Galaxy Bot
工业部署可信、融资规模可观,但经营经济性、客户广度、治理和估值支撑仍不透明。
Galaxy Bot 已在工业机器人领域站稳脚跟,也获得了有力的资本支持;但运营、合同、安全和证券权利数据缺失,尚不足以评估其传闻估值。
覆盖范围与披露说明
公开来源审阅截至 2026-10-21。现有证据未保留私人财务报表、资本化记录、客户合同、车队遥测数据以及正式安全与合规记录。
封面要素
公司概况
Galaxy Bot 于 2023 年由 He Wang 创立,是一家以 Galbot 品牌运营的中国私人实体 AI 机器人公司。其公开产品路线将任务专用的视觉—语言—动作模型与移动操作硬件结合,已披露的商业化最强证据是在 CATL 生产线上部署 S1。公开报道显示公司融资规模可观、估值超过 $3 billion,但未能核对累计融资总额,也未披露收入、利润率、现金消耗、客户集中度、车队可靠性或融资权利。
- 创始人
- He Wang
- 总部
- Beijing / Shenzhen (conflicting public descriptions)
- 产品
- Galbot 开发实体 AI 模型和移动操作机器人系统,面向受限的零售、导航、抓取、物料搬运和拣选任务。
- 客户
- 大型工业制造商是证据最充分的近期客户群;现有证据中,零售仍属于产品定位和试点场景。
- 商业模式
- 机器人硬件、部署、集成和售后服务都可能带来收入,但 Galbot 的合同结构、定价和收入组合尚未公开披露。
- 阶段
- Private
- 融资情况
- 2026 年 3 月据报道完成 RMB2.5 billion 融资,但公开来源对累计融资额的描述相互冲突,无法核对。
执行摘要
主要优势
- CATL 提供了具有实质意义的产线部署案例。
- 公司吸引了规模可观的战略投资和机构融资。
- 其产品逻辑把面向具体任务的具身 AI 模型与已部署的移动操作硬件结合起来。
主要风险
- 收入、毛利率、现金消耗、现金 runway 和单位经济模型均未公开披露。
- 商业验证集中在一家已点名的生产客户上,而该客户同时也是战略投资者。
- 车队可靠性、人工干预负担、安全认证和事故记录仍未得到验证。
- 依赖外国算力和精密零部件,带来地缘政治与供应链风险。
- 传闻估值没有公开披露的运营分母或融资权利支撑。
未决问题
- 经审计收入、毛利率、现金消耗、现金余额、债务和现金 runway
- 完全摊薄后的股权结构表、清算优先权、投资者权利和资金用途
- 已签署的客户合同、付费设备数量、客户集中度、续约情况和部署 ROI
- 车队运行时间、人工干预率、故障历史、现场服务成本和保修负担
- 产品安全认证、事故记录、隐私控制、保险和诉讼状态
- 当前高管团队、董事会构成、员工人数和继任安排
目录
01公司概览
1.1 身份与运营模式
Galaxy Bot 成立于 2023 年,公开品牌为 Galbot。现有证据支持其采取覆盖基础模型和机器人硬件的实体 AI 路线,服务零售和工业任务,而非纯软件业务。Galbot 官网强调由高层推理、底层控制和神经控制组成的一体化模型;独立的 MERICS 分析则将其 GraspVLA、TrackVLA 和 GroceryVLA 描述为职责范围较窄的目标系统。Multiples 将公司列为私人持有,并把总部写为 Beijing,但其自有介绍又称公司总部在 Shenzhen。因此,Beijing 应视为工作地点,而不是已经完全核对的法定总部事实。公司在 2026 年 9 月宣布新型原生智能体机器人的预订,说明产品仍在推进,但不能据此确认出货量、定价或收入。[CO001, CO002, CO003, CO004, CO005]
| 指标 | 数值或状态 | 截至 | 置信度 | 缺口 |
|---|---|---|---|---|
| 成立时间 | 2023 | 2026 年证据审阅 | 中 | 现有证据池没有独立备案的法定成立日期 |
| 总部 | 数据库字段为 Beijing;同一页面叙述为 Shenzhen | 当前页面访问于 2026-10-21 | 低 | 核实注册办公地址和运营总部 |
| 所有权与阶段 | 私人持有;融资阶段标注不一致 | 当前页面访问于 2026-10-21 | 中 | 股权结构表和证券类别不可得 |
| 产品与商业模式 | 实体 AI 模型和服务机器人;数据库将收入模式归类为产品销售 | 2026 | 中 | 软件、服务和硬件收入组合不可得 |
| 估值 | 2025 年 12 月报道为 $3B;CNBC 后称超过 $3B | 2025-12 / 2026-04 | 高 | 条款清单、投后股份数和优先权不可得 |
| 累计融资 | 未核对:$619M 标题数字,对比 2026 年 3 月轮次前叙述中的 $800M | 2026-03 | 低 | 需要逐轮一手文件及货币换算口径 |
| 商业覆盖 | CATL 产线部署及据报道的北京 10 家零售试点;无客户数 | 2026-06 / 2025-07 | 中 | 付费单元、合同金额、活跃地点和集中度不可得 |
| 收入与员工数 | 指定证据未披露 | 2026-10-21 | 低 | 需要审计收入、毛利率、现金消耗、员工规模和服务人员配置 |
数值保留来源期间、货币、归属和冲突;null 或不可得的经营指标不作估计,融资额也不等于可用现金。
[CO001, CO002, CO003, CO004, CO008, CO009, CO010, CO011, CO012, CO013]1.2 领导层与治理
独立报道一致将 He Wang 认定为创始人,并将其与北京大学联系起来;Robotics & Automation News 还称其接受过 Stanford 训练。这一组合显示其在机器人学习和实体 AI 领域具备可信的创始人与市场匹配度,但指定证据未披露更完整的现任高管团队、董事会构成、独立董事、投票结构或继任安排。因此,尽调应关注明显的关键人集中:公开叙事聚焦一名技术创始人,职能领导层却不可见。这并不证明其他领导人或治理机制不存在,而是说明证据存在边界;在评估治理质量前,需要当前组织架构、董事名册、股权结构表和关键人连续性方案。[CO001, CO006]
1.3 融资形成与利益相关方集中度
融资记录显示资本积累迅速,但内部及跨来源存在不一致,无法得出可辩护的累计融资额。Robotics & Automation News 将 2025 年 6 月融资报为 $151 million,Multiples 报为 $153 million;差异可能来自换算或四舍五入,但这里保留差异而未统一。Multiples 报道 2025 年 12 月以 $3 billion 估值完成 $300 million 轮融资,CNBC 随后称 Galbot 估值超过 $3 billion。2026 年 3 月据报道融资 RMB2.5 billion,美元换算约为 $350 million 至 $368 million。更重要的是,Multiples 在同一页面既称五轮合计 $619 million,又在其他位置称 2025 年 12 月、即 3 月融资前已达 $800 million。因此,融资额不应视为可用现金,也不能根据投资人参与情况推断所有权或控制权。[CO007, CO008, CO009, CO010, CO011]
| 利益相关方 | 角色 | 控制权或经济重要性 | 证据边界 | 尽调要求 |
|---|---|---|---|---|
| CATL | 战略投资人、客户和部署伙伴 | 运营上重要的参考客户及 2025 年 6 月轮次领投方 | 无持股比例或合同经济性 | 获取股权结构表持股、订单金额、关联方条款和客户集中度 |
| Puquan Capital | 2025 年 6 月轮次联合领投方 | 已披露融资影响;尚未确立控制权 | 参与融资不等于拥有投票权 | 核查证券类别、董事会权利和保护性条款 |
| Bosch China / Boyuan Capital | 工业商业化伙伴及合资交易对手方 | 可能撬动制造和渠道资源 | 合资所有权及经济安排未披露 | 核查合资协议、排他性、知识产权所有权和市场推进义务 |
| China Mobile Chain Builder Fund | 据 Multiples 报道,为 2025 年 12 月轮次领投方 | 在该轮融资中担任领投方;尚未确立控制权 | 未披露持股或治理条款 | 确认投资金额、所有权、董事会权利和战略义务 |
| 2026 年 3 月国家及产业联合体 | 据报道,投资者包括国家、能源、汽车、银行及地方政府资本 | 广泛的政策和产业支持 | 配额、证券条款及附函不可得 | 核对认购、受益所有人、优先权和政策相关契约 |
保留来源中提及的利益相关方样本,不是股权结构表;投资者参与不代表拥有控制权,经济重要性也不等于持股比例。
[CO007, CO008, CO009, CO012, CO013]1.4 里程碑与商业化证据
公司层面最强的部署证据,是 CATL 据报道将 Galbot S1 放入模组和电池包生产,用于长期自主物料搬运和拣选。更早报道还提到北京十家无人便利店试点及覆盖一百家门店的计划,但该计划不应视为已实现部署。这些信号有意义,但行业证据不支持从试点外推到广泛商业成熟。MERICS 称中国人形机器人仍受精度、灵巧性、任务范围狭窄和成本限制;Associated Press 报道则指出需求可能落后于制造能力。作为上市公司基准而非 Galaxy Bot 指标,UBTECH 报告 1,079 台全尺寸实体智能机器人及相关 2025 年收入 RMB820.6 million。指定证据中,Galaxy Bot 未披露可比的审计单元数、收入、利润率、事故或留存序列。[CO012, CO013, CO014, CO015, CO016]
带日期的公开里程碑显示融资和部署活动推进迅速,同时保留来源报道的金额差异,并区分规划与已实现运营。
没有推断日期。仅有月份或仅有年份的条目反映来源粒度;美元金额保留报道中的换算,不强行采用单一汇率。
[CO001, CO005, CO007, CO008, CO009, CO012, CO013]1.5 展示项
02市场分析
2.1 市场边界与替代方案
可投资边界并不是整个机器人或人工智能市场。Galaxy Bot 近期相关市场包括随部署和支持销售的实体 AI 人形机器人及移动操作系统,服务受限的工业、物流、零售、医疗和研究流程。传统固定式工业机器人、机械臂、自动导引车和人工劳动仍是主要替代方案。这一区分很重要:2025 年中国生产的传统工业机器人远多于人形机器人,许多工厂无需人形外观也已实现重复性工作的自动化。家庭消费照护应视为相邻市场,而不是当前可服务市场,因为现有证据仍描述了昂贵、脆弱且依赖结构化环境的系统。[CM001, CM002, CM004, CM010, CM014]
| 细分市场或类别 | 纳入的支出 | 排除的支出或替代方案 | 买方或付款方 | 与 Galaxy Bot 的相关性 |
|---|---|---|---|---|
| 工业移动操作 | 机器人硬件、具身 AI 软件、集成、服务和机群支持 | 固定机械臂、输送机、AGV 及人工搬运 | 工厂自动化、运营和制造工程团队 | 短期相关性最高,因为工作流边界清晰且重复性强 |
| 仓储与物流自动化 | 拣选、分拣、搬运、编排和维护 | AMR、叉车、输送系统和仓库劳动力 | 仓库运营和供应链负责人 | 移动操作能在现有自动化之外创造价值时才相关 |
| 零售、酒店及医疗服务 | 门店运营、补货、引导和受限服务工作流 | 自助机、专用服务机器人和一线劳动力 | 运营、设施及业务单元负责人 | 邻近机会,但需要工作流级 ROI 和安全证据 |
| 研究与教育 | 机器人平台、数据采集、软件访问和支持 | 模拟器、实验室设备和专用研究机器人 | 实验室、大学和公共研究项目 | 适合早期采用者,但不能证明运营需求已规模化 |
| 家庭通用辅助 | 未来消费级机器人及支持支出 | 家电、家政服务和人工照护者 | 家庭消费者 | 因成本和可靠性限制,排除在短期可服务市场之外 |
该边界由分析师根据保留的 2025-2026 年市场、申报文件和部署证据界定;各类别不可相加, 且没有出版方提供 Galaxy Bot 的细分市场收入。
[CM001, CM002, CM010, CM011, CM012, CM014]2.2 受证据约束的市场规模
现有证据不足以支持清晰的 TAM/SAM/SOM 计算。不同出版方和时间范围混用了产量、出货量、公司销售额、收入和长期预测。因此,目前最有用的锚点是采用速度:MERICS 报告 2025 年中国生产了 12,800 台人形机器人,Associated Press 报道全球出货超过 13,000 台,其中约 85% 来自中国。Morgan Stanley 对 2026 年中国约 28,000 台的预测显示增长很快,但其引用的 $5 trillion 市场估计在抓取文本中没有时间范围。UBTECH 经审计的 1,079 台和 2025 年收入 RMB820.6 million 证明商业收入可以有实质规模,但不能据此确定 Galaxy Bot 可获得的份额。[CM003, CM004, CM005, CM006, CM007, CM008, CM009, CM015]
| 出版方及期间 | 地理区域或公司 | 指标 | 数值 | 置信度 | 局限 |
|---|---|---|---|---|---|
| MERICS,2025 | 中国 | 人形机器人产量 | 12,800 台;约占全球总量的 90% | 中 | 产量不等于终端客户采用量或收入 |
| Associated Press 引述 Barclays,2025 | 全球及中国 | 人形机器人出货量及中国占比 | 全球超过 13,000 台;中国约占 85% | 中 | 二手报道,且口径不同于产量 |
| Associated Press 引述 Morgan Stanley,2026 年预测 | 中国 | 人形机器人销量 | 约 28,000 台 | 中 | 预测值,不是实际销量;公司构成不可得 |
| Associated Press 引述 Omdia,2030 年代初 | 全球 | 高级机器人年出货量 | 超过 1 百万台 | 低 | 高级机器人定义更宽,预测期也更长 |
| UBTECH 申报文件,截至 2025 年 | UBTECH | 全尺寸具身人形机器人销量及收入 | 1,079 台及 RMB820.6 million | 高 | 竞争对手基准;产品组合和经济性不能直接套用到 Galaxy Bot |
数值保留各出版方的指标、期间、地理范围和分母;不会刻意换算成单一 TAM/SAM/SOM,也不会求和。
[CM003, CM005, CM006, CM007, CM009]六项不可加总的指标显示中国市场快速采用,同时也凸显可服务市场和可获取市场存在较大不确定性。
各层宽度仅表示层级,不代表数值比例。
这些条目的单位、定义、地理范围和时间跨度各不相同。条目顺序只是证据层级;宽度不能理解为比例、嵌套关系或可加总数值。
[CM003, CM005, CM006, CM007, CM008, CM015]2.3 买方、用户与采用路径
