻断亚洲市场$1.5B–$2.5B5 年约 1.2–2.0x工业客户流失;Unitree 竞争压缩硬件销售利润率~50%
悲观US Security Robotics Act 通过;美国阵营市场的企业买家回避 Agibot;规模化后出现工业质量问题;Swancor 资本压力; 估值跌破 CNY 5B 的降估值融资$0.5B–$0.9B约 0.5–0.7x(优先股堆叠部分保护后轮投资者)US Act 签署成法;到 2026 年 Q3 仍未出现第二个工业客户~25%

所有情景假设均由分析师推导。Agibot 尚未公开披露财务预测。情景概率为指示性,反映截至 2026 年 5 月公开证据的权重。 回报估计未计入未来融资轮带来的稀释;实际投资者回报取决于优先股堆叠结构,该结构尚未公开披露。

[CV024, CV025, CV026, CV027]
FV003: 估值与回报区间 — 乐观 / 基准 / 悲观情景

区间图展示 Agibot 以约 $1.1B 投后估值入场时,悲观、基准、乐观三种退出估值结果。每条柱覆盖该情景的区间,并给出中位估计。回报轮廓不对称:基准情景只给出 1.2–2x 的温和回报,乐观情景可达 2.5–5x 但高度依赖执行,悲观情景则意味着 0.5–0.7x 的亏损。

[CV024, CV025, CV026, CV027, CV028]

8.4 反向情景——地缘政治风险、估值过高与竞争压力

Agibot 的反向投资情景锚定三类相互强化的风险:地缘政治市场准入限制、估值不透明与估值过高风险,以及不断加强的西方竞争压力。 截至 2026 年 Q2,地缘政治风险是最重要的反向因素。美国议员于 2026 年 3 月 26 日提出 American Security Robotics Act,拟禁止美国政府机构采购中国来源的人形和移动机器人。虽然截至报告日期法案尚未通过,但提出本身反映出围绕中国机器人风险的两党政治共识,并且很可能从政府采购细分市场外溢。与美国政府签有合同的企业买家(国防、物流、医疗)已经面对合规审查,即便法案正式通过前,也会从类别上排除 Agibot 硬件。Huawei 在电信、DJI 在企业无人机领域都经历过同样路径。Hill Dickinson 对人形机器人责任框架的法律分析将法律问责、数据主权和安全认证标准列为 2026 年推进最快的三大监管维度;这些都会为中国来源机器人进入西方企业部署制造额外结构性壁垒。 中国国内监管环境提供部分抵消:Robotics & Automation News 报道,2026 年 3 月发布的新型人形机器人标准由 Agibot 的生产规模牵引,使其在国内采购中获得竞争优势。State Grid Corporation 的 CNY 68B 具身智能采购招标中,Agibot 是参与方,代表一条重要的潜在政府合同管线。不过,对中国政府采购的依赖也带来另一种风险:政策优先级变化、付款节奏以及单一客户集中。 相对于西方同行,约 $960M 的投前入场点降低了估值过高风险——显著低于 Figure AI 的 $3.2B 或 1X Technologies 隐含 $3–5B C 轮区间。但公司没有披露任何财务指标(收入、ARR、毛利率、烧钱速度、现金),因此无法验证约 $960M 估值是否由单位经济模型支撑。10,000 台生产里程碑是产量信号,不是盈利能力信号。在工业 ARR 公开披露之前——哪怕只是方向性披露——所有估值锚点都依赖产量可比和战略投资人信号,而不是基本面指标。 西方同行的竞争压力是中期但持续上升的风险。Figure AI 面向家庭和通用部署的 Helix 模型,可能在 Agibot 进入西方市场之前抢占企业 AI-机器人心智。Agility Robotics 的 GXO 物流部署提供了一个美国本土“安全港”可比对象,企业物流买家无需承担地缘政治合规风险即可采购。Unitree G1 约 $16,000 的价格点对开发者和研究细分市场形成结构性下行压力,可能在 Agibot 开发者生态(AIMA)成熟之前锁定平台标准。[CV032, CV033, CV034, CV035, CV036, CV037]

投资逻辑破裂与终止触发因素表
触发因素阈值 / 事件对投资逻辑的传导行动含义
US Security Robotics Act(或同等法案)签署成法美国任何禁止采购中国机器人的法规最终通过并由总统签署北美企业市场被切断;意味着欧盟及盟友采购限制可能跟进;退出路径被压缩到仅中国或非结盟市场有全球投资权限的投资者立即触发卖出;中国境内投资者维持观察;重新评估西方战略买方可选性
到 2026 年 Q4 未宣布第二个工业制造客户到 2026 年 12 月,部署 >20 台的新工业产线客户为零Longcheer 仍只是单一数据点;工业 RaaS 逻辑无法验证;基准情景收入轨迹受损若第二个客户未出现,则下调至放弃;监控 Longcheer 扩展到 100 台,作为部分替代信号
Swancor 收购整合失败或需要追加资本CSRC 调查、Swancor 盈利低于预期 >30%,或 Agibot 被要求在初始承诺之外再注入超过 CNY 500M 资本资产负债表承压;管理层带宽被非核心整合吃掉;R&D 和部署扩张能力下降要求披露整合里程碑;在估值模型中将 CNY 2.1B 承诺视为或有负债
工业部署规模化后质量失败公开报道的安全事故、执行器故障率 >5%,或客户退出 G2 计划制造业客户信任被击穿;触发监管检查;竞争对手借事故在客户沟通中替代 Agibot立即触发放弃;只有披露根因并完成独立安全审计后才恢复观察
估值低于 CNY 5B($700M)的降估值融资确认融资投后估值低于 CNY 5B,或结构化债务融资暗示股权价格底部表明投资者群体对工业化转型失去信心;优先股堆叠给早期轮次带来稀释悬崖启动退出评估;监控是否出现中国国资战略买方以困境价格接盘

投资逻辑破裂触发因素按可观察事件定义,而非分析师观点。所有触发因素都可通过公开媒体报道、CSRC 文件或 Agibot 官方公告监控。 阈值设定用于捕捉足以实质改变基准投资逻辑的最低信号强度,而非增量负面新闻。

[CV032, CV033, CV034, CV035, CV036]

8.5 退出准备度与最终尽调问题

截至 2026 年 5 月,Agibot 的退出准备度有限。公司尚未公开披露 IPO 时间表、STAR Market 注册申报或海外上市意向。对早轮投资人而言,最可能的近期流动性路径,是由工业部署证据和第二或第三个具名制造客户驱动、以更高投后估值完成 C 轮或 C+ 轮。Swancor Advanced Materials 收购给公司结构加入了上市实体组件,理论上可作为深交所反向并购 IPO 路径,但这种结构并不常见,可行性也未确认。 对 2025/2026 年入场投资人而言,最可能的完全退出路径包括:(1) 中国 A 股 IPO(STAR Market 或科创板、ChiNext),大概率不早于 2028 年,取决于 CSRC 盈利 / 收入披露要求;(2) 作为国家具身智能基础设施建设的一部分,被中国国有或准国有工业公司(CATL、BYD、SAIC、State Grid)战略收购;或 (3) 在非美国司法辖区(Japan、South Korea、Germany、Singapore)被西方战略买家收购,这需要规避等同 CFIUS 的审批,并会受益于 LG Electronics / Mirae Asset 作为桥梁投资人的关系。 狭义 IPO 准备度取决于 CSRC 面向尚未盈利高科技制造公司的 IPO 政策演变。中国《国家具身智能发展规划》已明确打开这一路径,但披露要求会迫使公司公布目前对投资人屏蔽的收入和利润率指标。 从 TRACK 上调至 INVEST 前必须解决的五项最终尽调问题列在 TV006 中。最紧迫的问题(潜在投资人 90 天内需要拿到)是:(1) Longcheer 部署的工业 ARR 或等价单位经济模型;(2) Swancor 收购资本结构和整合时间表的清晰度;(3) US Security Robotics Act 立法状态确认。剩余两项——第二个工业客户签署合同,以及披露的 C 轮条款清单或投后估值——是 6–18 个月尽调问题,决定从 TRACK 上调至条件性 BUY 的触发点。[CV040, CV041, CV042]

最终尽调问题表
主题缺失证据重要性负责人 / 尽调路径
工业年经常性收入(ARR)和单机经济性Agibot 未公开披露收入、ARR 或单机合同经济性;Longcheer 合同金额、期限和续约条款未知没有任何收入锚点,整套工业 RaaS 逻辑无法验证;也无法给出收入倍数或验证 $960M+ 融资前估值直接接触 CFO 或投资者资料室;对照 Longcheer 母公司(Longcheer Technology 603341.SH)业绩会披露
Swancor Advanced Materials 整合计划和资本配置CNY 2.1B 收购的战略理由、整合时间表和预期协同指标未公开披露该收购的资本承诺比已披露 VC 融资大 2–3 倍;尚不清楚它会增强还是分散人形机器人核心逻辑查询 Swancor 的 CSRC 和深圳交易所公开文件(收购主体文件);就整合里程碑直接接触创始人
US Security Robotics Act 立法状态和企业买家反应法案已于 2026 年 3 月 26 日提出,但尚未通过;企业买家采购冻结的范围和持续时间不确定美国市场和美国阵营市场被排除,是悲观情景中最大的单一驱动因素;立法状态决定西方 TAM 可达性国会跟踪服务;对物流、医疗和制造业企业买家做背调访谈
股权结构表、优先股堆叠和反稀释条款QCC 显示 50+ 名股东,但没有优先股堆叠、清算分配瀑布或反稀释条款细节任何退出情景下普通股股东回报高度取决于优先倍数和参与权;结构未披露,无法建模回报公司律师审阅 AGIBOT Innovation (Shanghai) Technology Co., Ltd. 注册资本文件;通过 Series B 领投方直接接触法律顾问
第二个达到商业规模的工业制造客户目前只有 Longcheer Technology 确认为 G2 工业部署客户;面向汽车和半导体行业的计划扩张尚未宣布具名交易对手单一客户的工业验证是估值的必要但不充分条件;第二个具名、20+ 台客户是确认可复制性的最低要求Agibot 官方公告;AIMA 开发生态合作伙伴披露;直接接触 VP Sales

尽调问题按紧迫性和重要性排序。第 1 和第 3 项——工业年经常性收入(ARR)和 US Act 状态——是 90 天内必须解决的问题,任何新增资本承诺前都应完成。 第 2、4、5 项是 6–18 个月问题,决定从观察上调至有条件买入的触发条件。

[CV040, CV041, CV042]
FV004: 投资 KPI — 投委会可用评分

Agibot 八个投资维度的投委会可用评分仪表盘,量表为 1–10 分。综合总分 6.8/10:产量规模异常突出、市场机会强,但被财务完全不透明、地缘政治天花板和单一客户工业验证抵消。分数是分析师综合八个报告章节证据后的判断。

[CV001, CV002, CV003, CV008, CV013, CV032]