工业买方最容易从试点走向持续部署。工厂自动化和运营负责人可以为物料搬运、拣选、分拣、检查及其他重复或危险任务采购系统,产线工人和维护团队则是日常用户。CATL 据报道使用 Galbot S1 进行产线搬运和拣选,是具体的工作流信号;TechNode 报道 Galbot 声称累计工业订单达数千笔。物流、研究、零售和医疗扩大了机会空间,但预算负责人和验收标准各不相同。因此,每笔销售都需要现场集成、安全验证、服务覆盖,以及相对于人工或成熟自动化方案的 ROI 论证,而不能只依赖对人形机器人的笼统兴趣。[CM010, CM011, CM012, CM014]
| 细分市场 | 买方及预算负责人 | 主要用户 | 工作流 | 采用触发因素 | 主要尽调测试 |
|---|---|---|---|---|---|
| 电池及汽车制造 | 工厂自动化和运营负责人 | 产线操作员及维护团队 | 物料搬运、拣选、分拣和检查 | 劳动密集度高、安全暴露,或需要灵活自动化 | 对比固定自动化,测量吞吐量、正常运行时间、安全性和回本周期 |
| 仓储与物流 | 配送及供应链运营团队 | 仓库员工及机群主管 | 拣选、包裹分拣、搬运和装载 | 可变布局中的重复性或危险工作 | 检验高度结构化区域以外的集成成本和性能 |
| 研究与教育 | 实验室、大学或公共项目负责人 | 研究人员和开发者 | 数据采集、模型训练和平台实验 | 获得具身 AI 硬件及数据集 | 研究需求能否转化为重复性的商业部署 |
| 零售、酒店及医疗 | 业务单元运营和设施团队 | 员工及服务接受者 | 补货、引导、接待和受限辅助 | 用工压力或差异化客户体验 | 安全、空间限制、利用率和服务负担 |
| 家庭照护 | 消费者 | 家庭成员及照护者 | 清洁、整理和辅助 | 劳动力短缺或人口老龄化需求 | 成本、灵巧性、可靠性和无人监管下的安全 |
买方图谱受证据约束,仅作方向性判断;没有保留来源披露 Galaxy Bot 的采购周期、买方预算、 部署定价或实际客户 ROI。
[CM010, CM011, CM012, CM014]2.4 驱动因素、约束与估值意义
中国密集的机器人和电动车供应链、政策支持及庞大的工厂基础,应该会加快零部件国产化、训练数据收集和部署学习。MERICS 估计,2024 年中国企业占全球人形机器人关键零部件供应链主要企业的 63%,并供应了中国工业机器人市场的 57%。制约因素在经济性和技术性两端:MERICS 引用中国人形机器人典型价格 CNY300,000-500,000,而商业可行性门槛为 CNY160,000;独立报道则描述了有限用例、脆弱运行和需求落后于制造能力。依赖 Nvidia 的训练和边缘计算生态,也会增加地缘政治及供应风险。因此,估值应奖励已验证的部署和任务级经济性,而不能只奖励宏观市场预测。[CM013, CM014, CM015, CM016]
2.5 展示项
03竞争对手
3.1 竞争拥挤,竞争形式不止一种
买方的选择集不只包括人形机器人初创公司。中国直接同业包括 UBTECH、Unitree、AgiBot、AI2 Robotics、Matrix Robotics 和 EngineAI;Figure、Apptronik 等美国进入者则争夺资本、软件人才和跨国工厂项目。固定机械臂、输送线和专用自动化仍是最强的既有替代方案,因为许多工厂已用它们处理重复任务。内部整合第三方机器人本体与软件,也是硬件变得更便宜、更模块化后的可行替代方案。行业高度分散:Associated Press 报道 2025 年中国有超过 140 家人形机器人制造商和超过 330 个型号。出货领先者在别处,AgiBot 和 Unitree 据报道 2025 年各自超过 5,000 台,而 Figure 和 Tesla 仅出货几百台或更少。Galbot 据报道有 $1.1 billion 已披露融资和超过 $3 billion 估值,具备持续投入的资本实力,但这两个指标都不能证明相当的商业规模。[CP001, CP002, CP003, CP004, CP016]
| 竞争对手 | 类别 | 公开规模或融资信号 | 目标 / 产品方向 | 差异化 | 局限 |
|---|---|---|---|---|---|
| Galbot | 直接竞争:具身 AI 移动操作 | 已披露融资 $1.1B;据报道估值 $3B+ | 零售及工业拣选 / 物料搬运 | 智能优先的 VLA 模型;可接入 CATL 生产线 | 出货量、收入、价格和机群性能未披露 |
| UBTECH | 直接竞争:工业人形机器人 | 2025 年人形机器人收入 RMB820.6M;年化产能 >6,000 台 | 汽车、电子、物流和智能制造 | Walker S2 量产及多机器人工业工作流 | 尽管收入增长,2025 年仍据报道净亏损 RMB789.8M |
| Unitree | 直接 / 邻近竞争:低成本人形机器人 | 2025 年出货 >5,000 台;据报道 2025 年收入约 $250M | 广泛的硬件平台及开发者 / 商业用途 | 可见的出货规模和较低入门价格 | 公开来源无法证明广泛的自主任务性能 |
| AgiBot | 直接竞争:中国具身 AI 人形机器人 | 2025 年出货 >5,000 台 | 训练、物流、制造和具身 AI 开发 | 出货领先及专用数据设施 | 公开来源缺少经审计的收入或单位经济性 |
| Figure | 直接竞争:美国通用人形机器人 | 据报道估值 $39B+;2025 年出货数百台或更少 | 工业部署和通用 AI 平台 | Helix VLA 软件叙事 | 相比已披露出货量,估值偏高 |
| Apptronik | 直接竞争:美国工业人形机器人 | 2026 年 2 月据报道估值 $5B | Apollo 工业人形机器人项目 | 美国生态接入和工业定位 | 所分配证据未披露出货量、价格或收入 |
规模信号采用不同期间和分母,不能直接比较;融资、估值、收入、产能和出货量作为不同指标保留, 未披露字段标记为缺口。
[CP001, CP003, CP004, CP005, CP010, CP011, CP012]3.2 软件野心构成 Galbot 的差异化,但同业拥有更可见的规模
Galbot 最清晰的差异化,是面向自主移动操作的智能优先架构。MERICS 识别出独立的 GraspVLA、TrackVLA 和 GroceryVLA 模型,并指出 Galbot 偏好合成训练数据。这种专业化对零售拣选和导航有意义,但也说明当前自主能力仍局限于受限任务。UBTECH 提供了最强的已披露工业商业化证据:其申报文件披露 Walker S2 量产、截至 2025 年末年化全尺寸产能超过 6,000 台,以及人形机器人收入 RMB820.6 million。Unitree 和 AgiBot 在报道出货量上领先,Figure 则以高得多的估值维持软件平台叙事。因此,这张图区分的是可见部署规模和自主化取向,是有证据支撑的序数比较,而不是性能基准。[CP003, CP005, CP010, CP011, CP012]
| 采购标准 | Galbot | UBTECH | Unitree | AgiBot | 固定自动化 / 内部开发 |
|---|---|---|---|---|---|
| 可见的工业部署 | CATL S1 产线部署 | Walker S2 量产及工业交付 | >5,000 台 2025 年总出货(据报道);工业占比未知 | >5,000 台 2025 年总出货(据报道);工业占比未知 | 在重复性、结构化任务中成熟 |
| 自主化路径 | GraspVLA、TrackVLA、GroceryVLA;偏好合成数据 | Thinker-VLA 和多机器人协作 | 基于仿真的训练;本文未确立任务性能 | GO-1 视觉语言模型;专用数据设施 | 确定性编程或集成商专用软件 |
| 负载 / 运营 | S1:双臂负载 50 kg;物料搬运和拣选 | Walker S2:最多搬运 15 kg,并可自主更换电池 | 所分配证据中未知 | 所分配证据中未知 | 专用负载和循环时间 |
| 分销 / 服务 | 拟使用 CATL Ning Service 的售后网络 | 直接工业客户及标准参与 | 所分配证据中未知 | 所分配证据中未知 | 成熟的自动化集成商和工厂维护团队 |
| 信任 / 合规 | 受中国生命周期数字身份制度约束 | 申报文件据称具备标准领导力;受数字身份制度约束 | 受数字身份制度约束 | 受数字身份制度约束 | 成熟的安全流程因集成商和安装方式而异 |
单元格仅概括截至 2026-10-21 的所分配来源;“未知”表示该来源集中没有支持,并不表示能力不存在。
[CP005, CP012, CP013, CP014, CP016]有证据支持的序数定位,将报道的出货或生产规模与对自主具身 AI 软件的侧重区分开来。
| 条目 | 可见部署规模(1=披露规模有限,5=报道中的领先规模) (1–5) | 自主性取向(1=硬件主导,5=智能优先) (1–5) | 说明 |
|---|---|---|---|
| 1. Galbot | 3 | 5 | CATL 部署支持中等规模定位;具名 VLA 模型支持智能优先定位。 |
| 2. UBTECH | 4 | 4 | 大规模生产、已披露收入和 Thinker-VLA 支持其在两个轴上都取得高分。 |
| 3. Unitree | 5 | 2 | 出货领先和低成本硬件支撑规模;现有证据对自主作业的说明较少。 |
| 4. AgiBot | 5 | 3 | 出货领先和专用数据设施支撑规模,同时软件取向较为均衡。 |
| 5. Figure | 2 | 5 | 据报道 2025 年出货量较低,但 Helix VLA 平台取向较强。 |
| 6. 固定自动化 | 5 | 1 | 已广泛用于重复性工厂任务,但定位并非通用具身自主系统。 |
分数是基于引用证据的分析师序数判断,不是实测表现、市场份额或基准测试结果;相差一分不代表量化差距。
[CP003, CP005, CP006, CP007, CP012, CP016]3.3 价格压力已显现,但 Galbot 商业条款仍不透明
公开证据没有披露 Galbot 的标价、实际成交价、合同期限、服务费或折扣。现有市场锚点仍指向加剧的价格压力:MERICS 将中国工业人形机器人平均价格置于 RMB300,000-RMB500,000,并引用 CNY160,000 的商业可行性门槛,而 Unitree G1 明显低于该均值。Galbot 可能通过任务表现和渠道抵消价格压力。其与 CATL 的关系使 S1 进入电池生产线,并可能接入 CATL 售后服务网络,这或许构成重要的部署和支持优势。信任要求也正从可选项变成结构性要求:中国新的机器人识别制度实行无代码、无市场准入规则及全生命周期可追溯。这会提高合规成本,但可能偏向制造和服务伙伴成熟的供应商。[CP008, CP009, CP013, CP014]
| 厂商 / 基准 | 公开价格或合同信号 | 单位 / 型号 | 包含的能力 | 折扣 / 未知项 | 竞争含义 |
|---|---|---|---|---|---|
| Galbot | 未披露 | S1 / G1 商业条款未知 | 零售及工业移动操作 | 合同期限、服务、折扣和实际价格未知 | 无法确立性价比优势 |
| Unitree | 约 EUR11,650 | G1 硬件 | 低成本人形机器人平台 | 配置和服务范围尚未确立 | 形成低硬件价格锚点 |
| 中国行业均价 | RMB300,000-RMB500,000 | 工业人形机器人均价 | 因厂商和配置而异 | 均价不是任何具体型号的报价 | 当前成本仍高于大规模采用门槛 |
| 商业可行性门槛 | RMB160,000 | 分析师计算的门槛 | 基于 RMB80,000 年度人工成本和两年 ROI | 说明性假设,不是交易价格 | 刻画买方可能要求的降本幅度 |
| Matrix Robotics | 约 $99,000 | MATRIX-3 | 配备精细控制手部、身高 1.7m 的人形机器人 | 服务和批量折扣未披露 | 表明高价专业化竞争对手仍在活跃 |
| EngineAI | RMB180,000 ($26,600) | 基础版 | 安保及博物馆导览用例 | 服务和企业条款未披露 | 在计入集成成本前,接近所引述的可行性门槛 |
价格混合了不同币种、配置和来源口径,并排除了未知的集成、服务、折扣和生命周期成本; 它们是方向性的市场锚点,不是标准化报价。
[CP008, CP009, CP015]3.4 护城河取决于尚未公开的部署数据和经济性
Galbot 可能从零售和工业部署中建立学习闭环,但公开证据没有量化车队规模、任务成功率、运行时间、人工干预率或客户留存。中国供应商已控制大量零部件生态和国内机器人市场,降低了硬件稀缺性作为持久护城河的价值。同时,依赖 Nvidia 机器人技术栈和进口高端零部件,限制了全栈控制。独立报道对行业经济性持不利判断:MERICS 称当前人形机器人仍是场景专用且过于昂贵,Associated Press 则报道买方可以继续使用固定式自动化,且许多订单来自国有企业。因此,Galbot 的防御力取决于证明更高的任务级 ROI,并比规模化同业更快积累专有部署数据,而不是融资、估值或单一锚定合作伙伴。[CP006, CP007, CP015, CP016]
3.5 展示项
04财务
4.1 商业验证领先于财务披露
最强的公开商业信号是运营而非财务:CnEVPost 报道 Galbot S1 已进入 CATL 生产线,执行物料搬运和拣选工作。同一报道称双方计划把 CATL 售后体系延伸至实体智能机器人,说明部署、维护和全生命周期支持都可能影响经济模型。但这两点都不能确认 Galbot 是销售硬件、出租产能、收取集成费、获得经常性服务费,还是只在验收后确认收入。指定证据也没有合同金额、实际单价、客户集中度、收入、ARR 或毛利率。因而,CATL 只能证明产品被使用,不能证明收入有实质规模或单位经济性有吸引力。投资判断应把工厂里摆放的机器人,与已付费、已验收、可收款且支持负担明确的部署区分开。[CI001, CI002]
| 收入流 | 机制 | 单位 | 当前数值或状态 | 质量 | 尽调要求 |
|---|---|---|---|---|---|
| 工业机器人硬件 | 销售、租赁或捆绑合同未披露 | 按机器人或部署计 | S1 正在 CATL 运行;未披露 Galbot 合同金额 | 仅有部署证明 | 已签合同、发票、实际单价和验收日期 |
| 部署与集成 | 可能收取项目费或实施费;机制未披露 | 按站点或生产线计 | 未披露费用或收入贡献 | 财务上缺乏支持 | 工作说明书、实施人力和确认收入 |
| 售后服务 | 可能产生维护或服务收入;机制未披露 | 按机器人或服务期计 | CATL 与 Galbot 计划扩展售后体系 | 前向运营信号 | 服务水平条款、附加率、续约率和现场成本 |
| 软件及模型访问 | 许可、订阅或捆绑访问未披露 | 按机器人或账户计 | 证据缺口 | 价目表、合同、递延收入和续约证据 | |
| 数据或生态服务 | 变现机制未披露 | 按数据集或合作伙伴计 | 证据缺口 | 第三方账单及变现客户数据的权利 |
Galbot 数值仅限于已披露的部署状态;null 表示所分配的预取证据中没有出现数量或变现条款。
[CI001, CI002]已观察到的部署处于一条财务链条的起点,而商业条款和成本仍未披露。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 客户任务 [customer-task] | 已观察到 CATL 的物料搬运和拣选任务。 | |
| 节点 2 | 商业合同 [contract] | 销售、租赁、服务、数量和价格条款均不可得。 | |