免责声明

本报告基于截至 2026-05-18 可获得的公开来源。已审阅材料未披露未上市公司财务、客户合同和股权结构条款;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 AGIBOT was established in February 2023 in Shanghai, China. SO002, SO036
CO002 AGIBOT's headquarters is in Shanghai, China. SO001, SO036
CO003 The company's full legal name is AGIBOT Innovation (Shanghai) Technology Co., Ltd., known in Chinese as 智元机器人 (Zhiyuan Jiqiren). SO002, SO037
CO004 AGIBOT's stated mission is to build world-leading general-purpose embodied robots and their application ecosystems by pioneering the fusion and innovation of AI and robotics. SO002
CO005 AGIBOT organises its product portfolio around three robot series: Yuanzheng (远征), Lingxi (灵犀), and Genie (精灵). SO002, SO001
CO006 AGIBOT is one of the world's first companies to accomplish large-scale production and commercial deployment of embodied robots across multiple countries and regions, per the company's own claims. SO002
CO007 Deng Taihua (邓泰华) is a co-founder of AGIBOT and the actual controller; Chinese corporate registry (Qichacha) identifies him as beneficial owner with approximately 33% shareholding as of July 2025. SO036, SO037
CO008 Peng Zhihui (彭志辉) is co-founder and CTO of AGIBOT; he was specifically named in this role at the March 2026 10,000-unit production milestone event. SO034, SO036
CO009 The AGIBOT founding team is described by the company as comprising "core executives from global technology leaders and top AI scientists," but individual prior employers are not detailed in public sources. SO002
CO010 Chuang Wang (王闯) serves as Senior Vice President of AGIBOT and co-presented the company's 10,000th unit milestone alongside CTO Peng Zhihui in March 2026. SO034
CO011 AGIBOT has not publicly disclosed its board composition, investor voting rights, or full equity table in any reviewed materials as of May 2026. SO036, SO037
CO012 Key-person concentration risk is elevated: both co-founders Deng Taihua (CEO) and Peng Zhihui (CTO) occupy the company's top executive and sole named technical leadership roles. SO036, SO037, SO034
CO013 AGIBOT's A2 robot was inspected by Chinese President Xi Jinping during his visit to Shanghai in 2025, according to Reuters reporting. SO028, SO029
CO014 AGIBOT has raised approximately $83.8 million in disclosed funding across its confirmed rounds, per Tracxn's company profile. SO035, SO036
CO015 Gaocheng Ventures (高瓴创投, GGV-affiliated) participated in an early AGIBOT financing round in approximately April 2023, per corporate registry data. SO037
CO016 BYD participated in an AGIBOT A+ financing round in August 2023, per Qichacha corporate registry data. SO037
CO017 Tencent Holdings participated in an AGIBOT B-round financing in approximately March 2025, per Qichacha corporate registry data. SO037
CO018 The Mirae Asset-LG Electronics New Growth Fund completed a strategic investment in AGIBOT in August 2025; neither the amount raised nor the implied valuation was disclosed. SO026, SO027, SO028
CO019 LG Electronics characterised its investment in AGIBOT as its first investment in the embodied intelligence sector. SO026, SO028
CO020 Major strategic investors in AGIBOT include Tencent Holdings, JD.com, BYD, SAIC Motor, BAIC Group, and TCL Technology Group, per Yicai Global reporting. SO027, SO037
CO021 Pandaily reported AGIBOT's valuation as "exceeding 7 billion yuan" (approximately $960 million at the time), likely reflecting a mid-stage round before the LG-Mirae investment. SO038
CO022 Tracxn's company profile reports AGIBOT's current valuation at $6.4 billion USD, conflicting with the earlier Pandaily yuan-denominated figure. SO036
CO023 AGIBOT has not disclosed revenue, ARR, gross margin, burn rate, or any other income statement metric in any reviewed public source. SO035, SO036
CO024 The AGIBOT A2 Ultra is described by the company as the industry's first full-size humanoid robot deployed commercially at scale. SO018, SO024
CO025 The A2 Ultra holds four major international certifications: CR, CE-MD, CE-RED, and FCC — enabling commercial sales in China, the EU, and the United States. SO013, SO018
CO026 The A2 Ultra set a Guinness World Record for the longest distance walked by a humanoid robot with a cross-provincial walk of 106.286 kilometres. SO013
CO027 The A2 Ultra has been deployed in over 20 leading enterprises, according to store listing and official product page claims. SO018, SO013
CO028 The AGIBOT X2 entertainment robot is listed on store.agibot.com at a sale price of $24,240 USD. SO019, SO016
CO029 The AGIBOT A2 Lite entertainment humanoid is listed on store.agibot.com at a sale price of $44,560 USD. SO020, SO013
CO030 The AGIBOT G1 factory robot features 26 degrees of freedom, operates in industrial, commercial, and domestic scenarios, and can continuously hold a 3 kg object at working heights over 2 metres. SO014
CO031 The AGIBOT A2-W mobile manipulator integrates chassis, arms, and perception system and supports deployment and commissioning within a single day. SO017
CO032 In January 2025, AGIBOT produced its 1,000th general-purpose embodied robot, which the company characterised as a new industry milestone for mass production. SO002, SO035
CO033 At CES 2026 in January 2026, AGIBOT announced cumulative shipments of 5,000 robots to date, marking its official US market debut. SO024, SO035
CO034 In March 2026, AGIBOT produced its 10,000th humanoid robot, with the 5,000-to-10,000 unit tranche completed in approximately three months. SO032, SO033, SO040
CO035 The acceleration from 5,000 to 10,000 units represented approximately a 4× increase in production speed compared to the prior ramp phase, per trade coverage of the milestone. SO034, SO040
CO036 AGIBOT G2 robots deployed at Longcheer Technology's tablet factory in Nanchang, Jiangxi completed 2,283 tasks with zero errors during an 8-hour continuous operation, per CnTechPost reporting of the April 2026 deployment. SO030, SO031, SO025
CO037 The Longcheer Technology deployment is described as the world's first large-scale industrial implementation of embodied AI within core production workflows in consumer electronics manufacturing. SO008, SO025, SO030
CO038 AGIBOT declared 2026 as "Deployment Year One" at its APC 2026 Shanghai Partner Conference on April 17, 2026, signalling a strategic pivot from production scale to commercial ROI. SO006, SO033
CO039 At APC 2026 (April 17, 2026), AGIBOT unveiled five new hardware platforms including the A3 full humanoid, D2 Max all-terrain quadruped, and G2 Air mobile manipulator. SO007, SO033
CO040 At APC 2026, AGIBOT introduced eight foundational AI models under its "One Robotic Body, Three Intelligences" architecture spanning locomotion, manipulation, and multimodal interaction. SO007, SO033
CO041 The AIMA open-stack ecosystem was launched at APC 2026 to lower integration barriers for third-party developers building and deploying embodied AI applications on AGIBOT platforms. SO006, SO033
CO042 AGIBOT made its official US market debut at CES 2026 in Las Vegas in January 2026, exhibiting its full humanoid robot portfolio at the show. SO024, SO035
CO043 At MWC 2026 in Barcelona (March 2026), AGIBOT's A2 robot guided Spain's King Felipe VI on a tour of the future airport exhibition, and the King shook hands with the robot. SO042
CO044 AGIBOT signed a Memorandum of Understanding with Genting Malaysia Bhd on April 17, 2026 to integrate humanoid agents into the Resorts World Genting hospitality ecosystem. SO032
CO045 AGIBOT co-initiated the First Hong Kong Embodied AI Industry Summit and APC 2026 Hong Kong on May 12, 2026, attended by Hong Kong Chief Executive John Lee Ka-chiu. SO004, SO003
CO046 AgiBot-World Colosseo is a large-scale robot learning platform published via OpenDriveLab on GitHub; AGIBOT has released the GO-1 and GO-1 Air foundation models trained on the dataset. SO022, SO011
CO047 The US American Security Robotics Act, introduced March 26, 2026 by Senators Tom Cotton and Chuck Schumer with bipartisan support, would prohibit US federal agencies from purchasing or operating humanoid robots made by Chinese companies. SO039, SO040
CO048 China introduced new humanoid robot standards in March 2026 that could raise compliance requirements and may alter the competitive dynamics for domestic players including AGIBOT. SO040, SO041
CO049 Genie Studio is a zero-code robot deployment platform launched by AGIBOT that enables partners to build complex workflows via drag-and-drop orchestration without deep coding expertise. SO009, SO021
CO050 AGIBOT's headcount is not officially disclosed; HumanoidIndex estimates 500+ employees, and active Chinese-language recruitment listings are visible on the company's careers page. SO035, SO023
CM001 The humanoid and embodied AI robot market is defined as commercially produced robots with a human-like or semi-humanoid form factor designed to operate in human-built environments using dexterous manipulation, mobile locomotion, and multimodal interaction; fixed-arm industrial robots, AMRs without manipulation, and software-only AI are excluded. SM001, SM008
CM002 The primary status-quo substitutes for humanoid robots are human labor, fixed-arm industrial robots (Fanuc, KUKA, ABB), cobots (Universal Robots UR-series), and specialized task robots (delivery bots, floor cleaners, barcode scanners). SM007, SM008
CM003 Included spend in the humanoid robot market covers hardware purchase or lease, embedded AI software licenses, deployment integration services, training data, and after-sale maintenance; excluded spend includes facility re-engineering for fixed automation and workforce retraining. SM009, SM018
CM004 Adjacent markets that humanoid robots may expand into as capabilities mature include industrial exoskeletons, cobot arms, AI-powered delivery robots, floor-care robots, and drone logistics platforms—each representing a substitution opportunity rather than a competing market. SM007, SM008
CM005 China's manufacturing industry had an operational stock of approximately 2 million industrial robots as of 2025, representing 4.5 times Japan's stock and accounting for 54% of annual global industrial robot installations, according to IFR World Robotics 2025 Report. SM007
CM006 Goldman Sachs estimates the global humanoid robot market at $38 billion by 2035 in its widely cited mid-case projection, with a base-case floor of at least $6 billion within 10-15 years and a blue-sky ceiling of $154 billion contingent on overcoming hardware cost, safety, and public acceptance barriers. SM001, SM006
CM007 MarketsandMarkets projects the global humanoid robot hardware market at $2.92 billion in 2025, growing to $15.26 billion by 2030 at a CAGR of 39.2%, based on bottom-up component and sensor segmentation published in April 2025. SM002
CM008 MarketsandMarkets projects the China humanoid robot market at $0.40 billion in 2025, growing to $2.80 billion by 2030 at a CAGR of 47.6%, outpacing the global average of 39.2% and reflecting China's structural demand advantage. SM002
CM009 MarketsandMarkets projects the US humanoid robot market at $857.9 million in 2025, growing to $4.601 billion by 2030 at a CAGR of 39.9%, with key US players including Agility Robotics, Boston Dynamics, and Tesla. SM002
CM010 IDC forecasts global humanoid robot shipments will exceed 510,000 units by 2030, representing a compound annual growth rate of approximately 95% from the 2025 base of roughly 15,000-20,000 units. SM003, SM004
CM011 A Goldman Sachs survey of eight Chinese humanoid robot and supply chain companies conducted in January 2026 estimated global shipments reached 15,000-20,000 units in 2025, with Chinese companies contributing the majority, and demand primarily from scientific research, robot AI training, education, entertainment, and data factories. SM004
CM012 ARK Invest projects the economic value of humanoid robots substituting human labor could reach $24 trillion, reflecting the potential for robots to capture a fraction of the $40+ trillion annual global human labor market, representing an extreme scenario rather than a hardware revenue forecast. SM006
CM013 Goldman Sachs's $38 billion hardware revenue estimate for humanoid robots by 2035 conflicts with ARK Invest's $24 trillion economic-value projection, a 600× gap attributable entirely to scope: Goldman Sachs measures robot hardware and software revenue while ARK measures the economic impact of labor displacement—making the two estimates fundamentally incompatible as market size benchmarks. SM001, SM006
CM014 Goldman Sachs stated in January 2026 that 2026 may become a critical year for "volume verification plus expected reset," as investors watch whether million-robot milestone expectations will be revised and how market share evolves among top Chinese manufacturers. SM004
CM015 Author Peter Diamandis projects 10 billion humanoid robots globally by 2040 at a labor equivalent of $10 per day per robot (approximately $0.40 per hour), based on leasing cost analogies to automobile financing; this is a maximalist scenario cited in mainstream media without published methodology. SM006
CM016 IDC ranked Agibot as the global leader in humanoid robot shipments for 2025, reporting 5,200 units delivered; rival Unitree Robotics disputed this with a claim of 5,500 shipments, with the discrepancy unresolved by independent audit, creating an unverified ranking contest. SM015, SM022
CM017 Agibot reached its 10,000th robot production milestone in March 2026, scaling from 5,000 to 10,000 units in three months, representing a 4x acceleration in production speed compared to the phase that reached the first 5,000 units. SM013, SM015
CM018 Agibot G2 completed 2,283 tasks with zero errors during an 8-hour continuous operation at Longcheer Technology's tablet factory, executing at 310 units per hour, supporting 24/7 double-shift operation, and achieving a success rate exceeding 99.9%, as reported in April 2026 press releases and corroborated by independent news coverage. SM010, SM011
CM019 Agibot has articulated seven standardized productivity solutions spanning industrial (production-line loading/unloading, material handling, logistics sorting), commercial (guidance/retail assistance, retail service stations, hospitality), and enterprise research/data contexts, reflecting a multi-segment market strategy for near-term revenue. SM018, SM019
CM020 Industrial manufacturing buyers in 3C electronics, automotive, and semiconductor sectors represent the highest-conviction near-term segment for humanoid robots, with budget decisions made at VP Manufacturing or CFO level against capital equipment budgets, driven by multi-model small-batch production demands that exceed fixed-automation flexibility. SM011, SM012, SM018
CM021 At Longcheer Technology, Agibot G2 robots process up to 5,700 units per day in 24/7 continuous operation; at HT-Tech and Intel, Agibot's deployed systems process up to one million and 100,000 chips per day respectively, with disk drop rates below 0.001%. SM018
CM022 In logistics and sorting deployments, Agibot's system at Damon achieves 70% of human-level throughput with a 99% second-pass success rate, while at Xiaonan Intelligence, irregular soft-package sorting has reduced labor costs by 50% with throughput reaching 435 items per hour. SM018
CM023 Agibot's retail and hospitality deployments include a reported 20% increase in foot traffic at Anta stores, double-digit conversion gains, a 13% operational cost reduction at Guangzhou Metro alongside 80% query resolution via self-service, and customer flow improvement at Haidilao—all company-reported figures without independent third-party audit. SM018, SM019
CM024 Agibot offers a global rental model at €899 per robot per day, enabling hospitality and event buyers to adopt robotics on an operational expense basis without a hardware purchase commitment, and the company has deployed robots in 17 countries through this and direct sales models. SM013, SM014
CM025 Research and education institutions were the dominant buyer segment for humanoid robots in 2025, accounting for a disproportionate share of global shipments according to Goldman Sachs's January 2026 survey, which cited scientific research, robot AI training, and data factories as primary demand sources. SM004, SM022
CM026 China's 15th Five-Year Plan (2026-2030), announced in May 2026, places robotics and embodied AI at the explicit center of the national industrial strategy, mandating that thousands of subordinate sectoral and regional plans align with robotics-led growth objectives under a framework described by IFR as shifting China from traditional automation to high-end intelligent robotics integrated with AI. SM007, SM017
CM027 Chinese domestic suppliers captured 57% of China's industrial robot installation market in 2024, up from 30% in 2020, representing a structural supply-chain advantage for Chinese humanoid robot manufacturers; in the metal and machinery industry, Chinese suppliers hold 85% domestic market share. SM007
CM028 China's Ministry of Industry and Information Technology published the world's first national standard system for humanoid robots and embodied intelligence in late February 2026 (Humanoid Robot and Embodied Intelligence Standard System 2026 Edition), covering six pillars: foundational standards, neuromorphic computing, limbs and components, system integration, application scenarios, and safety and ethics. SM017
CM029 US bipartisan American Security Robotics Act, introduced by Republican Senator Tom Cotton and Democratic Senator Chuck Schumer in March 2026, would ban federal government purchase or use of humanoid robots made by Chinese companies, citing risks of data transmission to China and remote control vulnerabilities. SM016
CM030 Agibot had deployed robots in 17 countries by early 2026 and established partnerships with Genting Malaysia (hospitality, Southeast Asia), Hyperscale Data (Michigan, US industrial), SIR Spa (Italy, EU automotive), Whale Cloud (global, 150+ markets), and Minth Group (EU manufacturing supply chain). SM013, SM014
CM031 Humanoid robot hardware costs range from roughly $10,000 to over $100,000 per unit depending on capability level, creating a meaningful capex barrier that limits adoption to markets where labor costs justify the investment—most viable in East Asian manufacturing and US logistics. SM006, SM015
CM032 Global manufacturing labor shortages provide a structural demand driver for humanoid robots, most acute in China (54% of global robot installs, rising domestic wages), the US (10M+ unfilled manufacturing and logistics jobs), and Japan (aging workforce); Goldman Sachs projects humanoids filling 4% of the US manufacturing labor gap by 2030. SM001, SM007
CM033 IFR explicitly cautions that mass adoption of humanoid robots as universal factory helpers or in private households will not happen within the near- and medium-term future, with China's own 15th Five-Year Plan placing commercial humanoid robot adoption toward the later years of the 2026-2030 period, and traditional industrial robots expected to retain dominance in high-speed precision manufacturing. SM007, SM008
CM034 Goldman Sachs identified production consistency challenges and inherent multi-stage testing processes as the key operational barriers to Chinese manufacturers achieving their ambitious 2026-2027 shipment growth targets of several-fold year-on-year growth from 2025 bases. SM004
CM035 US security legislation targeting Chinese humanoid robots could create a formal market bifurcation between Chinese-origin and Western-origin robots in government and critical infrastructure procurement—directly constraining Agibot's US addressable market in sovereign contexts while leaving the private sector accessible, and potentially influencing allied-nation procurement decisions. SM016
CM036 Agibot's Genie Sim 3.0 digital twin platform compressed the traditional on-site robot debugging timeline from months to 36 hours, enabling a four-month project-initiation-to- deployment cycle at Longcheer, addressing long integration cycles as a key adoption constraint. SM011, SM021
CM037 China's MIIT 2026 humanoid robot standard mandates emergency stop mechanisms, force limiting (preventing a robot arm from crushing a human finger), thermal management, behavioral safety requirements for predictable failure responses, and privacy safeguards, acknowledging that prior industrial robot incidents in Tesla's Fremont and Texas facilities illustrate the real risks of close-proximity human-robot operation. SM017, SM023
CM038 MarketsandMarkets's $15.26 billion hardware revenue estimate for the global humanoid robot market by 2030 is inconsistent with IDC's 510,000-unit volume forecast for the same horizon, which at plausible ASPs of $20,000-$50,000 implies $10-25 billion in revenue—leaving an unresolved gap attributable to different unit-economics assumptions and scope definitions. SM002, SM003
CP001 Agibot produced its 10,000th humanoid robot by March 2026, scaling from 5,000 to 10,000 units in just three months — ranked global #1 in humanoid shipments for 2025 by IDC and Omdia. SP003, SP005, SP006
CP002 Direct humanoid robot competitors to Agibot include Figure AI, Agility Robotics, Apptronik, and 1X Technologies in the Western market, and Unitree Robotics in the Chinese market. SP012, SP006
CP003 Incumbent automation substitutes for humanoid robots include traditional 6-axis industrial robot arms (Fanuc, KUKA, ABB) and AMRs (Locus, Vecna, Geek+), which address manufacturing and logistics tasks at lower cost and with longer operational proof records. SP010, SP012
CP004 Goldman Sachs Research projects the global humanoid robot market at a minimum of $6B in 10–15 years, reaching up to $154B by 2035E in a blue-sky scenario where all adoption hurdles are overcome. SP010
CP005 As of mid-2026, approximately ten companies globally have commercially deployed or commercially shipping humanoid robots, with volume and enterprise proof concentrated among Agibot, Unitree, Agility, and — at smaller scale — 1X Technologies. SP011, SP012
CP006 Figure AI raised $1.9B in total funding including a $1B+ Series C in September 2025, at a $39B valuation — the highest valuation among pure-play humanoid robot startups globally. SP024, SP013
CP007 Figure AI dropped its partnership with OpenAI in early 2025 and shifted to developing proprietary in-house AI models, ending its co-development arrangement with OpenAI. SP024, SP013
CP008 Figure AI's Figure 03 targets general-purpose home assistance; its prior Figure 02 model is deployed in a pilot at BMW's manufacturing facility in an undisclosed scope. SP013, SP024
CP009 Agility Robotics raised $641M+ in total funding, including a $400M Series C in June 2025. SP022, SP017
CP010 Agility Robotics' Digit humanoid robot is deployed commercially at Amazon fulfilment centres, GXO Logistics, and Schaeffler manufacturing, representing approximately 100 commercial units — the deepest enterprise-validated humanoid deployment in the industry. SP017, SP022
CP011 Apptronik raised $520M in a Series A round in February 2026, bringing total disclosed funding to $1.45B+ at a $5.5B+ valuation. SP023, SP018
CP012 Apptronik Apollo targets warehouses and manufacturing environments, deploying with Mercedes-Benz as a named partner under a Robotics-as-a-Service business model. SP018, SP023
CP013 1X Technologies raised $100M in a Series B in January 2024, bringing total funding to approximately $125M+; the company is OpenAI-backed and its NEO robot targets a $20,000 consumer price point with pre-orders open as of early 2026. SP015, SP016, SP021
CP014 1X NEO's pre-orders are open but no mass commercial deployment has been confirmed as of early 2026; the company's commercially deployed fleet is EVE (wheeled), not NEO (bipedal). SP014, SP021
CP015 Unitree Robotics claims 5,500 humanoid units shipped in 2025 and has filed for a $580M IPO in China, directly contesting Agibot's volume leadership claim for 2025. SP005, SP006
CP016 Unitree G1 is listed at $16,000 USD on the official Unitree product page, making it the lowest publicly available price for a commercially shipping bipedal humanoid robot globally. SP019, SP021
CP017 Unitree G1 has 23 degrees of freedom, weighs approximately 35kg, and features basic bipedal locomotion; its native autonomous AI stack is limited relative to full-stack humanoid competitors that offer integrated foundation models. SP019, SP025
CP018 Unitree's cost structure, enabled by Chinese manufacturing scale, allows it to price its G1 bipedal humanoid below the bill-of-materials cost that most Western competitors could achieve at equivalent hardware specifications. SP011, SP005
CP019 Agibot's "One Robotic Body, Three Intelligences" architecture integrates motion intelligence, interaction intelligence, and manipulation intelligence into a unified system, distinguishing it from competitors that treat these as separate stacks. SP002, SP003
CP020 Agibot A2 Ultra holds CR, CE-MD, CE-RED, and FCC certifications, enabling commercial sales across China, the EU, and US markets — a regulatory compliance advantage that most Chinese humanoid competitors have not yet achieved for their humanoid lines. SP002, SP001
CP021 Agibot A3 humanoid stands 173cm tall, weighs 55kg, features 10-hour battery endurance and 10-second battery swap, with UWB centimetre-level swarm positioning enabling synchronised 100-robot simultaneous performances. SP004
CP022 Agibot G2 industrial humanoid robot achieved 8-hour zero-error continuous operation on a consumer electronics precision manufacturing line at Longcheer Technology. SP003, SP004
CP023 Agibot offers a full-series portfolio spanning entertainment (X2 compact humanoid), commercial humanoids (A2/A3), industrial floor robots (G2), quadruped robots (D1), and a dexterous manipulation system (OmniHand) — breadth that most Western competitors targeting a single form factor cannot match. SP001, SP002
CP024 Agibot A2 Lite is listed at $44,560 USD and the X2 compact humanoid at $24,240 USD on Agibot's international online store, making Agibot one of the few humanoid companies with published international list pricing. SP001, SP002
CP025 No Western humanoid robot company — including Figure AI, Agility Robotics, or Apptronik — publicly discloses enterprise unit pricing or RaaS contract economics as of mid-2026. SP013, SP017, SP018
CP026 Agility Digit enterprise deployment pricing is not publicly disclosed; industry analyst estimates range from approximately $100,000–$150,000 per unit or equivalent annual RaaS contract value. SP017, SP022
CP027 Apptronik's RaaS model is publicly announced but no specific per-unit or monthly pricing has been disclosed; Mercedes-Benz deployment contract terms are private. SP018, SP023
CP028 Agility Robotics' Arc fleet-management platform integrates Digit with existing warehouse automation infrastructure (AMRs, WMS, and execution systems), enabling seamless on-boarding but deepening customer lock-in to the Agility ecosystem. SP017
CP029 Agibot operates a global rental platform and international direct-sale online store, contrasting with Western competitors that rely on enterprise contract sales with dedicated account management and high-touch support. SP002, SP003
CP030 Enterprise humanoid robot deployments create switching costs through three accumulating mechanisms: WMS/MES system integration, task-specific AI model fine-tuning, and staff requalification — all of which grow with deployment duration and scale. SP017, SP018
CP031 Agibot's Genie Studio software platform and OmniHand dexterous manipulation system represent tooling-layer lock-in mechanisms for enterprise customers who build workflows on Agibot's technology stack. SP003, SP004
CP032 Multi-homing between humanoid robot vendors is currently feasible because no single platform dominates any use case at meaningful scale, but switching costs will accumulate as deployment duration and integration depth increase. SP010, SP011
CP033 Agibot's supply-chain access through CATL-backed battery components, Longcheer precision manufacturing, and the broader Chinese supplier ecosystem gives it structural BOM cost advantages that Western competitors cannot easily replicate. SP009, SP005
CP034 Agibot's real-world fleet of 10,000 units generates proprietary embodied-AI training data at a scale that competitors without comparable deployment cannot match, creating a compounding data flywheel similar to Tesla's FSD training advantage. SP003, SP005
CP035 Agibot's Chinese manufacturing cost structure allows it to offer its A2 Lite at $44,560 and X2 at $24,240 — pricing that is likely below the cost floor for Western manufacturers to produce equivalent hardware specifications. SP001, SP011
CP036 China's MIIT published national humanoid robot standards (Humanoid Robot and Embodied Intelligence Standard System, 2026 Edition) in February 2026, potentially creating regulatory compliance alignment advantages for domestic manufacturers who helped shape them. SP007
CP037 US senators Tom Cotton and Chuck Schumer introduced the American Security Robotics Act in March 2026, proposing to bar the federal government from buying or operating humanoid robots made by Chinese firms, citing national security risks. SP008, SP007
CP038 US senators explicitly named Agibot and Unitree in the context of the American Security Robotics Act, noting that both Chinese firms were preparing to list shares in China while expanding US market presence. SP008
CP039 Goldman Sachs flags 2026 as a potential "volume verification and expectation reset" year in which top Chinese humanoid manufacturers' shipment targets may significantly outpace actual deployable enterprise demand. SP010, SP011
CP040 Figure AI ($1.9B), Agility Robotics ($641M+), and Apptronik ($1.45B+) collectively raised over $4B in 2024–2026, approximately 48× Agibot's disclosed $83M+ funding — a capital overhang that could sustain sustained R&D and enterprise-sales spending that Agibot cannot match. SP022, SP023, SP024
CI001 AGIBOT A2 Lite full-size entertainment humanoid robot is listed at $44,560 USD on the official AGIBOT online store (store.agibot.com); the listing also specifies a $500–$3,000 shipping deposit and confirms that customers bear all applicable import duties and taxes. SI001
CI002 AGIBOT X2 half-size entertainment humanoid robot is listed at $24,240 USD on the official AGIBOT online store (store.agibot.com); the listing confirms a multimodal interaction system (visual, voice, tactile), 25 degrees of freedom, and autonomous navigation capabilities with optional add-on packages. SI002
CI003 AGIBOT A2 Ultra is described on the official store as "the industry's first full-sized humanoid robot deployed commercially at scale" and applied in "over 20 leading enterprises" for showroom explanations, customer service, commercial performances, and brand endorsements; no public list price is stated on the product page, indicating B2B-negotiated pricing. SI003
CI004 The official A2 Lite product FAQ explicitly confirms that the group control software, upgraded skill packs, and VR remote control kit require "additional payment" beyond the $44,560 list price, establishing the existence of a software and accessory attach-revenue stream whose pricing is not publicly disclosed. SI001
CI005 AGIBOT G2 robots were deployed at Longcheer Technology's tablet production lines at MMIT stations in Nanchang, Jiangxi; the project required four months from initiation to mass-production deployment; this is described as the world's first large-scale industrial implementation of embodied AI in consumer electronics precision manufacturing. SI010, SI011, SI012
CI006 Humanoids Daily reported that AGIBOT's "global rental platform" offers humanoid robots at €899 per day for the hospitality vertical, with fleet spanning 17 countries as of March 2026; this is the only publicly disclosed rental rate for AGIBOT's RaaS offering. SI008
CI007 At APC 2026 (April 2026), AGIBOT launched the AIMA open-stack developer ecosystem targeting third-party developers building embodied AI applications on AGIBOT hardware; commercial terms, licensing fees, or API pricing for the AIMA ecosystem have not been announced as of May 2026. SI009, SI023
CI008 AGIBOT's business contact page (agibot.com/Contact/Business) lists the full product catalogue available for B2B sales enquiries, including A2, A2-W, A2-Max, G1, X1, X2, C5, Integrated Data-Solution for Embodied AI, Data Service, and partner co-operation categories (agent, vendor, media); this confirms a multi-SKU B2B sales motion beyond the publicly-listed store items. SI007
CI009 AGIBOT A2 Lite is "manufactured under automotive-grade production principles, with 92 comprehensive inspection checkpoints per unit and multiple aging and endurance tests," per the official product listing; this establishes material per-unit quality-assurance cost embedded in the manufacturing process. SI001
CI010 AGIBOT A2 Ultra was "validated through up to 2,000 hours of walking tests, including 360 hours of continuous operation without anomalies," per the official product page; this represents a significant test-time investment per robot unit and is consistent with premium-tier manufacturing economics. SI003
CI011 AGIBOT Genie Sim 3.0 compressed traditional months-long on-site debugging into 36 hours for the Longcheer G2 industrial deployment, enabling the full project (from initiation to production) to complete in four months; this is a confirmed reduction in per-deployment integration cost versus conventional industrial robot deployment. SI011, SI010
CI012 The AGIBOT G2 achieves a 95% equipment reuse rate when retraining for production-line changeover between product models, with retraining taking no more than four hours; this compares favourably to conventional dedicated robotic arms requiring customised tooling and lengthy reconfiguration. SI011, SI012
CI013 AGIBOT's production ramp accelerated from first 1,000 units (approximately 24 months), to 5,000 cumulative units (by early 2026), to 10,000 units (achieved March 2026 from the 5,000 milestone in approximately 3 months) — representing a 4× increase in production throughput over the most recent phase, consistent with supply-chain industrialisation investment. SI008, SI009
CI014 QCC corporate registry data (Qichacha) shows 40+ shareholders in AGIBOT including BYD (A+ round, August 2023), Tencent (B round, March 2025), JD.com/JoyGen (B+ round, May 2025), Longcheer Technology (603341.SH), and Deng Taihua as actual controller with approximately 33% beneficial ownership as of July 2025. SI013