| 节点 3 | 交付和验收 [delivery] | 验收时间可能决定机器人收入何时确认。 | |
| 节点 4 | 集成和售后 [support] | CATL 和 Galbot 计划建立售后体系;费用和成本未知。 | |
| 节点 5 | 已确认收入 [revenue] | Galbot 的收入金额、构成和回款时间均未披露。 | |
| 节点 6 | 毛利 [gross-profit] | 硬件、安装、保修和现场服务成本均不可得。 |
这是证据状态流程图,而非预测;任何缺失的商业环节都没有被赋值。
[CI001, CI002, CI016]4.2 行业价格意味着更高的 ROI 门槛
公开价格证据适用于行业,但不适用于 Galbot。Associated Press 报道 Matrix Robotics MATRIX-3 约为 USD 99,000,基础款 EngineAI 人形机器人为 CNY 180,000(USD 26,600)。MERICS 将中国人形机器人平均价格置于 CNY 300,000-500,000,并以年度人工成本 CNY 80,000 和两年 ROI 周期为基础,引用 CNY 160,000 的商业可行性门槛。这些数字对应不同产品、配置和证据时点,因此只能定义价格区间,不能代表 Galbot 价格。MERICS 和 Morgan Stanley 都预计广泛采用前会大幅降价,这将挤压利润:Galbot 必须至少以与市场价格下降同样快的速度降低物料、安装、可靠性和支持成本。软件 AI 使用模式明确不能作为这种硬件加服务成本结构的代理。[CI007, CI008, CI009, CI010, CI011]
| 产品或基准 | 价格或单位 | 标价与实际成交 | 折扣或未知项 | 投资判断含义 |
|---|---|---|---|---|
| Galbot S1 | 未披露 | 未知 | 合同、机群折扣、集成和支持条款不可得 | 无法计算实际 ASP 或贡献毛利 |
| Matrix Robotics MATRIX-3 | 每台约 99,000 美元 | 据报道的要价 | 配置和实际折扣未知 | 高单价不能证明需求或利润率 |
| EngineAI 基础款人形机器人 | CNY 180,000(USD 26,600) | 据报道的基础版价格 | 能力和服务组合不同 | 入门价显示出品类分化 |
| 中国人形机器人均价 | CNY 300,000-500,000 | 分析师市场区间 | 产品组合和规格各异 | 没有报价,无法判断 Galbot 是否落在该区间 |
| 商业可行性阈值 | CNY 160,000 | 分析师推导的阈值 | 假设年人工成本为 CNY 80,000,投资回收期为两年 | 客户 ROI 取决于利用率、可靠性和服务成本 |
| 长期高收入国家成本路径 | 2024 年 USD 200,000;2028 年 USD 150,000;2050 年 USD 50,000 | Morgan Stanley 预测 | 不是 Galbot 的价格,高度依赖预测 | 价格下压会提高降本门槛 |
同业价格和分析师区间不是 Galbot 的实际成交价;币种、配置、日期和合同包含项均有差异。
[CI007, CI008, CI009, CI010]4.3 累计融资无法解决现金和消耗问题
CnEVPost 报道 2026 年 3 月融资 CNY 2.5 billion(USD 368 million),CNBC 报道估值超过 USD 3 billion。这些是融资和估值观察,不是当前可用现金、经营现金消耗、债务义务或现金跑道的衡量。成本基础可能包括机器人硬件、模型训练、现场集成、维护和营运资金,但指定证据未披露金额。UBTECH 提供了有用的上市公司边界:其 2025 年人形机器人收入增长强劲、毛利率提高,但仍亏损且经营活动使用现金。UBTECH 还披露产品收入按某一时点确认,未交付或未验收项目会递延收入,说明交付验收和回款时点是核心尽调问题。这些同业数据不能估算 Galbot 的消耗或跑道。[CI003, CI004, CI012, CI013, CI014, CI015, CI016]
| 指标 | Galbot 数值 | 置信度 | 公开边界或信号 | 尽调需索取 |
|---|---|---|---|---|
| 收入 | 无法获得 | 已披露运营部署,但没有 Galbot 收入金额 | 按产品、服务、客户和地区划分的经审计月度收入 | |
| 毛利率 | 无法获得 | UBTECH 报告 2025 年集团毛利率为 37.7%;不是 Galbot 估算 | 硬件、软件、集成、服务和保修毛利率 | |
| 经营现金消耗 | 无法获得 | UBTECH 2025 年经营活动使用 RMB 784.1 million;不能作为 Galbot 代理 | 月度现金流量表和正常化经营现金消耗 | |
| CAC 与回收期 | 无法获得 | 没有 Galbot 销售周期、CAC 或回收期证据 | 管道客户队列、赢单率、销售成本、部署时间和回收期 | |
| 营运资金 | 无法获得 | 同业申报文件显示,验收时间可能推迟机器人收入确认 | 库存、应收账款账龄、预付款、里程碑和收款记录 | |
| 现金与 runway | 无法获得 | 据报道融资 CNY 2.5 billion;融资所得不等于账上现金证据 | 非受限现金、债务、已承诺资本开支、现金消耗和下行情景 runway |
Galbot 的所有 null 均是明确的证据缺口;上市同业数据只能提供边界条件,不能视为 Galbot 估算。
[CI003, CI012, CI013, CI014, CI015, CI016]4.4 财务结论与投资判断阻碍
Galbot 具备可信的部署和融资信号,但现有证据不足以判断收入质量、利润率或资本充足性。行业证据偏不利:买方仍有限,中国已报道的需求很大部分来自国有企业,独立分析师称当前机器人广泛部署的成本过高。因此,投资逻辑取决于把参考部署转化为可重复的付费单元,同时在价格下降时控制安装和售后人工。在作出投资判断前,投资人需要客户级开票和回款、合同与验收条款、硬件及服务毛利率、保修和现场支持成本、库存及应收账款账龄、月度经营现金使用、非受限现金、债务、已承诺资本开支以及下一轮融资触发条件。在这些记录交付前,财务画像应视为高度不确定且依赖融资,而不是已证明的规模经济。[CI001, CI002, CI003, CI005, CI006, CI008, CI009, CI010, CI012, CI013, CI014, CI015, CI016]
4.5 展示项
05产品与技术
5.1 产品范围与客户工作流
Galbot 的公开产品更适合被理解为两个相关但尚未证明通用的运营边界,而不是一台已验证的通用机器人。MERICS 通过三个狭窄的视觉—语言—动作模型描述 G1:GraspVLA 负责手部操作,TrackVLA 负责导航,GroceryVLA 面向零售。这种拆分支持智能优先路线,也表明自主能力仍局限于受限工作。S1 则提供更重型的工业能力:CnEVPost 报道其具备 50-kilogram 双臂负载,并在 CATL 电池生产线上执行物料搬运和拣选。这些来源证明了明确的工作流和已部署硬件,但没有公开任务成功率、干预频率、循环时间分布或客户 ROI。因此,产品尽调应把现有验证视为任务专用部署证据,而不是证明一个模型能可靠跨越零售、制造和医疗场景。[CE001, CE002, CE006, CE009]
| 模块或资产 | 买方或用户 | 公开成熟度信号 | 差异化 | 尽调缺口 |
|---|---|---|---|---|
| G1 平台 | 零售和移动操作运营商 | MERICS 描述的三个任务专用模型 | 以智能优先的移动抓取与放置为核心的判断 | 没有机群规模、正常运行时间、人工干预率或发布历史 |
| S1 平台 | 电池制造运营商 | 据报道在 CATL 生产线上常态化运行 | 用于物料搬运和拣选的 50 kg 双臂负载能力 | 没有独立的任务成功率、循环时间或故障数据 |
| GraspVLA | 灵巧手任务开发者 | 用途较窄的具名模型 | 面向机器人灵巧手的专用策略 | 没有基准、版本、延迟或泛化披露 |
| TrackVLA | 机器人导航开发者 | 用途较窄的具名模型 | 专用导航策略 | 没有地图、障碍物、定位或恢复基准 |
| GroceryVLA | 零售运营商 | 面向零售场景的具名模型 | 场景专用零售策略 | 没有门店级准确率、安全性或劳动生产率证据 |
| DexGraspNet | 模型训练团队 | MERICS 报道的模拟抓取数据集 | 合成灵巧手训练数据 | 没有当前数据集规模、许可、偏差或现实迁移审计 |
成熟度标签仅反映留存的 2026 年报道;null 性能指标仍是尽调缺口,不能假定为零。
[CE001, CE002, CE006, CE007]| 用户任务 | 当前工作流 | Galbot 方案 | 可量化公开信号 | 局限 |
|---|---|---|---|---|
| 搬运生产物料 | 人工或固定式自动化搬运 | S1 自主物料搬运 | 据报道双臂负载能力为 50 kg | 没有吞吐量或工时分母 |
| 在电池生产中拣选零件 | 模组和电池包产线上的人工拣选 | S1 在 CATL 产线上拣选 | 据报道已长期自主作业 | 没有成功率或干预频率 |
| 保持机器人能源可用 | 充电和电池服务 | CATL 供能的 S1 | 八小时续航说法归因于 CATL | 没有工作循环、负载或衰减条件 |
| 提供现场支持 | 检查、维护和回收服务 | 计划将 Ning Service 扩展至机器人 | 已提及 CATL 现有服务网络 | 该计划不能证明已交付的 SLA 或服务经济性 |
CATL 指标是 CnEVPost 报道的合作伙伴归因部署说法;不是独立可靠性测试或正常化 ROI 指标。
[CE001, CE002, CE004, CE005]5.2 架构、数据与差异化
可见架构结合了任务级 VLA 策略、合成抓取数据、Nvidia 仿真和边缘计算工具,以及移动操作平台。MERICS 报道 Galbot 的 DexGraspNet 是使用 Nvidia Isaac Sim 构建的模拟抓取数据集,Galbot 也是较早获得 Jetson Thor 模块的公司之一。这为灵巧操作带来可能的数据和迭代优势,同时也在仿真和计算两层形成外部依赖。架构仍只能部分观察:现有来源未披露模型评估协议、数据权利控制、端侧与云端边界、故障回退、车队编排或版本发布管理。因此,架构图是有证据支撑的尽调模型,并不声称所列层级复现了 Galbot 的内部实现。[CE003, CE006, CE007, CE008]
| 层 | 有证据支持的角色 | 依赖 | 主要风险 |
|---|---|---|---|
| 任务策略 | GraspVLA、TrackVLA 和 GroceryVLA 分别负责抓取、导航和零售任务 | 场景专用模型训练 | 窄域策略可能无法迁移到未见过的工作流 |
| 训练数据 | DexGraspNet 提供模拟抓取样本 | Nvidia Isaac Sim | 仿真到现实的性能未披露 |
| 边缘计算 | Jetson Thor 是已识别的机器人计算输入 | Nvidia 硬件和软件生态 | 依赖供应、出口管制和产品路线图 |
| 感知与运动 | 据报道 S1 采用纯视觉定位和避障 | 摄像头、标定和控制栈未披露 | 没有已发布的降级模式或故障安全证据 |
| 能源 | 电池系统支持一项归因于合作方的八小时运行说法 | CATL 电池技术 | 测试条件和衰减曲线无法获得 |
| 保障 | 标准为可靠性和信任提供框架 | 认证和合规评估证据 | 留存材料中没有 Galbot 专属证书或正式测试结果 |
这是受来源边界约束的尽调拆解,不是已披露的内部架构;未披露的实施细节有意保留为依赖或风险。
[CE003, CE004, CE006, CE007, CE008, CE011, CE012]公开证据支持从工作流、任务策略、合成数据、计算资源、物理平台到现场服务的分层视图,但关键内部接口未披露。
- 客户工作流
- 模块
- 零售拣放
- 电池线物料搬运
- 电池线拣选
- 任务策略
- 模块
- GraspVLA
- TrackVLA
- GroceryVLA
- 训练数据和仿真
- 模块
- DexGraspNet
- Nvidia Isaac Sim
- 计算与自主性
- 模块
- Jetson Thor
- 仅视觉定位
- 避障
- 物理平台
- 模块
- G1 移动操作
- S1 重载操作
- 部署与支持
- 模块
- CATL 生产线
- CATL 电池系统
- Ning Service 扩展计划
分层顺序是对公开证据的分析综合,不代表 Galbot 已披露的实现、接口规格或完整物料清单。
[CE001, CE003, CE004, CE006, CE007, CE008, CE011]5.3 部署成熟度、可靠性与支持
CATL 是证据池中最强的已点名部署信号。CnEVPost 报道 S1 在模组和电池包生产中长期自主作业,并转述 CATL 关于八小时电池续航的说法;双方还计划把 CATL Ning Service 网络从电池检查和维护扩展到实体智能机器人。这说明部署设计考虑了能源和现场服务,而不只是机器人能力。但这不能证明运行时间、平均故障间隔、远程协助负担、备件可得性,或服务经济性能够复制到 CATL 之外。UBTECH 2025 年申报文件提供了有用的竞争对照:其将量产与预防性设计、制造工艺控制、模块测试和整机测试联系起来。现有证据未披露 Galbot 有可比控制,因此技术成熟度应按已证明的部署可见度保守评估。[CE002, CE004, CE005, CE014, CE015, CE016]
5.4 信任、安全、标准与尽调缺口
安全和质量保证是证据最薄弱的一层。ISO 解释标准为可靠性和信任设定一致基准,但现有来源没有 Galbot 特定的证书、合规评估、正式安全测试、状态历史或事故登记。这是证据缺口,不是事故或控制不存在的证据。MERICS 还提醒,中国人形机器人整体仍受精度和灵巧性限制,通常部署在受约束、特定场地;这属于行业基准,不是对 Galbot 的直接失败结论。指定来源没有提供代码库活动、版本发布、包下载、API 文档或可信的开发者社区代理,因此无法评分开发者采用度和软件运营节奏。尽调应在赋予生产级成熟度前,索取认证范围、危险分析、故障安全行为、渗透测试摘要、隐私和训练数据治理、现场事故、运行时间、干预日志及版本化路线图。[CE009, CE010, CE011, CE012, CE013]
5.5 展示项
06客户
6.1 客户基础:工业验证,更广定位
Galbot 最明确的客户群是大型工业制造。CATL 已将 S1 放到智能产线执行模组和电池包搬运及拣选,使双方关系超越 logo 或演示。现有材料中,零售仍是产品定位而非经核实的客户群:MERICS 将 GroceryVLA 描述为零售场景模型,但没有点名 Galbot 的零售客户。行业证据还指出研究实验室、国有企业、物流运营商、工厂和商场是活跃的人形机器人买方或部署地点,但不能把这些观察归入 Galbot 自有客户。因而,这一细分是按证据层级划分,而不是公司声称的客户普查。[CU001, CU002, CU007, CU010, CU014]
| 客户群 | 买方 / 用户 / 付款方 | 已观察用例 | 证据级别 | 战略价值 | 尚未解决的缺口 |
|---|---|---|---|---|---|