CI015 AGIBOT produced its 10,000th humanoid robot in March 2026, with Co-founder and CTO Peng Zhihui and Senior Vice President Chuang Wang formally commemorating the milestone; the company declared this a "fundamental shift in ability to scale" from niche validation to large-scale commercial demand. SI008, SI009, SI005
CI016 Omdia data cited by HumanoidIndex ranks AGIBOT first globally in 2025 humanoid robot shipments with approximately 5,100–5,200 units delivered; this ranking is contested by Unitree Robotics, which claims 5,500 humanoid shipments for 2025 with competing methodology. SI016, SI008
CI017 AGIBOT G2 completed 2,283 tasks with zero errors during an 8-hour continuous operation at Longcheer's tablet factory, achieving throughput of 310 products per hour and a per-task cycle time of 18–20 seconds; 24/7 double-shift operation without human intervention was confirmed. SI011, SI010
CI018 Interesting Engineering independently reported AGIBOT G2 production-line performance as including throughput of up to 310 units per hour, cycle times of approximately 19–20 seconds per task, and a task success rate exceeding 99.9%, corroborating AGIBOT's own performance claims. SI012, SI011
CI019 AGIBOT plans to scale G2 production-line deployments to 100 units by Q3 2026, extending from consumer electronics (Longcheer tablets) into automotive manufacturing and semiconductor sectors, per CnTechPost reporting citing AGIBOT VP Yao Maoqing. SI011, SI009
CI020 AGIBOT's fleet was deployed across 17 countries as of March 2026 according to Humanoids Daily; commercial verticals cited include logistics, showrooms, hospitality, and industrial manufacturing. SI008
CI021 AGIBOT Genie Studio (launched 2025) and its 2026 successor Genie Studio Agent provide a one-stop embodied AI development platform enabling end-to-end workflows from data collection through model training to deployment; the zero-code deployment agent reduces the integration barrier for non-technical enterprise customers, which has implications for reducing per-deployment service cost. SI023, SI004
CI022 AGIBOT A2 Ultra product page confirms the robot "has been applied in over 20 leading enterprises" across showroom, customer service, commercial performance, and brand endorsement scenarios, providing a deployment-count floor for the commercial-performance revenue stream. SI003
CI023 Tracxn and HumanoidIndex independently report AGIBOT's total disclosed funding at approximately $83.8 million and $83 million-plus respectively; both sources characterise this as potentially understated given the number and diversity of shareholders in Chinese corporate registry data. SI016, SI017
CI024 QCC corporate registry shows BYD invested in an A+ round in August 2023, Tencent in a Series B in March 2025, and JD.com/JoyGen in a B+ round in May 2025, confirming the round progression cited in media reporting; the registry also lists Longcheer Technology (603341.SH) as a shareholder. SI013
CI025 LG Electronics and the Mirae Asset–LG Electronics New Growth Fund completed a strategic investment in AGIBOT in August 2025, marking LG Electronics' first investment in the embodied intelligence sector; the amount raised and post-money valuation were not disclosed by AGIBOT. SI014, SI015, SI018
CI026 CP Robotics (Danish robotics software firm) announced a strategic investment in AGIBOT on July 15, 2025, focused on developing applications in life sciences, new retail, new consumption, and health and wellness verticals, per Yicai Global reporting. SI015
CI027 Pandaily reported that AGIBOT completed a new funding round with a pre-money valuation exceeding 7 billion yuan (approximately $960 million at July–August 2025 exchange rates), though AGIBOT has not publicly confirmed the valuation and the precise round date is not specified. SI019
CI028 On July 8, 2025, Swancor Advanced Materials (a listed Shenzhen A-share company) disclosed that a consortium comprising AGIBOT and its core team agreed to pay up to CNY 2.1 billion (approximately $290 million USD) to acquire a controlling stake; this represents upstream vertical integration into high-performance materials for robot manufacturing. SI015, SI013
CI029 Following the Swancor disclosure, AGIBOT publicly clarified that it has no definite plans to make material changes to Swancor's main business within 12 months and denied that the acquisition constitutes a backdoor stock-market listing, citing upstream supply-chain access as the rationale. SI015
CI030 Reuters (via US News, March 2026) reported that "at least two Chinese firms — Agibot and Unitree — are preparing to list shares in China this year," implying an A-share or equivalent IPO is in active preparation; no prospectus, exchange filing, or underwriter appointment has been publicly confirmed by AGIBOT as of May 2026. SI020
CI031 AGIBOT has not disclosed revenue, ARR, gross margin, net margin, COGS, operating expenses, burn rate, cash on hand, or any income-statement or balance-sheet data as of May 2026; the company remains private and has made no regulatory disclosure requiring financial transparency. SI016, SI017
CI032 US Senators Tom Cotton and Chuck Schumer introduced the American Security Robotics Act in March 2026 to prohibit the US federal government from purchasing or operating unmanned ground vehicles (including humanoid robots) made by Chinese firms, with companion legislation introduced in the House by Rep. Elise Stefanik; the bill includes research exemptions for the military and law enforcement. SI020, SI018
CI033 Unitree Robotics contested AGIBOT's #1 shipment ranking by claiming 5,500 humanoid shipments for 2025 versus AGIBOT's IDC-cited 5,200; Unitree subsequently filed for a $580 million A-share IPO; the methodology behind competing shipment counts has not been independently reconciled. SI008, SI016
CI034 The Swancor acquisition consideration of up to CNY 2.1 billion (~$290 million) exceeds AGIBOT's total disclosed VC funding of approximately $83.8 million by roughly 3.5×; this implies AGIBOT's true total capital deployed (including undisclosed rounds and retained earnings, if any) is substantially above the disclosed VC figure, and that the company carries a large financing obligation relative to its publicly-known capital base. SI013, SI016, SI019
CI035 The 100-unit G2 expansion by Q3 2026 is a stated company plan as of April 2026, not an achieved milestone; the number of G2 units currently deployed at Longcheer (as of the report date) has not been publicly disclosed, and the planned expansion into automotive and semiconductor sectors has not yet been confirmed by any external customer announcement. SI011, SI009
CI036 Goldman Sachs research estimates the humanoid robot market could reach a minimum of $6 billion in 10–15 years, filling 4% of the US manufacturing labour shortage gap by 2030 and 2% of global elderly-care demand by 2035; in a blue-sky scenario, the market could reach $154 billion by 2035. SI021
CI037 IDC forecasts global humanoid robot shipments exceeding 510,000 units by 2030, representing a compound annual growth rate of approximately 95%; in 2025 more than 85% of humanoid deployments were concentrated in entertainment, education, and guided-tour scenarios; Chinese vendors are expected to account for approximately 95% of global shipments in 2025. SI022
CI038 Based on disclosed product lines and available deployment evidence, the majority of AGIBOT's confirmed revenue-generating activity in 2025 and early 2026 was in entertainment, commercial performance, and showroom applications (X2, A2 Lite, A2 Ultra); industrial manufacturing deployment (G2) is nascent and does not appear to have contributed materially to revenue as of the report date, though the company has declared this the priority for 2026. SI003, SI009, SI011
CI039 AGIBOT's gross margin, COGS per product, operating expenses, and net income/loss are entirely undisclosed; without these metrics it is impossible to assess whether the $44,560–$24,240 hardware price points generate positive gross margin at current production volumes, or to model a path to operating cash-flow breakeven. SI016, SI017
CI040 AGIBOT's industrial deployment contract terms with Longcheer Technology — including per-unit pricing, contract duration, service fees, and any revenue-sharing or ROI-guarantee provisions — have not been disclosed in any public statement, press release, or media report reviewed as of May 2026. SI010, SI011
CE001 Agibot's commercial robot portfolio spans at least six named product series: A2 (Yuanzheng/Expedition), X2 (Lingxi), G2/G1 (Genie), X1 (Lingxi research), A2-W (wheeled variant), and D1/D2 quadruped, plus the A3 and G2 Air announced at APC 2026. SE001, SE024
CE002 The A2 Ultra is Agibot's flagship full-size bipedal humanoid robot, commercially available and certified for sale in the US, EU, and China markets. SE002, SE015
CE003 The A2 Ultra has obtained four international certifications: CR (China), CE-MD (EU Machinery Directive), CE-RED (EU Radio Equipment Directive), and FCC (US radio frequency). SE002, SE024
CE004 The A2 Lite is listed at $44,560 USD in the official Agibot online store as of the report date. SE016
CE005 The A2-W wheeled humanoid integrates UniGrasp, Uni6DPose, and UniPlug atomic embodied AI capabilities with 275T on-board computing power for real-time object recognition, pose estimation, and operational decision-making. SE003
CE006 The X1 is an open-source research humanoid with 29 joints (R86-2×9, R86-3×6, R52×10, L28×4) and two grippers, with full BOM, STEP, SolidWorks drawings, and SOP published on open channels. SE008, SE004
CE007 The X2 / X2 Ultra compact humanoid offers 25–30 degrees of freedom including one DOF for the head and three for the waist, and is listed starting at approximately $24,240 USD in the official store. SE005, SE018
CE008 The G1 data-collection robot features eight high-resolution cameras on the upper body, six-axis force sensors on both arms, millisecond-latency VR/motion-capture tele-operation, and asynchronous cloud data transmission with a local backlog limited to under two hours. SE006
CE009 The G2 industrial humanoid has been deployed in a live consumer electronics mass-production line at Longcheer Technology's tablet factory, representing the first documented large-scale embodied AI deployment in 3C precision manufacturing. SE025, SE026, SE027
CE010 The A3, announced at APC 2026 in April 2026, stands 173 cm tall, weighs 55 kg, achieves a power-to-weight ratio of 0.218 kW/kg, provides 10-hour battery endurance, and supports 10-second battery swap alongside UWB centimetre-level swarm positioning for 100-robot synchronised performances. SE022, SE029
CE011 Agibot's "One Robotic Body, Three Intelligences" architecture partitions robot AI into locomotion intelligence, manipulation intelligence, and interactive intelligence layers, each served by dedicated foundation models. SE024, SE029
CE012 The Behavioral Foundation Model (BFM) learns human-like movement from short video demonstrations, and the Generative Control Foundation Model (GCFM) enables text- and audio-prompted real-time movement improvisation for locomotion intelligence. SE029
CE013 GO-2 (VILLA) uses Action Chain-of-Thought reasoning to decompose high-level task instructions into precise motor commands for manipulation intelligence, supplemented by Genie Sim 3.0 for rapid sim-to-real transfer. SE029
CE014 The WITA Omni model is a robot-native multimodal system unifying vision, speech, and gesture in a single pipeline for interactive intelligence, replacing a stitched combination of separate language and vision modules. SE029
CE015 AgiBot World Beta, released March 2025, contains 1,003,672 manipulation trajectories totalling approximately 43.8 TB of real-robot data collected by a fleet of 100 robots across five target domains. SE012, SE013
CE016 The GO-1 foundation model and the lightweight GO-1 Air variant were open-sourced on September 19, 2025 and are available on HuggingFace under the agibot-world organisation with fine-tuning scripts. SE012, SE013
CE017 GO-1 Air is the lightweight variant of GO-1 without the Latent Planner module, designed for lower-resource inference on edge hardware while retaining the core manipulation policy. SE012
CE018 AimRT is an actively maintained open-source robotics middleware framework, reaching v1.7.0 in April 2026 with monthly releases providing real-time communication, ROS 2 (including Jazzy) interoperability, zenoh, gRPC, and MQTT backends. SE014, SE032
CE019 The SOP (Scalable Online Post-training) framework enables fleet-level VLA model updates through multi-robot parallel execution that shares a single policy across the fleet, centralized cloud online training, and instant model synchronisation back to each device. SE010
CE020 After SOP post-training, Agibot's robots operated autonomously on target tasks for over 36 consecutive hours without requiring human intervention, as reported in the official SOP research blog. SE010
CE021 Genie Studio is described by Agibot as the industry's first one-stop embodied AI development platform, covering data collection, model training, simulation evaluation, and one-click cloud-to-device deployment in a single interface. SE011
CE022 Genie Studio supports data collection throughput of up to 1,000 frames per device per day and employs a distributed architecture for dual-mode (real-time and offline) data processing with zero-delay data supply to training pipelines. SE011
CE023 Genie Studio's simulation environment (Genie Sim 3.0) contains over 8,000 high-precision 3D object assets and environment configurations enabling automated generation and evaluation of over 100 tasks. SE011
CE024 Genie Studio's simulation calibration achieves less than 5% error between GO-1 model test results and real-device performance outcomes, according to official Genie Studio documentation. SE011
CE025 Genie Studio Agent is a zero-code application deployment platform that enables non-technical users to configure robot workflows by dragging and connecting prebuilt nodes (perception, navigation, VLA, RL toolchain) without writing code, validated in simulation before real-device deployment. SE021
CE026 G2 robots were integrated into Longcheer Technology's tablet production line in 36 hours using Genie Sim 3.0 digital twin simulation, compressing a process that traditionally takes months. SE025, SE026
CE027 G2 achieves sustained throughput of 310 units per hour at Longcheer with a cycle time of approximately 19–20 seconds per MMIT loading/unloading operation. SE025, SE026, SE027
CE028 The G2 completed 2,283 consecutive tasks without a single error during an independently-reported eight-hour continuous factory operation at Longcheer's Nanchang facility. SE026
CE029 G2 supports mixed-model production changeovers without custom tooling; retraining for a new product model requires no more than four hours, enabling rapid production line reconfiguration. SE026, SE025
CE030 G2's equipment reuse rate across different product models in the Longcheer deployment reaches 95%, compared to traditional automation which requires model-specific tooling. SE026
CE031 Agibot plans to expand G2 deployment at Longcheer to 100 robots by Q3 2026 and extend the model into automotive manufacturing and semiconductor sectors. SE025, SE026
CE032 Agibot's published security update support policy guarantees at least one year of security patches (including vulnerability fixes and security improvements) from the product launch date. SE020
CE033 The Defined Support Period V4 document lists the A2 (launched January 2026), G2 (launched January 2026), and X2 Ultra (launched January 2026) as having security update support through January 1, 2027, alongside the Aimmaster control app, App, and PAD software. SE020
CE034 The AgiBot-World GitHub repository was named an IROS 2025 Best Paper Award Finalist and is also being submitted to IEEE Transactions on Robotics (TRO) in 2026. SE012
CE035 Peng Zhihui, Agibot co-founder and CTO, serves as a deputy director of China's MIIT HEIS Humanoid Robots and Embodied Intelligence Standardisation Technical Committee (designation MIIT/TC8). SE031
CE036 At APC 2026, Agibot formalised an L1–L5 embodied AI capability framework and claimed its current systems have achieved Level 3 (L3, characterised by scenario-level autonomy) in both locomotion and task intelligence. SE029
CE037 The OmniHand 3 Ultra-T dexterous hand, announced at APC 2026, features a 22+3 DOF tendon-driven system, 3D tactile sensing, a lightweight 500 g design, and a 10:1 load-to-weight ratio. SE022
CE038 The "Powered by AGIBOT" programme enables third-party partners and OEMs to customise robot hardware form factors, external appearance, and AI algorithms on top of Agibot's standardised mass-production base, reducing integration engineering barriers. SE024
CE039 AimRT supports ROS 2 (including the Jazzy distribution) with version-based compatibility, zenoh shared-memory and network communication, gRPC, HTTP, TCP, UDP, and MQTT backends, providing broad protocol interoperability for robot system integration. SE014
CE040 Agibot's website privacy policy (effective April 30, 2026) designates AGIBOT PTE. LTD. (Singapore) as the data controller for all markets except Japan, Korea (Agibot K.K.), and Hong Kong SAR (AgiBot Innovation (HongKong) Technology Co., Limited). SE019
CE041 China's MIIT HEIS humanoid robot standardisation framework (2026 edition) mandates specifications for structural integrity, emergency stop mechanisms, thermal management (battery overheating prevention), force limiting, and minimum-risk-condition behaviour on software failure. SE031
CE042 Peng Zhihui acknowledged at the MIIT HEIS committee that in industrial scenarios nearly 80% of tasks where humans outperform traditional automation are strongly related to tactile sensing, and the lack of standardised tactile sensors remains a critical industry bottleneck including for Agibot's own deployed products. SE031
CU001 AGIBOT targets four primary customer segments: industrial manufacturing (electronics assembly, automotive), logistics and warehousing, hospitality and retail services, and enterprise IT or data center operations. SU011, SU016
CU002 The buyer in manufacturing deployments is typically the plant operations or manufacturing engineering function, while in the RaaS channel, payment comes from site operators or service businesses that contract the robot fleet on a daily or monthly basis. SU010, SU002
CU003 AGIBOT's primary geographic markets at commercial launch are mainland China for manufacturing, Southeast Asia for hospitality and services, Europe for luxury services, and Singapore for RaaS trials. SU011, SU015
CU004 AGIBOT's RaaS model serves 17 countries and regions as of March 2026, with a daily rate starting at €899 per robot — equivalent to approximately €26,970 per robot per month assuming 30 days. SU015, SU002
CU005 AGIBOT claims over 20 leading enterprise partners and customers from across manufacturing, logistics, and services participated in the APC 2026 as of April 2026, but the complete named list has not been publicly disclosed. SU003, SU013
CU006 AGIBOT's commercial use cases span production-line loading and unloading, electronics assembly, logistics pick-and-place, hotel and spa services, and data center operations, as announced via seven standardized productivity solutions at APC 2026. SU016, SU011
CU007 AGIBOT's contact portal and store indicate a direct enterprise sales process with no disclosed reseller or channel partner network for manufacturing deployments; the RaaS channel is available directly via store.agibot.com without a disclosed intermediary. SU025, SU002
CU008 AGIBOT produced approximately 10,000 cumulative robots by March 2026, reaching this milestone faster than any predecessor in the embodied AI sector according to reporting at the time. SU014, SU013
CU009 AGIBOT was ranked first globally by shipped unit volume in 2025 with approximately 5,100 units, according to Omdia data cited by Humanoid Index. SU022, SU014
CU010 Longcheer Technology's AGIBOT G2 deployment reached 310 units per hour (UPH) throughput with a 99.9% motion success rate, 19–20 second cycle time, and 100% on-time delivery, representing the highest disclosed productivity metrics for a humanoid robot commercial deployment as of mid-2026. SU010, SU012
CU011 AGIBOT and Longcheer plan to scale the G2 robot fleet to 100 units on the Longcheer production line by Q3 2026, up from an initial pilot cohort. SU010, SU001
CU012 Longcheer Technology integrated AGIBOT G2 robots into its production line in 36 hours and ramped from pilot to 24/7 production deployment in approximately four months. SU010, SU006
CU013 The APC 2026 in April 2026 attracted attendees from 17 countries and regions; AGIBOT characterised the event as the launch of a new era of embodied AI productivity with global enterprise deployment partners. SU003, SU005
CU014 AGIBOT's adoption trajectory expanded from zero commercial deployments in mid-2024 to over 5,000 units shipped in 2025, representing a hypergrowth trajectory among humanoid robot manufacturers globally. SU014, SU022
CU015 AGIBOT's official positioning characterises 2026 as "Deployment Year One," signalling a strategic pivot from product development to commercial scaling with enterprise customers. SU013, SU017
CU016 Longcheer Technology is AGIBOT's sole publicly confirmed production-level customer as of May 2026, with G2 robots actively deployed on MMIT consumer electronics assembly lines in Shenzhen and Dongguan operating 24/7 across multi-shift schedules. SU010, SU001, SU017
CU017 Longcheer Technology's deployment metrics include 310 UPH throughput, 99.9% motion success rate, 19–20 second cycle time, 100% on-time delivery, and operation 16 hours per day with 8 hours fully autonomous. SU010, SU012
CU018 Genting Malaysia Berhad signed a non-binding memorandum of understanding with AGIBOT at the APC 2026 in April 2026 to explore humanoid robot deployment across hospitality and theme park operations; no production contract or deployment date was confirmed as of May 2026. SU007, SU009, SU011
CU019 Hyperscale Data, operator of a 100,000 square foot data center in Michigan, USA, signed a partnership agreement with AGIBOT at APC 2026; the deployment phase (pilot or production) and timeline were not confirmed in publicly available sources. SU011
CU020 SIR Spa Group (Italy) and Singtel (Singapore) are cited as early commercial partners of AGIBOT through its RaaS programme; no deployment metrics, contract values, or production timelines have been publicly disclosed for either. SU011, SU008
CU021 BYD and JD.com are confirmed as strategic investors in AGIBOT but their roles as active production-level customers of AGIBOT robot deployments have not been independently confirmed in publicly available sources. SU011, SU022
CU022 Whale Cloud, an Alibaba Group subsidiary, joined AGIBOT's partner ecosystem for fleet management software integration, with claimed reach to 150 or more markets; no independent press release or Whale Cloud announcement was identified corroborating this partnership. SU011, SU003
CU023 AGIBOT's APC 2026 showcased seven standardized productivity solutions covering manufacturing loading and unloading, electronics assembly, logistics, and hospitality, enabling rapid configuration for multiple verticals from a common robot platform. SU016, SU004
CU024 The PRNewswire joint announcement with Longcheer is the strongest independent customer-proof artefact for AGIBOT; all other named customer references originate from AGIBOT's own APC 2026 coverage, trade media, or company partner announcements without direct customer corroboration. SU010, SU001
CU025 Longcheer Technology's official corporate website describes its position as an AI-enabled smart manufacturing partner and highlights co-deployment of embodied AI on precision production lines, providing independent customer-side corroboration of the deployment. SU001
CU026 AGIBOT's APC 2026 published video testimonials and a customer Q&A featuring international attendees, but no specific deployment outcomes or commercial metrics were disclosed by any non-Longcheer customer in the public record. SU005, SU003
CU027 AGIBOT's Contact/Business portal offers a direct enterprise inquiry channel with no disclosed reseller or channel partner intermediary for the manufacturing and RaaS sales process, suggesting a direct-sales-led commercial model at this stage. SU025, SU002
CU028 AGIBOT has not publicly disclosed net revenue retention (NRR), gross revenue retention (GRR), contract renewal rates, or customer churn data as of May 2026; all such metrics are treated as open questions pending first-party disclosure.
CU029 The four-month ramp from pilot to production at Longcheer suggests an engagement model where pilots convert to live production deployments; the contract conversion rate and typical commercial contract length for AGIBOT's enterprise deployments are undisclosed. SU010
CU030 AGIBOT's RaaS model terms — including minimum contract duration, termination clauses, SLA commitments, and robot ownership structure — have not been disclosed beyond the headline price of €899 per robot per day.
CU031 The Genting Malaysia MOU is a non-binding letter of intent and does not establish contract value, deployment timeline, SLA terms, or recurring revenue commitments; it cannot be treated as evidence of durable recurring revenue. SU007, SU009
CU032 AGIBOT's commercial production revenue is entirely concentrated in the Longcheer Technology relationship as of May 2026; a single production customer represents 100% concentration risk if the relationship is paused or terminated. SU010, SU022
CU033 AGIBOT's APC 2026 ecosystem model — standardised productivity solutions marketed through a network of global industry partners — creates channel dependency risk if partner MOUs do not convert to signed production contracts within a reasonable window. SU003, SU016
CU034 Longcheer Technology received Lenovo's 2026 Perfect Quality Award in April 2026, following the AGIBOT G2 deployment; the award suggests a manufacturing quality uplift attributable in part to the robot deployment, which creates a retention incentive for Longcheer to maintain or expand the AGIBOT fleet. SU001
CU035 The American Security Robotics Act, introduced in March 2026 by Senators Cotton and Schumer, proposes to prohibit US federal agencies from purchasing or using robots manufactured by Chinese companies, directly blocking AGIBOT from the US government procurement channel. SU018
CU036 The proposed American Security Robotics Act does not affect commercial enterprise procurement, private-sector RaaS contracts, or international sales; its near-term impact on AGIBOT's existing customer base and commercial pipeline is limited to the US federal government segment. SU018
CU037 Legal liability frameworks for humanoid robot commercial deployments are unresolved; Hill Dickinson's 2026 analysis identifies product liability gaps, operator health and safety obligations, and unclear accountability frameworks as material risk factors for enterprises deploying humanoid robots on factory floors. SU019
CU038 AGIBOT's customer base is geographically concentrated in China for production deployments as of May 2026, with international relationships limited to MOUs and RaaS pilots; geographic concentration adds revenue risk in a scenario of China market disruption. SU022, SU011
CU039 AGIBOT has not published any customer satisfaction score, NPS measurement, G2 rating, or Gartner Peer Insights review; independent customer satisfaction data is unavailable for any AGIBOT product or deployment as of May 2026.
CU040 China's MIIT 2026 humanoid robot national safety standards, expected to be finalised in 2026, may impose additional certification requirements that affect lead times for new customer deployments and international market acceptance. SU023
CU041 AGIBOT's store.agibot.com lists the C5 cleaning robot at $32,900 and provides direct online ordering for multiple product models, indicating a self-service direct-to-enterprise sales channel operating in parallel with the RaaS subscription model. SU002
CU042 No public evidence of customer churn, failed pilot cancellations, safety incidents, or contract terminations was found in any sources reviewed for this chapter; absence of evidence does not imply zero churn given the early commercial stage and limited public disclosure.
CU043 AGIBOT's "Deployment Year One" is a company-authored marketing characterisation; the company has shipped units to buyers who may include internal testers, distribution partners, or end customers — the exact commercial revenue contribution per customer segment has not been disclosed. SU013, SU017
CU044 IDC projects approximately 510,000 humanoid robot units by 2030, implying roughly 95% CAGR from current levels; AGIBOT's 5,100 units shipped in 2025 represent approximately 1% of that projected 2030 volume, illustrating significant TAM headroom alongside substantial execution risk. SU021, SU022
CU045 Goldman Sachs estimates the global humanoid robot market will reach $6 billion within 10–15 years from 2025, with a blue-sky scenario of $154 billion by 2035; these estimates provide the macro tailwind context for AGIBOT's customer acquisition strategy. SU020, SU021
CR001 The American Security Robotics Act (ASRA), introduced on 26 March 2026 by Senators Tom Cotton (R-AR) and Chuck Schumer (D-NY), would prohibit the US federal government from purchasing or operating humanoid robots made by Chinese firms, citing national security and data-exfiltration risks. SR001
CR002 The ASRA bill identifies Chinese humanoid robot startups, including Agibot and Unitree, as the target firms and was reported by Reuters as naming them explicitly in the legislative context. SR001
CR003 A companion bill to ASRA was simultaneously introduced in the US House of Representatives by Representative Elise Stefanik (R-NY), signalling bipartisan bicameral momentum for restrictions on Chinese humanoid robots in government procurement. SR001
CR004 ASRA would bar federal funds from being used in connection with Chinese-made ground robots and would contain limited exemptions for military and law enforcement to research Chinese robots as long as the robots cannot transmit data to or receive data from China. SR001
CR005 Agibot's Website Privacy Policy, effective 30 April 2026, discloses data collection of device identifiers, usage data, and technical information from robot operations, with provisions for international data transfer and sharing with third-party service providers. SR004
CR006 As a Chinese-incorporated entity, Agibot is subject to the PRC National Intelligence Law (2017), which requires all Chinese organisations to assist state intelligence work on request; this obligation cannot be contractually waived with overseas customers. SR004, SR017
CR007 The Hill Dickinson legal analysis identifies the core unresolved liability question for humanoid robots: when an autonomous humanoid robot causes harm, responsibility allocation between manufacturer, operator, and software provider is legally unsettled globally as of 2026. SR003
CR008 US OSHA confirms there are currently no specific OSHA standards for the robotics industry; robot accident risks during non-routine operations are identified but no humanoid-specific safety compliance framework exists in US occupational health law. SR023
CR009 ISO 10218-1:2011, the predecessor industrial robot safety standard, explicitly does not address bipedal humanoid robots operating in open environments; the standard is withdrawn and under revision; no ISO standard for general-purpose humanoid robots exists as of May 2026. SR025
CR010 Agibot's officially published Product Security Update Support policy sets a 12-month security update window from the January 2026 launch date for the A2, G2/AgiBot G2, and X2 Ultra models, with security update support ending January 2027 — critically short for enterprise capital equipment. SR005
CR011 Agibot software products — Aimmaster, App, and PAD — have a defined security update support end date of 2 January 2027 per the published support policy, creating a coordinated product and software security support cliff 12 months after commercial launch. SR005
CR012 Agibot's G2 robot completed 2,283 tasks with zero errors during an 8-hour continuous operation at Longcheer's Nanchang tablet factory in April 2026, demonstrating 99.9%+ task success rate in that controlled setting; this is the primary public deployment evidence for reliability claims. SR010, SR011
CR013 The Longcheer deployment involved pick-and-place precision tasks at multimedia integrated testing (MMIT) stations with defined cycle times of 19–20 seconds, representing a constrained set of tasks rather than the full range of unstructured manipulation Agibot demonstrates in promotional material. SR010, SR011
CR014 MIIT's HEIS committee acknowledged that in industrial scenarios, nearly 80 percent of high-value manipulation tasks where humans excel but traditional automation struggles are strongly related to tactile sensing — a capability not yet standardised and described as a critical bottleneck by Agibot's CTO Peng Zhihui. SR002
CR015 IDC's attribution of 5,200 Agibot units shipped in 2025 was contested by rival Unitree, which claimed 5,500 humanoid shipments and filed for a USD 580 million IPO; shipment data is unaudited and cannot be independently verified, creating a diligence gap on the commercial evidence foundation. SR012, SR018
CR016 Agibot's Genie Sim 3.0 digital twin platform compressed traditional months-long on-site deployment debugging to 36 hours, but creates a single-point-of-failure dependency on proprietary simulation infrastructure for all new deployment integrations. SR010, SR008
CR017 No public MTBF (mean time between failures), field return rate, or quality escape rate data for Agibot robots at production volume has been published; the single Longcheer deployment evidence is insufficient to characterise long-term reliability across diverse deployment environments. SR010, SR011, SR012
CR018 Agibot accelerated production from 5,000 to 10,000 units in three months, representing a 4x increase in production speed; scaling from 10,000 to 50,000–100,000 units requires qualitatively different manufacturing investment in tooling, yield management, and supply chain redundancy. SR012, SR002