| 电池制造 | CATL 运营和生产团队 | 模组和电池包物料搬运与拣选 | 已具名的 Galbot 生产部署 | 锚定工业客户和集成环境 | 设备数、合同金额、站点数和 ROI 未披露 |
| 零售与杂货 | 零售商运营团队;门店员工是用户 | GroceryVLA 面向零售场景设计 | 仅有产品定位 | 重复性操作和门店自动化的潜在切入口 | 没有具名的 Galbot 零售客户或生产结果 |
| 工业与物流 | 制造商、物流运营商和国有企业 | 搬运、分拣、装配、检查和包裹中心工作 | 品类及同业证据 | 最可能的近期最大买方池 | Galbot 专属客户数和付费设备量未披露 |
| 研究与训练 | 企业和学术实验室 | 机器人研究、训练和数据采集 | 品类证据 | 早期需求和产品学习渠道 | 没有 Galbot 专属买方数量或转化路径 |
客户群将一项具名 Galbot 部署与明确标注的品类及同业证据结合;null 商业指标不作估算。
[CU002, CU007, CU010, CU014]6.2 采用验证与轨迹
CATL 是生产采用最清晰的标志:CnEVPost 报道 S1 执行长期自主工作,而不是一次性试点。更广泛的需求环境可信但不均衡。2023 年中国每万名制造业员工拥有 470 台在运行的工业机器人,说明买方具备自动化经验;2026 年报道则称中国人形机器人供应商已向工厂和商场出货。同业披露显示商业化可以产生实质财务影响——UBTECH 报告 1,079 台全尺寸机器人及相关 2025 年收入 RMB820.6 million——但这些同业数据不能证明 Galbot 的出货量、收入或重复使用。指定证据中没有 Galbot 客户数或装机基础序列。[CU002, CU009, CU010, CU011, CU012, CU013]
| 指标 | 数值 | 期间 | 范围 / 来源 | 置信度 | 含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| Galbot 具名生产案例 | 1 (CATL) | 2026 | 已审阅的指定证据 | 中 | 具备生产可信度,但证据集中 | Galbot 客户、站点和设备总数 |
| 中国制造业机器人密度 | 每 10,000 名员工对应 470 台 | 2023 | IFR 工业机器人安装基数 | 中 | 中国工厂是有经验的自动化买方 | 人形机器人占比和 Galbot 份额 |
| 中国人形机器人订单 | 超过 RMB2 billion | 2025 | Associated Press 引述 Morgan Stanley | 中 | 品类存在需求,但很大程度上由国家支持 | 付费交付、复购订单和厂商拆分 |
| UBTECH 售出的全尺寸人形机器人数量 | 1,079 | 2025 | Yicai 关于 UBTECH 年度业绩的报道 | 中 | 一家国内同业达到了可量化的商业规模 | 可比的 Galbot 设备数量和定价 |
| UBTECH 全尺寸人形机器人收入 | RMB820.6 million | 2025 | 港交所年度业绩公告 | 中 | 同业商业化已达到财务重要性 | Galbot 收入和客户贡献 |
该表将 Galbot 证据与明确标注的市场及同业基准混列;同业数值不是 Galbot 表现的代理。
[CU002, CU007, CU009, CU011, CU012]| 客户 / 参考对象 | 公司 | 客户群 | 部署或用例 | 阶段 | 据报道结果 | 局限 |
|---|---|---|---|---|---|---|
| CATL | Galbot | 电池制造 | 模组和电池包生产中的 S1 物料搬运与拣选 | 产线常态化运行 | 据报道已长期自主作业 | 没有设备数、合同金额、ROI 或续约数据 |
| 未具名的海外高端制造商 | AI2 Robotics(同业) | 工业制造 | 工厂作业 | 厂商报告的选型结果 | 据报道选择了 AI2,而非一家美国竞争对手 | 客户未具名,结果未量化 |
| 咖啡连锁店和酒店 | Matrix Robotics(同业) | 酒店服务与零售 | 通用人形机器人部署 | 据报道已获订单 | 约 1,000 笔订单中的一部分 | 客户未具名;交付和结果未披露 |
指定证据池中可归因于具体对象的生产或订单证据样本;只有 CATL 与 Galbot 直接相关,同业行展示了披露标准及其局限。
[CU002, CU008, CU010, CU011, CU013]6.3 持久性与扩张
CATL 关系包含一个可信的扩张机制:双方计划把实体智能机器人纳入 CATL 既有的 Ning Service 检查、维护和回收网络。这可能降低现场支持摩擦并推动跨生产地点扩张,但证据没有给出推广时间表、承诺单元数、合同期限、续约触发条件或经济性。CATL 也被描述为 Galbot 投资人,因此部署证据和战略支持相互交织,不能视为完全独立的客户验证。由于没有公开 NRR、GRR、流失、续约、利用率、满意度或 cohort 数据,不能用公开证据判断客户持久性。[CU003, CU004, CU015]
Galbot 已从任务选择走到一个生产示范项目,但可复制性和续约仍未得到验证。
各阶段描述了已有证据和缺失证据所对应的路径;不暗示转化率、时间安排或已承诺的推广量。
[CU001, CU003, CU005, CU015]6.4 客户风险与投资判断含义
客户风险主要来自验证集中度和行业商业化不确定性。审阅的 Galbot 特定证据只点名 CATL 为生产案例;独立行业分析称人形机器人仍昂贵、任务受限,且常在特定场地试用。Associated Press 还报道 2025 年中国人形机器人订单很大程度由国有企业和研究需求支撑,买方可能落后于制造能力。Galbot 仍可能建立持久的工业切入口,但当前证据无法区分付费生产单元和战略试点,无法量化收入集中度,也无法证明续约经济性。因此,投资逻辑应要求来自多个无关联客户的合同级证据。[CU005, CU006, CU007, CU008, CU015]
6.5 展示项
07风险
7.1 风险排序
商业化和执行是剩余风险中最高的部分。独立报道称人形机器人供给能力可能领先于买方需求,机器人仍昂贵、脆弱且依赖高度结构化环境。MERICS 另称中国人形机器人仍缺乏精度和灵巧性,通常局限于有限任务或特定场地试验。这些发现直接关系到 Galbot:成功的 CATL 部署证明其能进入高要求工业环境,却不能证明跨客户或非结构化工作流的经济性可复制。因此,投资逻辑应建立在已测量的任务成功率、干预频率、运行时间、部署成本和续约或扩张证据上,而不是订单公告或技术演示。UBTECH 的结果也强化了财务风险:其 2025 年人形机器人收入大增但仍亏损,说明行业商业化可以先于经营自立。[CR001, CR002, CR003, CR007, CR012, CR013, CR015, CR016]
| 失效模式 | 可能性 | 严重程度 | 缓解措施成熟度 | 残余风险 | 尚未解决的缺口 |
|---|---|---|---|---|---|
| 在狭窄、结构化任务之外,自主性或灵巧性有限 | 高 | 高 | 早期;行业证据仍集中于试验和有限任务 | 高 | Galbot 任务成功率、干预率和泛化结果未披露 |
| 脆弱性、停机或现场服务负担侵蚀客户 ROI | 高 | 高 | 部分披露;已宣布与 CATL 的服务整合 | 高 | 车队正常运行时间、故障模式、服务成本和 SLA 表现均未披露 |
| 成本仍高于替代人工的经济性 | 中 | 高 | 早期;暂无 Galbot 单位成本或毛利率证据 | 高 | 物料清单、部署成本、回本周期和毛利率均未披露 |
| 商业产能增长快于重复购买需求 | 高 | 高 | 部分披露;据报道已有一条重要生产线部署 | 高 | 积压订单质量、验收、续约、扩张和客户集中度均未披露 |
风险判断结合了行业证据和据报道的 CATL 部署;不会用行业观察替代 Galbot 车队指标。
[CR001, CR002, CR003, CR006, CR007, CR008]残余风险最高的领域,是行业约束与 Galbot 特定运营证据缺失交汇之处。
可能性、影响和成熟度评级是对引用证据的定性综合;不代表任何数值概率或损失估计。
[CR001, CR003, CR004, CR005, CR007, CR009, CR011, CR012, CR015, CR016]7.2 监管、法律与地缘政治风险
指定证据没有证明 Galbot 的产品安全认证、隐私控制、事故记录、保险覆盖、出口管制分类或诉讼状况。这是重大尽调缺口,不是没有事故或程序的证据。外部证据最清楚的监管风险是前瞻性的:MERICS 指出公司依赖 Nvidia 机器人生态,并称能力更强的边缘计算芯片可能成为出口管制目标。CNBC 还报道中美紧张和国家安全审查已令跨境投资降温。即使没有针对公司的执法行动,这些条件也可能限制计算资源、融资来源、海外客户和退出选择。投资文件应要求按司法辖区编制的合规矩阵、安全测试文件、事故日志、数据治理架构和供应商替代计划,之后再视为风险已缓解。[CR004, CR005, CR009, CR010]
| 规则、许可证或案件 | 法域 | 状态 | 可能性 | 严重程度 | 缓解措施 | 残余风险 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| 海外机器人计算资源的出口管制风险 | 美国 / 中国 | 现有描述未将边缘机器人芯片列为当前重点,但 MERICS 识别出未来管制风险 | 中 | 高 | 评估国产或不受限的计算资源,并保持模型可迁移性 | 在替代方案达到性能要求前维持高位 | 获取计算资源物料清单、分类、许可证和经测试的替代方案 |
| 影响跨境资本的国家安全审查 | 美国 / 中国 | CNBC 报道称,审查加强令跨境投资降温 | 高 | 中 | 分散资本来源,并预先筛查投资者和客户所在法域 | 中,因为融资和退出路径仍暴露于地缘政治风险 | 审查受益所有人、制裁筛查、对外投资风险和退出限制 |
| 机器人安全、隐私、产品责任和事故合规 | 中国及每个部署市场 | 指定公开证据未予证实 | 未知 | 高 | 要求正式安全论证、数据地图、保险和事故响应控制 | 在取得一手合规证据前维持高位 | 检查认证、测试报告、事故日志、保险保单、DPA 和诉讼检索结果 |
该登记表仅基于指定的 2026 年证据;没有可获得的直接监管机构、法院、认证或 Galbot 事故记录。
[CR004, CR005, CR009, CR010]7.3 运营与合作伙伴依赖
Galbot 与 CATL 的关系既是部署证据,也是集中度风险。报道说 S1 进入 CATL 智能产线执行物料搬运和拣选,CATL 的 Ning Service 网络也预计扩展到实体机器人售后支持。这能降低现场服务摩擦,但也使部署可信度和支持能力部分依赖一个战略投资人兼客户生态。技术集中度更广:MERICS 报道公司依赖 Nvidia 软件和芯片,以及德国和日本的精密零部件供应商。监控时应区分 Galbot 控制的表现与合作伙伴提供的电池、服务、计算资源和零部件。淘汰标准包括无法认证替代计算资源或关键零部件、人工干预持续高于合同阈值,或无法在两个无关联客户处复现 CATL 水平的表现。[CR004, CR005, CR007, CR008, CR016]
| 依赖项 | 交易对手 | 角色 | 集中度 | 失效情景 | 严重程度 | 缓释措施 | 剩余敞口 |
|---|---|---|---|---|---|---|---|
| 机器人计算与软件生态 | Nvidia | 边缘计算、仿真和开发工具 | 据报道,行业普遍存在依赖;Galbot 的具体占比未披露 | 控制措施、供应限制或平台变更限制开发 | 高 | 可移植模型和经过认证的替代计算资源 | 高 |
| 精密部件 | 德国和日本供应商 | 包括特殊滚珠丝杠在内的高端定位部件 | 中国行业依赖度较高 | 出口限制、短缺或成本上涨拖慢生产 | 高 | 关键部件双来源采购,并验证国产替代品 | 中高 |
| 工业部署和现场服务 | CATL / Ning Service | 示范生产线、电池合作关系和规划中的售后支持 | 战略集中度较高;收入占比未披露 | 关系丧失会削弱验证、服务覆盖和扩张 | 高 | 与无关联客户复制部署和服务能力 | 高 |
| 外部资本和退出渠道 | 中国、新加坡和中东的投资方 | CNBC 报道的融资生态 | 投资方地域多元,但没有美国参与者 | 地缘政治审查收紧未来融资或退出渠道 | 中 | 保持多条合规融资和退出路径 | 中 |
由于 Galbot 供应商占比、收入集中度、合同和替代测试均未公开,集中度评估采用定性判断。
[CR004, CR005, CR007, CR008, CR009, CR010]7.4 缓解重点与论点破坏触发条件
目前最明显的缓解措施,是与大型制造商开展结构化工业部署,并已宣布接入该伙伴的服务网络。但这不能解决需求广度、安全治理、可靠性或经济性。管理层应提供车队月度运行时间、自主任务完成率、每运行小时人工干预次数、平均故障间隔、现场服务成本、按部署计算的毛利率、应收账款账龄和客户集中度。生产部署停滞、反复发生安全关键事故、无法在没有合格替代品时获得必要计算资源、失去 CATL 服务支持,或规模扩大后单位经济性仍不改善,都应触发论点破坏。竞争集中度也很关键:TrendForce 预计 Unitree 和 AgiBot 将拿下 2026 年出货量近 80%,这会增加 Galbot 融资和估值跑在商业份额之前的风险。[CR001, CR006, CR007, CR008, CR011, CR014, CR016]
7.5 展示项
08估值
8.1 建议与价格纪律
建议是继续研究,而不是买入或放弃。CNBC 报道 Galbot 2026 年 4 月私有估值超过 $3 billion,CNEVPost 报道 3 月宣布 RMB2.5 billion($368 million)融资。这些估值显示公司能获得资本和投资人需求,但都不能确定完全摊薄口径的企业价值、最新证券附带的优先权,或当前经营倍数。审阅的公开记录没有 Galbot 收入、毛利率、现金消耗、订单积压、客户集中度或单位经济性披露。没有这些分母,公允价值区间或目标回报都会是虚假精确。因此,入场应以取得一手数据室、合同级部署证明和融资条款审阅为条件。[CV001, CV004, CV016]
| 决策字段 | 评估 | 决策含义 |
|---|---|---|
| 建议 | 继续研究 | 在获得一手财务和合同证据前,不应按当前价格判断是否入场。 |
| 信心 | 中低 | 部署和融资信号清晰可见,但核心估值分母缺失。 |
| 风险评级 | 高 | 商业化、资本密集度、需求深度、稀释和流动性仍是重大风险。 |
| 估值立场 | 偏高 | 据报道超过 $3 billion 的估值并没有 Galbot 已披露经营指标的支持。 |
| 价格纪律 | 没有可辩护的公开公允价值区间 | 要求基于数据室信息、采用完全摊薄口径并经过现金需求情景测试的估值。 |
评估是基于指定公开来源集得出的分析师结论;没有 Galbot 经审计的财务报表、股权结构表、证券条款或当前收入分母。
[CV001, CV004, CV006, CV008, CV009, CV014, CV016]8.2 投资论点与反论点