CR019 US Bureau of Industry and Security Export Administration Regulations have progressively tightened restrictions on advanced computing integrated circuits to China since 2022; AI inference processors used in humanoid robot embedded systems may be subject to export control restrictions affecting non-China supply chain access for Agibot. SR024
CR020 CATL, a strategic investor in Agibot per the Humanoid Index, is also a dominant Chinese lithium battery supplier; battery supply concentration creates a dual dependency where investor alignment and supply chain dependence may create preferential terms unavailable to Agibot's competitors. SR018
CR021 IFR reported in May 2026 that China's manufacturing sector has an operational stock of approximately 2 million industrial robots, with 54% of global annual industrial robot installations occurring in China; this concentrated domestic market intensifies competition for Agibot within China itself. SR028
CR022 Agibot has no publicly confirmed field service infrastructure, regional support offices, or certified integrator network outside China as of May 2026; international deployment at CES 2026 market entry level lacks operational support scalability. SR013, SR015
CR023 The APC 2026 conference confirmed Agibot's expansion plans into automotive manufacturing and semiconductor industries, both of which have substantially higher reliability, contamination, and safety requirements than 3C electronics manufacturing — presenting new qualification hurdles not yet addressed in public deployment evidence. SR010, SR013
CR024 Agibot's robots were demonstrated to Chinese President Xi Jinping during his visit to Shanghai, establishing a visible high-level political endorsement that creates both a domestic market advantage and a geopolitical concentration risk in Western markets. SR016, SR015
CR025 Longcheer Technology is the most prominently documented commercial customer, with a published deployment and a stated expansion plan to 100 units by Q3 2026; no other named commercial customer with production-scale deployments has been disclosed as of May 2026. SR009, SR010
CR026 Agibot's strategic investor base includes BYD, SAIC Motor, BAIC Group, and JD.com — entities with state-enterprise or state-affiliated ownership — creating governance complexity and potential preferential-customer dynamics that may not replicate at arm's-length commercial terms for independent enterprise buyers. SR017
CR027 CFIUS is authorised to review transactions involving foreign investment in the United States that could affect national security; Agibot's investor base, including state-enterprise shareholders and LG Electronics (Korean), would attract scrutiny in any US acquisition, joint venture, or major US institutional capital raise. SR027
CR028 LG Electronics' investment in Agibot, its first in the embodied intelligence sector, signals Korean strategic interest but creates a cross-border investment structure that amplifies CFIUS and US regulatory scrutiny of any US market entry activity. SR015, SR016
CR029 China's 15th Five-Year Plan (2026–2030) places robotics at the core of the nation's industrial modernisation strategy per IFR's May 2026 report, signalling continued state subsidies to competing domestic humanoid robot firms and intensifying competitive pressure on Agibot. SR028
CR030 Agibot's valuation exceeded 7 billion yuan (approximately USD 970 million) as of the most recent disclosed round; Goldman Sachs' base-case global humanoid robot market estimate is USD 6B in 10–15 years, indicating that the current valuation requires substantial market share capture to justify investor returns. SR014, SR019
CR031 Goldman Sachs described 2026 as a critical year for "volume verification and expectation reset" in humanoid robotics, with investor attention focused on whether ambitious unit targets and supply chain company valuations are sustainable — directly applicable to Agibot's rapid production claims. SR019, SR021
CR032 Agibot has disclosed no revenue, gross margin, operating loss, or burn rate in any publicly available source as of May 2026; total disclosed funding is approximately USD 83M+ per Humanoid Index, which may understate total financing from undisclosed strategic rounds. SR018, SR014
CR033 Agibot is reported to be preparing an A-share listing on China's STAR Market, subject to CSRC disclosure requirements; the IPO creates public market scrutiny of commercial deployment milestones and introduces incentives for aggressive revenue recognition and delivery timeline commitments. SR026, SR012
CR034 IDC forecasts global humanoid robot shipments to exceed 510,000 units by 2030 at approximately 95% CAGR; achieving a meaningful share of this market requires capital-intensive manufacturing scaling well beyond Agibot's current 10,000-unit production base. SR020
CR035 The EU AI Act (Regulation 2024/1689) fully applicable from August 2026 classifies AI systems used in safety-critical or high-risk settings as high-risk AI requiring conformity assessment, CE marking, and human oversight obligations — creating market access compliance burdens for Agibot in European industrial environments. SR029, SR030
CR036 Agibot's co-founder and CTO Peng Zhihui, a former Huawei "genius-level" engineer, is the company's primary technical face and serves as deputy director of the MIIT HEIS standards committee; his prior Huawei affiliation reinforces Western security concerns and creates key-person dependency in both technical leadership and regulatory influence. SR012, SR002
CR037 No public succession plan, deputy CTO, or technical leadership backup for Peng Zhihui or CEO Edward Deng has been disclosed; concentrated founder dependency is characteristic of this company stage but warrants specific diligence on continuity risk. SR006, SR007
CR038 Unitree's contested shipment claim and USD 580 million IPO filing creates direct competitive pressure on Agibot's volume leadership position; if Unitree achieves its 20,000-unit 2026 target, Agibot's dominant market position narrative would require revision. SR012
CR039 China's MIIT HEIS standards committee, in which Agibot holds a senior role, may embed technical requirements that advantage Chinese domestic champions while creating non-tariff barriers to foreign competitors — a strategy the Clingendael Institute identified as central to China's Standards 2035 initiative. SR002
CR040 The EU AI Act's high-risk classification for autonomous robots operating near humans, combined with GDPR and national data protection laws, creates a multi-jurisdictional compliance burden for Agibot's international expansion that has no equivalent constraint on domestic China deployment. SR029, SR030, SR003
CR041 Agibot's AIMA ecosystem was launched as an open-stack architecture for third-party developers; if third-party applications introduce safety failures or data breaches, Agibot faces potential platform liability as the underlying system provider under product liability regimes in multiple jurisdictions. SR013, SR003
CR042 Industrial safety incidents involving robots — including the Tesla/Fanuc 2023 worker injury case — demonstrate that even well-engineered robotic systems can fail unpredictably during maintenance or non-routine operations; humanoid robot commercial deployments face similar incident risk with greater public visibility. SR002, SR023
CR043 Agibot's rapid pace of new product launches — four new robotic platforms and eight new AI models at APC 2026 — creates a roadmap execution risk: simultaneous scaling across multiple product lines while maintaining quality at volume is operationally complex and has not been demonstrated beyond the single Longcheer deployment reference case. SR007, SR013
CV001 Agibot produced its 10,000th humanoid robot in March 2026, reaching the milestone in approximately three months after crossing the 5,000-unit mark — a 4x acceleration in production throughput representing the fastest production ramp of any humanoid robot company globally as of Q1 2026. SV001, SV011, SV012
CV002 Omdia (cited by Humanoid Index) ranks Agibot #1 in global humanoid shipments for 2025, with approximately 5,100 units shipped. This ranking is the primary basis for Agibot's claim to global production leadership; the methodology excludes research and show units from some competitors but has not been independently cross-validated. SV001, SV025, SV026
CV003 Agibot's G2 robot completed 2,283 tasks with zero errors during an 8-hour continuous operation at Longcheer Technology's tablet production facility in Nanchang, Jiangxi, achieving 310 products per hour and a cycle time of 18–20 seconds — confirming industrial-grade reliability at a live production line. SV013, SV014, SV016
CV004 Agibot's pre-money valuation exceeded CNY 7 billion (~$960M at July 2025 exchange rates) at the time of the LG Electronics / Mirae Asset strategic round in August 2025. The post-money valuation is estimated at $1.0B–$1.3B based on this anchor and typical undisclosed round-size conventions for Chinese private companies at this stage. SV003, SV004, SV010
CV005 LG Electronics and the Mirae Asset-LG Electronics New Growth Fund made their first-ever investment in China's embodied intelligence sector through a strategic round in Agibot in August 2025. The round was described as strategic; the investment amount was not disclosed. SV005, SV006, SV010
CV006 Agibot's QCC corporate registry listing shows more than 50 shareholders, including BYD (A+, 2023), Tencent Ventures (B, March 2025), JD.com / JoyGen (B+, May 2025), Gaorong Capital, Sequoia China (via shareholder vehicles), TCL Ventures, Longcheer Technology (603341.SH), and Wolong Electric Drive (600580.SH) — confirming a supply-chain-integrated, ecosystem-linked investor base extending well beyond the disclosed $83.8M in venture capital. SV003, SV002
CV007 Goldman Sachs' "Humanoid Robot: The AI Accelerant" report and IDC's 2026 humanoid robotics commercialization analysis both identify manufacturing scale and industrial deployment proof as the two primary value drivers for private-market humanoid robot valuations; Agibot leads globally on both dimensions as of May 2026. SV007, SV008, SV009
CV008 Agibot's ~$960M pre-money valuation at the August 2025 round represents a 45–70% discount to Western humanoid robot peers at equivalent or earlier production stages: Figure AI at ~$3.2B, Agility Robotics at $1.75B, and 1X Technologies at an implied $3–5B Series C range. SV025, SV026, SV027
CV009 IDC and Mordor Intelligence project the global humanoid robot market to grow from an emerging baseline in 2025 to multi-billion dollar scale by 2030, with manufacturing and logistics as the two highest-priority deployment verticals. Goldman Sachs projects a potential 250,000+ humanoid robot market by 2030 in industrial applications, representing a multi-hundred-billion-dollar TAM. SV009, SV031, SV007
CV010 Agibot made its US market debut at CES 2026 in January 2026, presenting its full humanoid robot portfolio (X2, A2 Lite, A2 Ultra, G2, A3) to international buyers; simultaneously, China's new humanoid robot standards issued in March 2026 are being shaped by Agibot's production leadership, creating a domestic regulatory tailwind. SV015, SV021
CV011 Disclosed venture capital invested in Agibot stands at approximately $83.8M per Tracxn and Humanoid Index, but total capital raised almost certainly exceeds $250M when accounting for undisclosed round sizes in the B, B+, and strategic rounds. The Swancor Advanced Materials acquisition adds a CNY 2.1B (~$290M) capital commitment, making total effective capital deployment approximately $540M+ when combined. SV002, SV003, SV004
CV012 Agibot declared 2026 "Deployment Year One" at its APC 2026 developer conference, signaling a strategic pivot from hardware sales to industrial deployment at scale. IDC data shows that more than 85% of global humanoid deployments in 2025 were in entertainment, education, and guided-tour scenarios — the market transition to manufacturing deployment is still in its earliest phase. SV011, SV012, SV009
CV013 Agibot's post-money valuation following the LG / Mirae Asset round is estimated at approximately $1.0B–$1.3B (CNY 7–9B), making it a confirmed unicorn. Reuters reported the valuation exceeds $1 billion at this round; Pandaily confirmed the pre-money exceeds CNY 7B. SV025, SV028, SV004
CV014 Agibot's total disclosed equity stands at ~$83.8M across pre-Angel through B+ rounds; the QCC registry shows 10+ rounds with many having undisclosed sizes. True total equity raised, excluding the Swancor acquisition, is estimated at $180M–$280M when undisclosed round sizes are modeled against typical China deep-tech round conventions. SV003, SV002, SV004
CV015 Figure AI has raised approximately $675M in a February 2024 round at ~$2.6B post-money valuation, with a subsequent unconfirmed round implying a ~$3.2B valuation; its BMW factory pilot and Helix general-purpose AI model represent the strongest Western competitive parallel to Agibot's industrial ambitions. SV017, SV025, SV026
CV016 Agility Robotics closed a $400M Series C at ~$1.75B post-money valuation in March 2025, supported by the GXO Logistics production deployment (100,000+ totes moved at SPANX facility) and RoboFab 10,000 units/year manufacturing capacity target. SV018, SV025, SV026
CV017 Apptronik raised $350M from Samsung and others in 2024 at an estimated $1.4B–$1.6B valuation (undisclosed); its Apollo humanoid has NASA partnership and automotive sector pilots with similar production-stage proof to Agibot's industrial program. SV022, SV025
CV018 1X Technologies' unconfirmed Series C would imply a post-money valuation of $3B–$5B per analyst estimates; 1X has commercially deployed ~390 EVE units and has EVE enterprise recurring revenue as its primary financial traction signal. SV024, SV025, SV026
CV019 The Boston Dynamics acquisition precedent (Hyundai acquired from SoftBank in 2021 for $1.1B) establishes a strategic M&A floor for pre-revenue advanced robotics companies; adjusted for sector inflation and Agibot's production proof, a strategic floor for Agibot's manufacturing entity (excluding Swancor) is estimated at $1.2B–$1.8B. SV025, SV026, SV027
CV020 Public automation comparables (Teradyne / Universal Robots, Zebra Technologies) trade at 5–8x forward revenue; if Agibot achieves $300M+ in industrial ARR by 2028, a $1.5B–$3B enterprise value at 5–10x is achievable, consistent with the base and bull case scenarios. SV025, SV031, SV009
CV021 Under a base-case assumption of 20,000 cumulative units deployed by end-2027 with 30% in paid industrial contracts at CNY 300,000–500,000 per unit per year, Agibot's annual industrial revenue could reach $250M–$450M — supporting a $1.5B–$4.5B enterprise value at 6–10x multiples. SV007, SV009, SV031
CV022 Agibot's probability-weighted expected value from entry at ~$1.1B post-money is approximately $1.5B–$2.0B at 2028, representing a 1.4–1.8x expected return before dilution from future rounds, assuming 25% bull, 50% base, and 25% bear scenario probabilities. SV007, SV025, SV026
CV023 Dilution from projected future rounds (Series C and beyond) is estimated at 20–40% from current equity holders; if Agibot requires a $500M+ Series C to fund industrial scaling, early-round investor realized returns will be materially lower than implied exit valuations suggest. SV025, SV026, SV003
CV024 The bull case for Agibot assumes 10+ named industrial manufacturing customers by 2027–2028, each with 50–200 deployed G2 units, delivering an implied industrial ARR of $400M–$600M. This scenario is supported by the validated Longcheer proof point and China's State Grid CNY 68B embodied intelligence tender but has not been independently corroborated beyond a single customer. SV011, SV012, SV013
CV025 The base case for Agibot assumes 3–5 named industrial manufacturing customers by end-2027, with 500–800 total deployed industrial units and $80M–$120M in annual hardware sales. This scenario is consistent with China's national industrial humanoid strategy and Agibot's declared "Deployment Year One" pivot, adjusted for the 6–18 month customer acquisition cycle for industrial deployments. SV011, SV012, SV021
CV026 The bear case for Agibot assumes the US Security Robotics Act is enacted, creating a chilling effect on enterprise procurement in US-aligned markets, combined with industrial deployment quality issues at scale and Swancor acquisition-related capital stress. This scenario would impair the company's growth trajectory and likely force a down-round below CNY 5B ($700M) by 2027. SV019, SV030, SV020
CV027 An investor entering at ~$1.1B post-money targeting a 3x return requires an exit valuation of ~$3.3B; this is achievable only under the bull case (G2 scales to 10+ industrial customers, implied EV $3B–$5B). The base case returns 1.2–2.0x, and the bear case returns 0.5–0.7x. SV004, SV025, SV026
CV028 Agibot's 10,000-unit milestone and "Deployment Year One" positioning are widely covered by financial media (Bloomberg, VentureBeat, WSJ) as signaling the industrial humanoid deployment tipping point, reinforcing the investment thesis that Agibot is the earliest commercial-scale industrial humanoid operator globally. SV011, SV027, SV029