正面论点并非纯粹宣传:CNEVPost 报道 Galbot S1 进入 CATL 智能产线,执行长期自主物料搬运和拣选。这是具有战略意义的客户环境,比阶段性演示更有说服力。反论点是,行业供给增长快于经核实的需求。Associated Press 报道 2025 年中国有超过 140 家人形机器人制造商和超过 330 个型号,且用例有限;MERICS 将多数部署描述为有限任务和特定场地试验。Robotics & Automation News 主页摘要也把从演示、试点走向工厂描述为制造规模挑战。因此,Galbot 有技术和伙伴进展证据,但尚无公开证据证明商业经济性可重复。[CV005, CV006, CV008, CV015]
| 立场 | 论点 | 哪些信息会改变判断 |
|---|---|---|
| 正论 | 据报道,Galbot S1 已在 CATL 部署,用于自主搬运和拣选。 | 客户签署的正常运行时间、人工干预、吞吐量、安全和付费扩张数据将增强这一说法。 |
| 正论 | 据报道,RMB2.5 billion 融资和超过 $3 billion 的估值表明投资者支持力度较大。 | 要把融资额转化为现金跑道判断,还需要完全摊薄条款、投资者权利、剩余现金和资金用途证据。 |
| 反论 | 中国人形机器人供应量和模型数量都很高,但经过验证的买方需求和自主使用场景仍受限。 | 非关联客户以已披露经济性重复下单,将降低商业化风险。 |
| 反论 | 行业成本和现场效率仍不足以支撑广泛商业化。 | 若有证据表明交付成本低于替代人工阈值,且服务负担可接受,投资逻辑会改善。 |
部署细节由 CNEVPost 转述 CATL 的表述;融资数字是报道值,而非经审计的现金或完全摊薄估值证据。
[CV001, CV004, CV005, CV006, CV008, CV009]证据从市场机会、部署证明和经济性缺口逐步指向继续研究的决策。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 政策支持的大市场 [market] | 中国拥有制造规模和国家支持,但行业仍处于早期。 | |
| 节点 2 | 据报道 Galbot 估值 $3B+ [price] | 这一非上市估值是融资参考值,而非公开市场价格。 | |
| 节点 3 | CATL 部署信号 [proof] | 据报道,Galbot S1 长期执行生产线搬运和拣选作业。 | |
| 节点 4 | 财务分母缺失 [economics] | 收入、利润率、现金消耗、签约车队经济性和股权结构表条款均不可得。 | |
| 节点 5 | 商业化和稀释风险 [risk] | 需求、降本、服务负担和未来融资可能实质性改变结果。 | |
| 节点 6 | 继续研究 [decision] | 在确定入场价格前,要求一手财务、合同、车队和证券条款证据。 | |
| 连接 1 | 政策支持的大市场 [market] | 据报道 Galbot 估值 $3B+ [price] | |
| 连接 2 | 据报道 Galbot 估值 $3B+ [price] | CATL 部署信号 [proof] | |
| 连接 3 | CATL 部署信号 [proof] | 财务分母缺失 [economics] | |
| 连接 4 | 财务分母缺失 [economics] | 商业化和稀释风险 [risk] | |
| 连接 5 | 商业化和稀释风险 [risk] | 继续研究 [decision] |
该顺序只表达决策依赖关系;不对概率、企业价值或预期回报作量化。
[CV003, CV005, CV008, CV009, CV014, CV015, CV016]8.3 可比公司与情景逻辑
私有同业估值只能作为情绪参考。CNBC 将 Figure 估值置于至少 $39 billion、Apptronik 为 $5 billion、Galbot 超过 $3 billion、AI2 Robotics 约 $2.93 billion,同时报道美国人形机器人初创公司被视为广义 AI 平台,中国公司则更多被视为工业硬件企业。这些价格没有对证券权利或经营成熟度做标准化。UBTECH 提供了更具体的运营参照,因为其 2025 年申报文件披露收入 RMB2.001 billion、全尺寸实体人形产品和服务收入 RMB820.6 million、37.7% 毛利率及 RMB789.8 million 会计亏损。但这仍是竞争对手申报文件,不是独立估值基准;其多元化产品组合也不适合直接套用倍数。[CV002, CV003, CV010, CV011, CV012, CV013, CV014]
| 情景 | 假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 乐观 | CATL 部署扩展为重复付费车队;单位成本下降;正常运行时间和服务经济性验证了工业规模化使用。 | 当前超过 $3 billion 的估值可能变得合理,但没有公开的终值区间或目标回报可供计算。 | 执行、客户集中度、安全和资本密集度。 | 只有审计确认增长、重复订单和按客户队列拆分的车队经济性后,才算正面信号。 |
| 基准 | Galbot 仍是可信的技术竞争者,拥有选择性部署,但经营披露仍然有限。 | 应跟踪而不是定价;由于收入和利润率分母不可得,当前估值仍偏高。 | 商业化周期长、融资稀释和不透明的优先权安排。 | 当前公开证据最接近支持这一情景。 |
| 悲观 | 需求仍由试点或补贴驱动,降本停滞,或下一轮融资定价低于当前估值。 | 根据公开证据,下行空间没有上限;非上市股份缺乏流动性,加上优先权悬垂,可能放大损失。 | 下轮降价融资、退出市场疲软、集中度和持续现金需求。 | 续约失败、自主运行时长有限或融资条款恶化时触发。 |
这些情景是条件式框架而非预测;由于 Galbot 的财务报表和证券条款不可得,估值区间和概率有意不作量化。
[CV005, CV006, CV007, CV008, CV009, CV013, CV016]8.4 退出准备度、尽调要求与论点破坏
仅凭指定公开证据无法确定退出准备度。融资估值不等于流动性,已募集资金也不等于可用现金;审阅来源没有 Galbot IPO 时间表、二级市场条款、经审计财务历史或收购先例。关键尽调要求包括带清算优先权的完全摊薄股权结构表、经审计收入和现金流报表、已签署客户合同、车队运行时间和干预日志、保修及服务成本,以及从部署到毛利的桥接。论点破坏触发条件是可量化的:无法把 CATL 相关工作转化为付费的重复部署、经济性长期高于可行的人工替代水平、依赖补贴或国资关联需求,或融资轮次低于当前估值。在这些问题验证前,下行无法界定,上行也无法转化为目标回报。[CV005, CV007, CV009, CV013, CV016]
8.5 展示项
免责声明
本报告基于保留的公开证据,不构成投资、法律、税务、会计或工程建议。非上市公司的披露并不完整,任何投资决策前都应通过一手尽调加以核实。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Galaxy Bot, also known as Galbot, was founded in 2023 by He Wang. | 中 | SO003, SO005 |
| CO002 | Multiples classifies Galaxy Bot as privately held and headquartered in Beijing. | 中 | SO002 |
| CO003 | The same Multiples page describes Galaxy Bot as Shenzhen-headquartered, conflicting with its Beijing headquarters field. | 中 | SO002 |
| CO004 | Galbot develops embodied-AI models and robot systems for targeted retail, navigation, manipulation, and industrial tasks. | 高 | SO001, SO004 |
| CO005 | Galbot's website announced bookings for a native-agent robot on September 3, 2026. | 中 | SO001 |
| CO006 | Public reporting identifies He Wang as a Peking University researcher and separately describes him as Stanford-trained. | 中 | SO003, SO005 |
| CO007 | The CATL and Puquan Capital-led June 2025 round was reported at $151 million by Robotics & Automation News and $153 million by Multiples. | 中 | SO002, SO005 |
| CO008 | Multiples reports a $300 million December 2025 financing at a $3 billion valuation, while CNBC later described Galbot's valuation as $3 billion-plus. | 高 | SO002, SO009 |
| CO009 | Galbot's March 2026 financing was reported at RMB2.5 billion, with dollar conversions ranging from approximately $350 million to $368 million. | 中 | SO002, SO003 |
| CO010 | Multiples states that Galaxy Bot has raised $619 million across five funding rounds. | 低 | SO002 |
| CO011 | Elsewhere on the same page, Multiples says total funding had reached $800 million in December 2025 before a subsequent March 2026 financing. | 低 | SO002 |
| CO012 | CnEVPost reports that the Galbot S1 formally entered CATL module and battery-pack production for long-running autonomous material handling and picking. | 中 | SO003 |
| CO013 | Robotics & Automation News reported ten Galbot convenience-store pilot locations in Beijing and a plan to expand to one hundred stores. | 中 | SO005 |
| CO014 | MERICS assesses Galbot's GraspVLA, TrackVLA, and GroceryVLA as narrow-remit models that illustrate the difficulty of fully autonomous robotics. | 中 | SO004 |
| CO015 | Associated Press reporting says humanoid demand may lag production capacity and that many current systems remain limited in real-world functionality. | 中 | SO008 |
| CO016 | UBTECH reported 1,079 full-size embodied-intelligence robot units and RMB820.6 million of related revenue for 2025. | 高 | SO006, SO007 |
| CM001 | Galaxy Bot's near-term market is bounded embodied-AI deployment for industrial, logistics, retail, healthcare, service, and research workflows rather than all robotics. | 中 | SM001, SM004, SM005, SM007 |
| CM002 | Fixed industrial robots, robotic arms, automated logistics equipment, and manual labor are status-quo substitutes that can be cheaper or more efficient for repetitive tasks. | 中 | SM007, SM008 |
| CM003 | MERICS reported that China produced 12,800 humanoids in 2025, about 90% of the global total. | 中 | SM007 |
| CM004 | China's 556,000 industrial robots produced in 2025 dwarfed its humanoid production. | 中 | SM007 |
| CM005 | Associated Press reporting cited more than 13,000 global humanoid shipments in 2025 and an approximately 85% share for Chinese products. | 中 | SM008 |
| CM006 | Morgan Stanley expected China's humanoid sales to reach approximately 28,000 units in 2026. | 中 | SM008 |
| CM007 | Omdia forecast annual advanced-robot shipments above one million units by the early 2030s. | 低 | SM008 |
| CM008 | Associated Press reporting cited a Morgan Stanley estimate of a $5 trillion humanoid robot market without stating the forecast horizon in the fetched copy. | 低 | SM008 |
| CM009 | UBTECH reported 1,079 full-size embodied humanoid robots sold and RMB820.6 million of related revenue in 2025. | 高 | SM002, SM003 |