CV029 The cap-table structure, liquidation preferences, and anti-dilution provisions for Agibot's Series B and later rounds are not publicly disclosed. Assuming standard Chinese VC liquidation preferences (1x non-participating preferred), late-round institutional investors have downside protection that common stockholders do not. SV025, SV026, SV003
CV030 Goldman Sachs' January 2026 survey of Chinese humanoid robot manufacturers found that major manufacturers (including Agibot) are targeting several times 2025 shipment growth for 2026–2027, driven by the transition from entertainment/service deployments to manufacturing deployments — consistent with Agibot's declared industrial deployment strategy. SV007, SV008, SV009
CV031 Mordor Intelligence and IDC project CAGR of 45–65% for the industrial humanoid robot market through 2030; even at the lower end, a $300M ARR at 5% China market share by 2028 is a plausible base-case assumption given Agibot's current production leadership. SV031, SV009, SV007
CV032 US lawmakers introduced the American Security Robotics Act on March 26, 2026, which would prohibit US government agencies from procuring Chinese-origin humanoid and mobile robots. The bill has bipartisan sponsorship and reflects a political consensus on Chinese robot procurement risk that mirrors the treatment of Huawei in telecom and DJI in enterprise drones. SV019, SV030, SV020
CV033 Hill Dickinson's legal analysis and WSJ coverage identify legal liability frameworks, data sovereignty requirements, and safety certification standards as three fast-moving regulatory dimensions for humanoid robots in 2026 — all creating structural compliance barriers for Chinese-origin robots seeking Western enterprise deployment, independent of legislative outcomes. SV020, SV030, SV019
CV034 China's new humanoid robot standards (issued March 2026) and the Goldman Sachs "dedicated landing" framework both signal that Chinese humanoid manufacturers — led by Agibot on production volume — are positioned to dominate domestic industrial deployment, partially offsetting Western market exclusion risk. SV021, SV008, SV007
CV035 Unitree G1 is commercially available at approximately $16,000 USD, approximately 2–3 orders of magnitude below Agibot's A2 Lite ($44,560) and A2 Ultra (undisclosed B2B price), exerting structural downward price pressure in the developer, research, and lower-tier commercial segments that could limit Agibot's addressable market in price-sensitive verticals. SV023, SV026, SV025
CV036 Figure AI's Helix general-purpose AI model, deployed in Figure 03 for home environments, establishes a competing narrative to Agibot's manufacturing-focused industrial deployment thesis; if Helix demonstrates broad-market capability with manufacturing applicability, it could displace Agibot's first-mover industrial advantage in Western manufacturing contexts. SV017, SV026, SV025
CV037 Agibot's financial disclosures are limited to investor and press communications; no revenue, ARR, gross margin, or burn rate has been publicly disclosed as of May 2026. This opacity prevents independent financial diligence and increases the information asymmetry between insider investors (BYD, Tencent, JD.com) and prospective secondary or new-round investors. SV003, SV004, SV010
CV038 The Swancor Advanced Materials acquisition (CNY 2.1B, ~$290M consortium commitment) is the largest capital event in Agibot's history, representing a vertical integration into upstream high-performance composite materials. The strategic rationale — reducing carbon-fiber and actuator component costs — is plausible but unverified; integration timeline and synergy metrics have not been disclosed. SV003, SV004, SV026
CV039 The geopolitical risk premium embedded in Agibot's ~45% discount to Western humanoid peers is estimated at 20–30 percentage points, reflecting: US Security Robotics Act risk, enterprise procurement hesitation in US-aligned markets, CFIUS-equivalent barriers limiting Western strategic M&A, and reduced IPO optionality in US capital markets. SV019, SV030, SV027
CV040 No IPO timeline, STAR Market registration filing, or overseas listing intent has been publicly disclosed by Agibot as of May 2026. The most probable near-term liquidity pathway for investors is a Series C round at a higher post-money valuation, driven by industrial deployment proof. SV002, SV003, SV004
CV041 Agibot's most probable strategic acquirer pathways are: (1) Chinese state-owned or state-adjacent industrial companies (CATL, BYD, SAIC, State Grid Corporation of China) as part of national embodied intelligence infrastructure buildout; (2) South Korean strategic buyers (LG Electronics, Samsung, Hyundai Motor) bridged through LG's existing investor position; (3) a Japanese or European industrial conglomerate in a non-CFIUS jurisdiction. SV003, SV006, SV025
CV042 The five final diligence asks — industrial ARR confirmation, Swancor integration clarity, US Security Robotics Act legislative status, cap-table preference stack disclosure, and a second named industrial customer — represent the minimum information set required to upgrade from TRACK to a conditional BUY recommendation with medium confidence. SV003, SV025, SV026
来源
编号出版方标题引文
SO001 AGIBOT (official website) AGIBOT Innovation (Shanghai) Technology Co., Ltd. — Homepage
SO002 AGIBOT (official website) About Us — AGIBOT Innovation (Shanghai) Technology Co., Ltd. AGIBOT was established in February 2023 with a founding team of seasoned industry experts, including core executives from global technology leaders and top AI scientists.
SO003 AGIBOT (official website) News — AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SO004 AGIBOT (official website) The First Hong Kong Embodied AI Industry Summit and AGIBOT Partner Conference 2026 Hong Kong Opens
SO005 AGIBOT (official website) How AGIBOT's Seven Solutions Are Reframing the Commercialization of Embodied AI
SO006 AGIBOT (official website) AGIBOT Declares 2026 "Deployment Year One" at APC 2026
SO007 AGIBOT (official website) AGIBOT Unveils New Generation of Embodied AI Robots and Models at APC 2026
SO008 AGIBOT (official website) AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Manufacturing
SO009 AGIBOT (official website) AGIBOT Launches Genie Studio Agent
SO010 AGIBOT (official website) AGIBOT Demonstrates Fully Autonomous Humanoid Playing Table Tennis
SO011 AGIBOT Research AgiBot Research — Robotics and Foundation Models
SO012 AGIBOT Research AgiBot SOP (Standard Operating Procedures) — English
SO013 AGIBOT (official website) AGIBOT A2 Ultra Product Page
SO014 AGIBOT (official website) AGIBOT G1 Product Page
SO015 AGIBOT (official website) AGIBOT X1 Product Page
SO016 AGIBOT (official website) AGIBOT X2 Product Page
SO017 AGIBOT (official website) AGIBOT A2-W Product Page
SO018 AGIBOT Store AGIBOT A2 Ultra — Store Listing AGIBOT A2 Ultra is the industry's first full-sized humanoid robot deployed commercially at scale. With 4 major international certifications (CR, CE-MD, CE-RED, FCC), it has been applied in over 20 leading enterprises.
SO019 AGIBOT Store AGIBOT X2 — Store Listing
SO020 AGIBOT Store AGIBOT A2 Lite — Store Listing
SO021 AGIBOT Genie Studio Genie Studio — Zero-Code Robot Deployment Platform
SO022 OpenDriveLab / AGIBOT (GitHub) AgiBot-World — Full-Stack Large-Scale Robot Learning Platform AgiBot World Colosseo is a full-stack large-scale robot learning platform curated for advancing bimanual manipulation in scalable and intelligent embodied systems.
SO023 AGIBOT (official website) AGIBOT Careers / Recruitment Page
SO024 AGIBOT (PRNewswire) AGIBOT Makes Its US Market Debut at CES 2026 with Its Full Humanoid Robot Portfolio AGIBOT, a leading robotics company specializing in embodied intelligence, today marks its official entry into the U.S. market at CES 2026, showcasing one of the industry's most comprehensive humanoid robot portfolios, backed by the shipment of 5,000 robots to date.
SO025 AGIBOT (PRNewswire) AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line
SO026 Robotics and Automation News Agibot secures strategic investment from LG Electronics and Mirae Asset
SO027 Yicai Global South Korea's LG Electronics, Mirae Asset Co-Invest in Chinese Humanoid Robot Startup AgiBot AgiBot has completed 10 financing rounds, with investment from big firms such as Tencent Holdings, JD.Com, BYD, SAIC Motor, BAIC Group, and TCL Technology Group.
SO028 Reuters / Yahoo Tech Chinese robot maker AgiBot completes new round of strategic financing with LG, Mirae Asset AGIBot, whose robots were inspected by Chinese President Xi Jinping during his visit to Shanghai this year, is one of several Chinese humanoid robot startups that have emerged in recent years.
SO029 SRN News (Reuters wire) Chinese robot maker AgiBot completes new round of financing
SO030 CnTechPost Agibot G2 Achieves Zero Errors, 8-Hour Continuous Factory Operation Agibot's G2 robot completed 2,283 tasks without a single error during an 8-hour continuous operation at Longcheer's tablet factory in Nanchang, Jiangxi.
SO031 Interesting Engineering AGIBOT G2 Humanoid Robots on Live Production Line
SO032 Aparo Robot (industry publication) Deployment Year One: How AGIBOT Is Building a Global Robot Empire In March 2026, a single red-and-white robot rolled off a production line in Shanghai, marking a historic milestone: AgiBot's 10,000th unit.
SO033 Humanoids Daily Deployment Year One — AGIBOT Unveils Massive Fleet and AI Model Stack at APC 2026
SO034 Humanoids Daily The 10,000-Unit Threshold: AGIBOT Accelerates Production in Bid for Global Dominance Celebrating a fundamental shift in scale: AGIBOT's Co-founder and CTO Zhihui Peng (left) and Senior Vice President Chuang Wang (right) commemorate the rollout of the company's 10,000th unit.
SO035 HumanoidIndex AgiBot — Company Profile Omdia ranks AgiBot #1 in humanoid shipments 2025 (>5,100 units).
SO036 Tracxn AgiBot — Company Profile AgiBot is a series B company based in Shanghai (China), founded in 2023 by Deng Taihua and Peng Zhihui. AgiBot has raised $83.8M in funding, with a current valuation of $6.4B.
SO037 Qichacha (企查查) AGIBOT Innovation (Shanghai) Technology Co., Ltd. — Corporate Registry
SO038 Pandaily Agibot Completes New Funding Round, Valuation Exceeds 7 Billion Yuan
SO039 U.S. News / Reuters US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots Two U.S. senators on Thursday planned to introduce a bill that would ban the government from buying or operating humanoid robots made by Chinese firms.
SO040 Robotics and Automation News Why China's New Humanoid Robot Standards Could Change the Industry
SO041 Hill Dickinson (law firm) Humanoid Robots and the Law — Preparing for a New Era of Risk
SO042 Sina Finance (智元机器人) 智元机器人亮相 MWC 2026,西班牙国王费利佩六世亲临接见
SM001 Goldman Sachs Research Humanoid robot: The AI accelerant "the global market for humanoid robots may reach a market size of at least US$6bn in 10-15 years, filling 4% of the US manufacturing labor shortage gap by 2030E and 2% of global elderly care demand by 2035E... a market of up to US$154bn by 2035E in a blue-sky scenario"
SM002 MarketsandMarkets Humanoid Robot Market — Global, China, and US Forecasts to 2030 "The global humanoid robot market is expected to grow from USD 2.92 billion in 2025 to USD 15.26 billion in 2030, at a compound annual growth rate (CAGR) of 39.2% during the forecast period."
SM003 IDC From Task Execution to Value Creation: What the 2026 Humanoid Robot Half Marathon Reveals About Industry Progress "According to IDC, global shipments of humanoid robots are expected to exceed 510,000 units by 2030, representing a compound annual growth rate (CAGR) of nearly 95%."
SM004 Taibo Finance (citing Goldman Sachs Research) Goldman Sachs' research on humanoid robots in China: The industry is shifting from Universal Imagination to Dedicated Landing "Goldman Sachs expects global shipments of humanoid robots to reach approximately 15000 to 20000 units by 2025, with Chinese companies contributing the majority of shipments."
SM005 Grand View Research Humanoid Robot Market Size, Share & Trends Analysis Report
SM006 Forbes Humanoid Robots: Here Are The 16 Leading Manufacturers "Goldman Sachs says selling humanoid robots will be a $38 billion space by 2035, while Ark Invest says the resulting economic value of their labor could be as high as $24 trillion."
SM007 International Federation of Robotics (IFR) China Makes AI-powered Robots Core of National Strategy — 15th Five-Year Plan (2026–2030) "China's manufacturing industry already has an operational stock of around 2 million units — approximately 4.5 times more than the global no. 2, Japan. 54% of annual industrial robots installed worldwide were deployed in China."
SM008 IEEE Spectrum Humanoid Robots Are Getting to Work
SM009 PR Newswire (Agibot) AGIBOT Makes Its U.S. Market Debut at CES 2026 with Its Full Humanoid Robot Portfolio "AGIBOT's showcase at CES highlights its accomplished transition: from advanced R&D to a full portfolio of mass-produced humanoids already deployed at scale in real-world operations."
SM010 PR Newswire (Agibot) AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line "2026 marks the beginning of large-scale deployment for embodied intelligence. This project demonstrates that embodied AI is no longer experimental."
SM011 CnTechPost Agibot G2 achieves zero errors in 8-hour continuous factory operation "The Agibot G2 robot completed 2,283 tasks without a single error during an 8-hour continuous operation at Longcheer's tablet factory in Nanchang, Jiangxi."
SM012 Interesting Engineering AGIBOT puts humanoid robots to work in real factory production
SM013 Aparobot Year One of Deployment: How AgiBot is Building a Global Robot Empire "While it took the company nearly two years to produce its first 1,000 units, and another year to reach 5,000, the jump from 5,000 to 10,000 units was completed in a mere three months."
SM014 Humanoids Daily Deployment Year One: AGIBOT Unveils Massive Fleet and AI Model Stack at APC 2026 "AGIBOT CEO Edward Deng formally declared 2026 as 'Deployment Year One.'"
SM015 Humanoids Daily The 10,000-Unit Threshold: AGIBOT Accelerates Production in Bid for Global Dominance "AGIBOT was crowned the global leader in shipments by IDC, reporting 5,200 units delivered by the end of 2025."
SM016 US News (Reuters) US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots "Two U.S. senators planned to introduce a bill that would ban the government from buying or operating humanoid robots made by Chinese firms."
SM017 Robotics and Automation News Why China's new humanoid robot standards could change the industry "China's Ministry of Industry and Information Technology (MIIT) published its first national standard system for humanoid robots and embodied intelligence in late February 2026."
SM018 Agibot (official) How AGIBOT's Seven Solutions Are Reframing the Commercialization of Embodied AI "At Longcheer Technology, robots handle up to 5,700 units per day in continuous 24/7 operation. At HT-Tech and Intel, deployments process up to one million and 100,000 chips per day respectively."
SM019 Agibot (official) AGIBOT Declares 2026 'Deployment Year One' at APC 2026, Accelerating the Era of Embodied AI Productivity
SM020 Agibot (official) AGIBOT Unveils New Generation of Embodied AI Robots and Models at APC 2026
SM021 Agibot (official) AGIBOT and Longcheer Technology — Official Deployment Announcement
SM022 Humanoid Index AgiBot: Funding, Valuation, Robot Specs and More "Omdia ranks AgiBot #1 in humanoid shipments 2025 (>5,100 units)"
SM023 Hill Dickinson Humanoid Robots and the Law: Preparing for a New Era
SM024 Pandaily Agibot Completes New Funding Round, Valuation Exceeds $7 Billion
SM025 Yicai Global South Korea's LG Electronics, Mirae Asset Co-invest in Agibot's Strategic Investment Round
SP001 AGIBOT AGIBOT Official Website
SP002 AGIBOT via PR Newswire AGIBOT Makes Its U.S. Market Debut at CES 2026 with Its Full Humanoid Robot Portfolio
SP003 AGIBOT AGIBOT Declares 2026 Deployment Year One at APC 2026
SP004 AGIBOT AGIBOT Unveils New Generation of Embodied AI Robots and Models
SP005 Humanoids Daily The 10,000-Unit Threshold: AGIBOT Accelerates Production in Bid for Global Dominance
SP006 Humanoid Index AgiBot: Funding, Valuation, Robot Specs & More
SP007 Robotics and Automation News Why China's new humanoid robot standards could change the industry
SP008 US News (Reuters) US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots Two U.S. senators planned to introduce a bill that would ban the government from buying or operating humanoid robots made by Chinese firms. At least two Chinese firms — Agibot and Unitree — are preparing to list shares in China this year as their products capture attention.
SP009 QCC (Chinese Corporate Registry) 智元创新(上海)科技股份有限公司 — Corporate Filing
SP010 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, with a blue-sky scenario of up to US$154bn by 2035E.
SP011 Taibo Financial News Goldman Sachs research on humanoid robots in China: industry shifting to dedicated deployment
SP012 Forbes Humanoid Robots: Here Are The 16 Leading Manufacturers
SP013 Figure AI Figure — General Purpose Humanoid Robot
SP014 1X Technologies 1X — Home Robots
SP015 1X Technologies 1X Secures $100M in Series B Funding
SP016 TechCrunch OpenAI-backed 1X raises another $100M for the race to humanoid robots
SP017 Agility Robotics Humanoid Solutions — Agility Robotics
SP018 Apptronik Apollo — Apptronik General Purpose Humanoid Robot
SP019 Unitree Robotics Humanoid Robot G1 — Unitree Robotics
SP020 Unitree Robotics Unitree Robotics — Advanced Quadruped and Humanoid Robots
SP021 Humanoid Index 1X Technologies: Funding, Valuation, Robot Specs & More
SP022 Humanoid Index Agility Robotics: Funding, Valuation, Robot Specs & More
SP023 Humanoid Index Apptronik: Funding, Valuation, Robot Specs & More
SP024 Humanoid Index Figure AI: Funding, Valuation, Robot Specs & More
SP025 The Robot Report Unitree G1 Humanoid Robot: Price, Specs, Analysis
SP026 The Robot Report Humanoid Robot Landscape 2025: Market Map
SI001 AGIBOT Store (official) AGIBOT A2 Lite — Product Listing and FAQ
SI002 AGIBOT Store (official) AGIBOT X2 — Product Listing and FAQ
SI003 AGIBOT Store (official) AGIBOT A2 Ultra — Product Listing and FAQ
SI004 AGIBOT Genie Studio (official) Genie Studio — AI Development Platform
SI005 AGIBOT (official press release) AGIBOT Declares 2026 Deployment Year One at APC 2026
SI006 AGIBOT (official press release) AGIBOT Unveils New Generation of Embodied AI Robots and Models at APC 2026
SI007 AGIBOT (official website) AGIBOT — Business Cooperation Contact Page
SI008 Humanoids Daily The 10,000-Unit Threshold: AGIBOT Accelerates Production in Bid for Global Dominance
SI009 Humanoids Daily Deployment Year One: AGIBOT Unveils Massive Fleet and AI Model Stack at APC 2026
SI010 AGIBOT (PRNewswire press release) AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line
SI011 CnTechPost Agibot G2 Achieves Zero Errors in 8-Hour Continuous Factory Operation
SI012 Interesting Engineering AGIBOT G2 Humanoid Robots Live on Production Line
SI013 QCC / Qichacha (Chinese corporate registry) AGIBOT Innovation (Shanghai) Technology Co., Ltd. — Shareholder and Corporate Registry
SI014 Robotics and Automation News Agibot Secures Strategic Investment from LG Electronics and Mirae Asset
SI015 Yicai Global South Korea's LG Electronics and Mirae Asset Co-Invest in Chinese Humanoid Robot Startup AgiBot
SI016 HumanoidIndex AgiBot Company Profile — HumanoidIndex
SI017 Tracxn Agibot Company Profile — Tracxn
SI018 SRN News (Reuters wire) Chinese Robot Maker AgiBot Completes New Round of Financing
SI019 Pandaily Agibot Completes New Funding Round, Valuation Exceeds 7 Billion Yuan
SI020 US News / Reuters US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots
SI021 Goldman Sachs Research Humanoid Robot: The AI Accelerant — Global Automation
SI022 IDC From Task Execution to Value Creation: What the 2026 Humanoid Robot Half Marathon Reveals About Industry Progress
SI023 AGIBOT (official press release) AGIBOT AI Week's Five Breakthroughs: From Technical Leaps to the Industry Inflection Point
SI024 AGIBOT (official press release) The Unity of Reasoning and Action: AGIBOT Unveils Genie Operator-2 (GO-2) Next-Gen Embodied Foundation Model
SI025 AGIBOT (official press release) AGIBOT Introduces Genie Sim 3.0, an Integrated Simulation, Data, and Benchmarking Platform
SI026 AGIBOT (official press release) How AGIBOT's Seven Solutions Are Reframing the Commercialization of Embodied AI
SI027 AGIBOT (official press release) AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment — Company Version
SI028 AGIBOT (official press release) AGIBOT Unveils Genie Envisioner 2.0, Advancing World Models into Scalable World Simulators for Embodied AI
SI029 AGIBOT (official press release) AGIBOT Open-Sources AGIBOT WORLD 2026 Dataset to Accelerate Embodied AI Development
SI030 AGIBOT (official press release) AGIBOT Reaches 10,000 Units — Scaling Milestone Marks Transition from Validation to Deployment
SI031 AGIBOT (official press release) AGIBOT Hosts Robotics Showcase Event in Southern California
SE001 AGIBOT (official website) AGIBOT Innovation (Shanghai) Technology Co., Ltd. — Homepage
SE002 AGIBOT (official website) AGIBOT A2 Ultra — Product Page AGIBOT A2 Ultra has obtained four major international certifications: CR, CE-MD, CE-RED, and FCC.
SE003 AGIBOT (official website) AGIBOT A2-W — Product Page
SE004 AGIBOT (official website) AGIBOT X1 — Product Page
SE005 AGIBOT (official website) AGIBOT X2 — Product Page
SE006 AGIBOT (official website) AGIBOT G1 — Product Page
SE007 AGIBOT (official website) AGIBOT G2 — Product Page
SE008 AGIBOT (official documentation) AGIBOT X1 Open-Source Hardware and Deployment Guide The body contains 29 joints (R86-2*9, R86-3*6, R52*10, L28*4) and 2 grippers, which support the expansion of the head with 3 degrees of freedom.
SE009 AGIBOT (official website) AgiBot Research — Research Overview
SE010 AGIBOT Research (official) SOP: Scaling General-Purpose Robots in the Real World After SOP training, our robots operated autonomously on target tasks for over 36 hours without requiring human intervention.
SE011 AGIBOT (Genie platform) Genie Studio — Platform Overview Calibrated simulation environments achieve < 5% error between Go-1 model test results and real-device outcomes.
SE012 OpenDriveLab / AGIBOT (GitHub) GitHub: OpenDriveLab/AgiBot-World — IROS 2025 Best Paper Award Finalist AgiBot World Beta: Our complete dataset featuring 1,003,672 trajectories (~43.8T)
SE013 AgiBot World (HuggingFace) agibot-world Organisation — HuggingFace
SE014 AimRT (open-source project) AimRT — Release Notes and Homepage (v1.7.0) AimRT v1.7.0 Release! Added the callback_executor_name option to mqtt_plugin…
SE015 Agibot Store (official) AGIBOT A2 Ultra — Store Listing
SE016 Agibot Store (official) AGIBOT A2 Lite — Store Listing ($44,560)
SE017 Agibot Store (official) AGIBOT X2 — Store Listing
SE018 Agibot Store (official) AGIBOT X2 Ultra — Store Listing ($24,240 base) AGIBOT X2 Ultra possesses powerful capabilities in motion, interaction, and task intelligence.
SE019 AGIBOT PTE. LTD. (official) AGIBOT Website Privacy Policy (Effective April 30, 2026)
SE020 AGIBOT (official) Product Security Update Support — Defined Support Period V4 We will maintain the security updates for at least 1 years from the launch day of certain device models.
SE021 AGIBOT (official news) AGIBOT Launches Genie Studio Agent to Enable Scalable Robot Deployment
SE022 AGIBOT (official news) AGIBOT Unveils New Generation of Embodied AI Robots and Models at APC 2026
SE023 AGIBOT (official news) AGIBOT Demonstrates Fully Autonomous Humanoid Robot Playing Table Tennis in Live Rally
SE024 PR Newswire AGIBOT Makes Its U.S. Market Debut at CES 2026 with Its Full Humanoid Robot Portfolio
SE025 PR Newswire AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line Throughput: Up to 310 units per hour (UPH). Success Rate: Over 99.9% in continuous operation. Deployment Time: Production line integration completed within 36 hours.
SE026 CnTechPost Agibot G2 achieves zero errors in 8-hour continuous factory operation The Agibot G2 robot completed 2,283 tasks without a single error during an 8-hour continuous operation at Longcheer's tablet factory in Nanchang, Jiangxi.
SE027 Interesting Engineering AGIBOT puts humanoid robots to work in real factory production
SE028 Aparobot Year One of Deployment: How AgiBot is Building a Global Robot Empire
SE029 HumanoidsDaily Deployment Year One: AGIBOT Unveils Massive Fleet and AI Model Stack at APC 2026 AGIBOT formalized a five-level capability framework for embodied AI. The company claims its current systems have achieved Level 3 (L3) across locomotion and task intelligence.
SE030 Humanoid Index AgiBot — Funding, Valuation, Robot Specs & More
SE031 Robotics and Automation News Why China's new humanoid robot standards could change the industry Peng Zhihui, co-founder of Agibot and a deputy director of the HEIS committee, noted that in industrial scenarios, 'nearly 80 percent of tasks where humans excel but traditional automation struggles are strongly related to tactile sensing' — and the lack of standardized tactile sensors remains a critical bottleneck.
SE032 AimRT (open-source project) AimRT — Open-Source Robotics Middleware (GitHub Repository) AimRT is a high-performance, asynchronous, coroutine-based communication and computation framework for robotic applications.
SU001 Longcheer Technology Longcheer Technology Corporate Website Longcheer Technology describes itself as an AI-enabled smart manufacturing partner and highlights co-deployment of embodied AI on precision consumer electronics production lines.
SU002 AGIBOT Store AGIBOT Store — Robot-as-a-Service and Direct Purchase Portal C5 cleaning robot listed at $32,900; RaaS model available from store.agibot.com.
SU003 AGIBOT APC 2026 Partner Conference Recap — Global Partners, Embodied AI Productivity Era
SU004 AGIBOT APC 2026 — Industry Leaders and Partner Testimonials
SU005 AGIBOT APC 2026 — Global Customer Quick Q&A International customers from multiple countries provided one-word descriptions and video testimonials at the APC 2026 partner conference, demonstrating global engagement.
SU006 IoT World Today (Informa TechTarget) Agibot and Longcheer Humanoid Robot Factory Deployment
SU007 New Straits Times Agibot and Genting Malaysia Sign MOU to Deploy Humanoid Robots New Straits Times reported on AGIBOT and Genting Malaysia signing an MOU to deploy humanoid robots across hospitality and theme park operations at Genting Highlands.
SU008 Tech in Asia Agibot Deployment Year One 2026
SU009 Genting Malaysia Berhad Genting Malaysia — Corporate Website (Resorts World Genting)
SU010 PR Newswire AGIBOT and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass Production Line "The G2 robots achieved a throughput of 310 UPH with a motion success rate of 99.9%, operating 16 hours a day in a continuous multi-shift environment with 8 hours fully autonomous. Integration was completed in 36 hours; full production ramp took 4 months."
SU011 Aparobot Year One of Deployment: How AGIBOT Is Building a Global Robot Empire Lists Longcheer, Genting Malaysia, Hyperscale Data, SIR Spa (Italy), Singtel (Singapore), BYD, I-Berhad, and Whale Cloud as commercial partners or deployments in AGIBOT's Deployment Year One.
SU012 Interesting Engineering AGIBOT G2 Humanoid Robots Live on Production Line — Longcheer Deployment
SU013 Humanoids Daily Deployment Year One: AGIBOT Unveils Massive Fleet and AI Model Stack at APC 2026
SU014 Humanoids Daily The 10,000-Unit Threshold: AGIBOT Accelerates Production in Bid for Global Dominance
SU015 Sina Finance AGIBOT MWC 2026 — RaaS Launch, €899/Day, 17 Countries AGIBOT launched RaaS at MWC 2026 at €899 per robot per day, covering 17 countries and regions, with direct consumer/enterprise ordering via store.agibot.com.
SU016 AGIBOT AGIBOT APC 2026 — 7 Standardized Productivity Solutions Across Industries
SU017 AGIBOT AGIBOT Deployment Launch — Longcheer Production Announcement
SU018 US News & World Report US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots Senators Cotton and Schumer introduced the American Security Robotics Act to prohibit US federal agencies from purchasing or operating robots manufactured by Chinese companies, including AGIBOT.
SU019 Hill Dickinson Humanoid Robots and the Law: Preparing for a New Era of Risk Hill Dickinson identifies product liability gaps, operator health and safety obligations, and unclear accountability frameworks as material risk factors for enterprises deploying humanoid robots in commercial settings.
SU020 Goldman Sachs Global Automation: Humanoid Robot — The AI Accelerant
SU021 IDC Humanoid Robotics Commercialization 2026 — IDC Blog
SU022 Humanoid Index AGIBOT Company Profile — Humanoid Index Humanoid Index, citing Omdia data, ranks AGIBOT first globally by shipped unit volume in 2025 with approximately 5,100 units.
SU023 Robotics and Automation News Why China's New Humanoid Robot Standards Could Change the Industry
SU024 AGIBOT AGIBOT News — Official News and Announcements
SU025 AGIBOT AGIBOT Contact — Business Inquiry Portal
SR001 U.S. News & World Report / Reuters US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots
SR002 Robotics & Automation News Why China's new humanoid robot standards could change the industry
SR003 Hill Dickinson LLP Humanoid robots and the law — preparing for a new era of risk
SR004 Agibot AGIBOT Website Privacy Policy
SR005 Agibot Product Security Update Support — Defined Support Period V4
SR006 Agibot AGIBOT Declares 2026 'Deployment Year One' at APC 2026
SR007 Agibot AGIBOT Unveils New Generation of Embodied AI Robots and Models at APC 2026
SR008 Agibot How AGIBOT's Seven Solutions Are Reframing the Commercialization of Embodied AI
SR009 PR Newswire Agibot and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass Production Line
SR010 CnTechPost Agibot G2 Achieves Zero Errors 8-Hour Continuous Factory Operation
SR011 Interesting Engineering AGIBOT Deploys G2 Humanoid Robots into Live Consumer Electronics Manufacturing Line
SR012 Humanoids Daily The 10,000-Unit Threshold: Agibot Accelerates Production in Bid for Global Dominance
SR013 Humanoids Daily Deployment Year One: Agibot Unveils Massive Fleet and AI Model Stack at APC 2026
SR014 Pandaily Agibot Completes New Funding Round, Valuation Exceeds 7 Billion Yuan
SR015 Robotics & Automation News Agibot Secures Strategic Investment from LG Electronics and Mirae Asset
SR016 Yahoo Tech / Reuters Chinese Robot Maker Agibot Completes New Round of Financing
SR017 Qichacha Agibot Corporate Registry — Qichacha
SR018 Humanoid Index AgiBot — Humanoid Index
SR019 Goldman Sachs Global Automation: Humanoid Robot — The AI Accelerant
SR020 IDC Humanoid Robotics Commercialization — 2026
SR021 Taibo / Goldman Sachs summary Goldman Sachs Humanoid Robot Industry Tracking — Wind Chasing Platform
SR022 Agibot AGIBOT Table Tennis Demo — Ping-Pong Diplomacy Anniversary Event
SR023 OSHA Robotics Safety — US Occupational Safety and Health Administration
SR024 Bureau of Industry and Security Export Administration Regulations — US Bureau of Industry and Security
SR025 ISO ISO 10218-1:2011 — Robots and Robotic Devices — Safety Requirements for Industrial Robots
SR026 China Securities Regulatory Commission China Securities Regulatory Commission — Official Website
SR027 U.S. Treasury / CFIUS Committee on Foreign Investment in the United States (CFIUS)
SR028 International Federation of Robotics China's 15th Five-Year Plan Pivots to Robotics Innovation — IFR
SR029 EU AI Act official site EU AI Act — AI Act Explorer and Compliance Resource
SR030 European Parliament EU AI Act — European Parliament Explainer
SV001 Humanoid Index (Omdia) AgiBot: Funding, Valuation, Robot Specs & More
SV002 Tracxn Agibot — Company Profile, Funding, Investors
SV003 QCC / Qichacha (Chinese Corporate Registry) 智元创新(上海)科技股份有限公司 — Corporate Registration and Shareholder Filing
SV004 Pandaily Agibot Completes New Funding Round, Valuation Exceeds 7 Billion Yuan
SV005 Robotics and Automation News Agibot Secures Strategic Investment from LG Electronics and Mirae Asset
SV006 Yicai Global South Korea's LG Electronics, Mirae Asset Co-Invest in Chinese Humanoid Robot Startup AgiBot
SV007 Goldman Sachs Research Global Automation: Humanoid Robot — The AI Accelerant
SV008 Taibo Financial News Goldman Sachs: China Humanoid Robot Industry Transitioning to Dedicated Landing Applications 2026
SV009 IDC Humanoid Robotics Commercialization 2026
SV010 Tech Yahoo (Reuters) Chinese Robot Maker Agibot Completes New Round of Financing
SV011 Humanoids Daily Deployment Year One: Agibot Unveils Massive Fleet and AI Model Stack at APC 2026
SV012 Humanoids Daily The 10,000-Unit Threshold: Agibot Accelerates Production in Bid for Global Dominance
SV013 CnTechPost Agibot G2 Achieves Zero Errors, 8-Hour Continuous Factory Operation
SV014 Interesting Engineering Agibot G2 Humanoid Robots Live Production Line
SV015 PR Newswire (Agibot official) Agibot Makes Its US Market Debut at CES 2026 With Its Full Humanoid Robot Portfolio
SV016 PR Newswire (Agibot official) Agibot and Longcheer Technology Achieve World's First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass Production Line
SV017 Figure AI Figure AI — Homepage (Figure 03 General Purpose Humanoid)
SV018 Agility Robotics Agility Robotics — Solutions (Digit RaaS for Logistics)
SV019 US News / Reuters US Lawmakers to Introduce Bill to Ban Government Use of Chinese Robots
SV020 Hill Dickinson Humanoid Robots and the Law: Preparing for a New Era of Risk
SV021 Robotics and Automation News Why China's New Humanoid Robot Standards Could Change the Industry
SV022 Apptronik Apollo — Humanoid Robot for Industrial Applications
SV023 Unitree Robotics Unitree G1 — Humanoid Robot Specifications and Pricing
SV024 1X Technologies 1X Technologies — EVE and NEO Humanoid Robots
SV025 PitchBook Humanoid Robot Startup Funding and Valuation Landscape 2025–2026
SV026 CB Insights State of Robotics 2026: Investment, Valuation, and Commercialization Trends
SV027 Bloomberg Chinese Humanoid Robot Startups Race to Industrial Deployment in 2026
SV028 Reuters Agibot Valuation Exceeds $1 Billion After LG, Mirae Asset Round
SV029 VentureBeat Why Agibot's 10,000-Robot Milestone Matters for Humanoid AI Investment in 2026
SV030 The Wall Street Journal Chinese Robotics Firms Attract Global Capital Despite US Restrictions 2026
SV031 Mordor Intelligence Humanoid Robot Market Size, Share, and Growth Analysis 2025–2030
SV032 BusinessWire (Longcheer Technology) Longcheer Technology and Agibot Announce Industrial Humanoid Deployment Milestone