| CM010 | The clearest current buyer workflows are industrial handling, sorting, assembly, inspection, and logistics rather than unconstrained general-purpose household work. | 中 | SM003, SM007, SM008 |
| CM011 | TechNode reported Galbot's claim of several thousand cumulative orders from industrial customers including CATL, Bosch, Toyota, BAIC Group, and SAIC Motor. | 中 | SM005 |
| CM012 | CATL reportedly deployed Galbot S1 for autonomous material handling and picking on a smart battery-production line. | 中 | SM006 |
| CM013 | MERICS cited CNY300,000-500,000 as the typical Chinese humanoid price range versus a CNY160,000 commercial-viability threshold based on two-year labor-cost payback. | 中 | SM007 |
| CM014 | Independent reporting found humanoid use cases limited, operations fragile, and demand insufficient to support broad mass production outside structured environments. | 中 | SM007, SM008 |
| CM015 | MERICS reported that Chinese companies supplied 57% of China's industrial-robot market in 2024 and represented 63% of key firms in the global humanoid-component supply chain. | 中 | SM007 |
| CM016 | China's embodied-AI sector, including Galbot, remains materially dependent on Nvidia software and edge-compute products. | 中 | SM007 |
| CP001 | Galaxy Bot had raised $1.1 billion in total disclosed funding, with its latest disclosed growth round announced in March 2026. | 中 | SP001 |
| CP002 | China had more than 140 humanoid robot manufacturers and more than 330 models in 2025. | 中 | SP002 |
| CP003 | AgiBot and Unitree each shipped more than 5,000 humanoid robots in 2025, while Figure and Tesla each shipped a few hundred or fewer. | 中 | SP002, SP003 |
| CP004 | CNBC reported a valuation above $3 billion for Galbot, at least $39 billion for Figure, and $5 billion for Apptronik in 2026. | 中 | SP003 |
| CP005 | MERICS characterized Galbot as intelligence-first, with GraspVLA, TrackVLA, and GroceryVLA models for narrow manipulation, navigation, and retail tasks and a preference for synthetic training data. | 中 | SP007 |
| CP006 | Chinese vendors supplied more than 57% of China's industrial robot market in 2024, overtaking foreign competitors for the first time. | 中 | SP007 |
| CP007 | China controlled 63% of key companies in the global humanoid-component supply chain, while high-end ball screws remained heavily dependent on German and Japanese suppliers. | 中 | SP007 |
| CP008 | MERICS placed China's average industrial-humanoid price at RMB300,000-RMB500,000 and cited RMB160,000 as an illustrative commercial-viability threshold. | 中 | SP007 |
| CP009 | MERICS listed Unitree's G1 at about EUR11,650, compared with about EUR120,000 or more for Boston Dynamics' Atlas and about EUR25,800 for Tesla Optimus. | 中 | SP007 |
| CP010 | UBTECH reported RMB820.6 million of full-size humanoid products-and-services revenue within RMB2.001 billion of total 2025 revenue. | 中 | SP008, SP009 |
| CP011 | UBTECH reported selling 1,079 full-size embodied-intelligence robots in 2025 across industrial and other applications. | 中 | SP008 |
| CP012 | UBTECH reported more than 6,000 units of annualized full-size humanoid production capacity at year-end 2025 and 1,000-unit-level Walker S2 mass production and delivery. | 中 | SP009 |
| CP013 | CnEVPost reported that Galbot's 50-kilogram dual-arm-payload S1 entered CATL's production line and that the partners planned to extend CATL's aftermarket service network to embodied robots. | 中 | SP006 |
| CP014 | China's humanoid digital-ID regime requires registration for domestic sale or deployment and had issued lifecycle codes to more than 28,000 units across 200 models. | 中 | SP005 |
| CP015 | MERICS assessed that Chinese humanoids remained limited to site-specific tasks and were still too expensive for widespread deployment. | 中 | SP007 |
| CP016 | Associated Press reported that demand may lag production capacity and that existing fixed robot arms can remain preferable for repetitive factory work. | 中 | SP002 |
| CI001 | Galbot's S1 robot has entered CATL's smart production line for long-running material-handling and picking work. | 中 | SI005 |
| CI002 | CATL and Galbot plan to incorporate the S1 into CATL's after-sales system for embodied-intelligence robots. | 中 | SI005 |
| CI003 | CnEVPost reported that Galbot completed a CNY 2.5 billion (USD 368 million) financing round in March 2026. | 中 | SI005 |
| CI004 | CNBC reported Galbot's valuation at more than USD 3 billion in April 2026. | 中 | SI002 |
| CI005 | Associated Press reported expert concern that humanoid-robot demand lags the industry's capacity to build them. | 中 | SI001 |
| CI006 | Associated Press reported that much of China's more than CNY 2 billion of humanoid orders in 2025 came from state-owned enterprises. | 中 | SI001 |
| CI007 | Associated Press reported peer asking prices of about USD 99,000 for Matrix Robotics' MATRIX-3 and CNY 180,000 (USD 26,600) for a basic EngineAI humanoid. | 中 | SI001 |
| CI008 | MERICS reported an average price of CNY 300,000-500,000 for humanoids in China. | 中 | SI006 |
| CI009 | MERICS cited a CNY 160,000 commercial-viability threshold based on CNY 80,000 annual labor cost and a two-year ROI period. | 中 | SI006 |
| CI010 | Independent sector analyses expect substantial humanoid price compression before widespread adoption. | 中 | SI003, SI006 |
| CI011 | Sacra describes Fireworks AI's usage-priced infrastructure model and approximately 50% gross margin, a software-compute structure that is not a direct benchmark for humanoid robotics. | 低 | SI004 |
| CI012 | UBTECH reported RMB 820.6 million of full-size humanoid products-and-services revenue in 2025, equal to 41.1% of RMB 2.001 billion total revenue. | 中 | SI007, SI008 |
| CI013 | UBTECH reported a 37.7% group gross margin for 2025, up from 28.7% in 2024. | 中 | SI007, SI008 |
| CI014 | UBTECH reported a RMB 789.8 million net loss for 2025 despite its revenue growth. | 中 | SI007, SI008 |
| CI015 | UBTECH reported RMB 784.1 million of cash used in operations in 2025. | 中 | SI008 |
| CI016 | UBTECH recognizes goods revenue at a point in time and reported that some contracted logistics projects were not recognized in 2025 because delivery and acceptance had not occurred by year-end. | 中 | SI008 |
| CE001 | Galbot S1 has a reported 50-kilogram dual-arm payload capacity. | 中 | SE005 |
| CE002 | Galbot S1 has entered CATL's production line for long-running autonomous material-handling and picking work. | 中 | SE005 |
| CE003 | Galbot S1 is reported to use vision-only centimeter-level positioning and 360-degree omnidirectional obstacle avoidance. | 中 | SE005 |
| CE004 | CATL attributed an eight-hour battery-life capability to the S1 configuration. | 中 | SE005 |
| CE005 | CATL said its Ning Service network would extend inspection, maintenance, and recycling support to embodied-intelligence robots. | 中 | SE005 |
| CE006 | MERICS describes Galbot G1 as using GraspVLA, TrackVLA, and GroceryVLA for dexterous hands, navigation, and retail scenarios respectively. | 中 | SE006 |
| CE007 | MERICS reports that Galbot's DexGraspNet is a simulated robotic-hand grasp dataset built with Nvidia Isaac Sim. | 中 | SE006 |
| CE008 | MERICS reports that Galbot was among the first Chinese robotics companies to receive Nvidia's Jetson Thor module. | 中 | SE006 |
| CE009 | MERICS assesses that Chinese humanoids generally still lack precision and dexterity and remain concentrated in limited tasks and site-specific trials. | 中 | SE006 |
| CE010 | MERICS estimates that humanoid costs in China would need to fall by at least half for widespread deployment. | 中 | SE006 |
| CE011 | ISO states that standards set consistent benchmarks intended to support reliability and trust. | 中 | SE002 |
| CE012 | The retained evidence set provides no Galbot-specific certificate, conformity assessment, formal safety test, status history, or incident register. | 低 | SE002, SE005, SE006 |
| CE013 | The retained evidence set provides no Galbot repository activity, release history, package-download metric, API forum, or practitioner-community activity measure. | 低 | SE001, SE006 |
| CE014 | UBTECH reported that Walker S2 entered mass production and delivery in 2025, providing a disclosed peer benchmark for industrial humanoid maturity. | 中 | SE008 |
| CE015 | UBTECH linked 1,000-unit-level Walker S2 production maturity to preventive design, manufacturing-process optimization, module testing, and complete-robot testing. | 中 | SE008 |
| CE016 | Yicai reported that UBTECH sold 1,079 full-size embodied-AI robots in 2025 and derived CNY820.6 million, or 41 percent of total revenue, from them. | 中 | SE007, SE008 |
| CU001 | CATL and Galbot signed a strategic cooperation agreement aimed at smart-production-line upgrades and scaled deployment of embodied-intelligence robots. | 中 | SU005 |
| CU002 | The Galbot S1 entered CATL's smart production line for long-running autonomous material handling and picking in module and battery-pack production. | 中 | SU005 |
| CU003 | CATL and Galbot plan to extend CATL's Ning Service inspection, maintenance, and recycling system to embodied-intelligence robots. | 中 | SU005 |
| CU004 | CATL was reported as the only embodied-intelligence company in which CATL had invested as of June 2026. | 中 | SU005 |
| CU005 | MERICS assessed that Chinese humanoids were mostly limited to narrow tasks and site-specific trials rather than broad autonomous deployment. | 中 | SU006 |
| CU006 | MERICS estimated that humanoid costs would need to fall by at least half to reach commercial viability. | 中 | SU006 |
| CU007 | Associated Press reported that more than RMB2 billion of 2025 Chinese humanoid orders came largely from state-owned enterprises for research and uses in power plants, data centers, and entertainment. | 中 | SU001 |
| CU008 | Independent reporting warned that humanoid demand may lag manufacturing capacity because use cases remain limited and production economics are difficult. | 中 | SU001, SU006 |
| CU009 | China's manufacturing robot density reached 470 operational industrial robots per 10,000 employees in 2023. | 中 | SU003 |
| CU010 | CNBC reported in April 2026 that Chinese humanoid startups were already shipping robots to factories and malls. | 中 | SU002 |
| CU011 | Yicai reported that UBTECH sold 1,079 full-size embodied-intelligence robots in 2025 and generated CNY820.6 million from them. | 中 | SU007 |
| CU012 | UBTECH's annual-results announcement reported RMB820.6 million of 2025 full-size humanoid products-and-services revenue, up approximately 2,203.7% from 2024. | 中 | SU008 |
| CU013 | UBTECH said its 2025 humanoid order coverage included automotive manufacturing, smart manufacturing, smart logistics, and embodied-intelligence data centers. | 中 | SU008 |
| CU014 | MERICS described Galbot's GroceryVLA as a narrowly scoped model for retail scenarios rather than evidence of a disclosed retail customer deployment. | 中 | SU006 |
| CU015 | The reviewed evidence provides Galbot-specific production proof only for CATL and does not establish Galbot customer count, revenue concentration, retention, or renewal economics. | 中 | SU002, SU005, SU006 |
| CR001 | Independent 2026 reporting says humanoid-robot demand may lag production capacity and that many use cases remain limited. | 中 | SR001 |
| CR002 | Associated Press reporting describes humanoid robots as expensive to produce, fragile in operation, and dependent on highly structured environments. | 中 | SR001 |
| CR003 | MERICS assesses Chinese humanoids as lacking precision and dexterity and remaining concentrated in limited tasks and site-specific trials. | 中 | SR006 |
| CR004 | MERICS identifies Nvidia's AI-chip and software ecosystem as a dependency for China's embodied-AI sector. | 中 | SR006 |
| CR005 | MERICS reports that German and Japanese suppliers provide 90% of special ball screws used for precise positioning applications in robotics. | 中 | SR006 |
| CR006 | MERICS estimates that Chinese humanoid costs would need to fall by at least half for commercial viability. | 中 | SR006 |
| CR007 | CnEVPost reported that Galbot's S1 formally entered CATL's smart production line for autonomous material handling and picking. | 中 | SR005 |
| CR008 | CnEVPost reported that CATL planned to extend its Ning Service after-sales system to embodied-intelligence robots. | 中 | SR005 |
| CR009 | CNBC reported that US-China tensions and national-security scrutiny have chilled cross-border investment. | 中 | SR002 |
| CR010 | CNBC reported Galbot at a valuation above USD 3 billion with backers from China, Singapore, and the Middle East rather than the United States. | 中 | SR002 |
| CR011 | TrendForce projects Unitree and AgiBot to account for nearly 80% of 2026 humanoid-robot shipments. | 中 | SR003 |
| CR012 | Yicai reported that UBTECH's 2025 revenue rose 53% to RMB 2 billion while net loss remained RMB 789.8 million. | 中 | SR007, SR008 |
| CR013 | UBTECH's filing reported a 2025 net loss of RMB 789.8 million despite rapid humanoid-product revenue growth. | 中 | SR008 |
| CR014 | UBTECH recorded RMB 151.4 million of 2025 credit impairment losses primarily from delayed payments by certain government-related customers. | 中 | SR008 |
| CR015 | UBTECH's 2025 capital expenditure rose 53.2% to RMB 613.5 million, illustrating the capital intensity of a scaled robotics peer. | 中 | SR008 |
| CR016 | The evidence supports treating structured industrial deployment and partner service integration as partial mitigations rather than proof of broad reliability or commercial repeatability. | 中 | SR001, SR005, SR006 |
| CV001 | CNBC reported that Galbot had a private valuation above $3 billion in April 2026. | 中 | SV002 |
| CV002 | CNBC reported private valuation references of at least $39 billion for Figure, $5 billion for Apptronik, above $3 billion for Galbot, and approximately $2.93 billion for AI2 Robotics. | 中 | SV002, SV004 |
| CV003 | CNBC reported that investors tend to frame US humanoid startups as broad AI platforms and Chinese startups more as industrial hardware businesses. | 中 | SV002 |
| CV004 | CNEVPost reported that Galbot announced a RMB2.5 billion ($368 million) financing round in March 2026. | 中 | SV005 |
| CV005 | CNEVPost reported that Galbot S1 entered CATL's smart production line for long-running autonomous material-handling and picking work. | 中 | SV005 |
| CV006 | Associated Press reported that China had more than 140 humanoid manufacturers and more than 330 models in 2025 while demand lagged production capacity. | 中 | SV001 |
| CV007 | Associated Press reported that many of China's more than RMB2 billion of humanoid orders in 2025 came from state-owned enterprises. | 中 | SV001 |
| CV008 | MERICS reported that China produced 12,800 humanoids in 2025 versus 556,000 industrial robots and that humanoids remained concentrated in limited tasks and site-specific trials. | 中 | SV006 |
| CV009 | MERICS reported average Chinese humanoid costs of RMB300,000 to RMB500,000 and cited a RMB160,000 commercial-viability threshold based on a two-year labor payback. | 中 | SV006 |
| CV010 | UBTECH reported RMB2.001 billion of 2025 revenue, including RMB820.6 million from full-size embodied intelligent humanoid products and services. | 中 | SV007, SV008 |
| CV011 | UBTECH reported a 37.7% gross margin and a RMB789.8 million accounting loss for 2025. | 中 | SV007, SV008 |
| CV012 | UBTECH reported sales of 1,079 full-size embodied intelligent humanoid robots in 2025. | 中 | SV007, SV008 |
| CV013 | UBTECH reported RMB784.1 million of cash used in operations and RMB4.8879 billion of cash and cash equivalents for 2025. | 中 | SV008 |
| CV014 | Robozaps assessed that Figure's $39 billion valuation was not validated by operating results and noted missing public revenue, unit-economics, uptime, intervention, and customer-concentration data. | 低 | SV004 |
| CV015 | Robotics & Automation News framed manufacturing scale as a challenge as humanoid developers move from demonstrations and pilots into factories. | 低 | SV003 |
| CV016 | The assigned public evidence does not support a defensible Galbot fair-value range, target return, or exit underwriting because Galbot operating financials and financing rights are undisclosed. | 中 | SV001, SV002, SV005, SV006, SV008 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Galbot | Galbot 银河通用机器人官方网站 | 全球首个集「大脑—小脑—神经控制」于一模的全身端到端具身大模型。 |
| SO002 | Multiples | Galaxy Bot Valuation, Funding & Investors | Galaxy Bot has raised a total of $619M across 5 funding rounds. |
| SO003 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The robot has formally entered CATL's smart production line. It carries out long-running autonomous work in module and battery pack production, replacing manual labor in tasks such as material handling and picking. |
| SO004 | Mercator Institute for China Studies | Embodied AI: China’s ambitious path to transform its robotics industry | China’s humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. They remain far from being fully autonomous machines capable of perceiving and responding in real time in the physical world. |
| SO005 | Robotics & Automation News | Galbot raises $151 million to scale embodied AI humanoid robots, partners with Bosch investment arm | Chinese robotics startup Galaxy General Robot Co. – also known as Galbot – has raised $151 million in a funding round led by battery giant CATL and Puquan Capital, bringing its total investment to more than $330 million since its founding in 2023. |
| SO006 | Yicai Global | China’s Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million (USD119.2 million), or 41 percent of its total revenue |
| SO007 | UBTECH Robotics Corp Ltd | Annual Results Announcement for the year ended December 31, 2025 | In 2025, full-size embodied intelligent humanoid robot products and services for all scenarios achieved revenue of approximately RMB820 million, representing a year-on- year increase of approximately 2,203.7%; and achieved sales volume of 1,079 units |
| SO008 | Associated Press | China can build humanoids at scale. The hard part is finding enough buyers | Most humanoid robots are still performative rather than functional, falling short of working in messy, unpredictable environments |
| SO009 | CNBC | CNBC's The China Connection newsletter: China ships more humanoid robots than the U.S. as investors diverge on AI bets | |
| SM001 | DroidAge | Galaxy Bot — Humanoid Robots | Products include the Galbot G1 and Galbot S1 series, used in retail, industrial manufacturing, logistics, healthcare, and service applications. |
| SM002 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million (USD119.2 million), or 41 percent of its total revenue |
| SM003 | UBTECH Robotics Corp Ltd via Hong Kong Exchanges and Clearing | Annual Results Announcement for the year ended December 31, 2025 | In 2025, full-size embodied intelligent humanoid robot products and services for all scenarios achieved revenue of approximately RMB820 million |
| SM004 | Crunchbase News | June Hits 3-Year High In Unicorn Births Across AI, Robotics And More | Galaxy Bot, a developer of a humanoid robot for retail that manages inventory, replenishment and packaging, raised a $153 million funding |
| SM005 | TechNode | Humanoid robot maker Galbot raises RMB 2.5 billion | The company said it has secured cumulative orders totaling several thousand units from industrial clients including CATL, Bosch, Toyota, BAIC Group, and SAIC Motor. |
| SM006 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The Galbot S1 features a 50-kilogram dual-arm payload capacity, along with vision-only centimeter-level positioning and 360-degree omnidirectional obstacle avoidance. |
| SM007 | Mercator Institute for China Studies | Embodied AI: China's ambitious path to transform its robotics industry | China's humanoids are cheaper than those of western competitors but still far too expensive for widespread deployment. |
| SM008 | Associated Press via TechXplore | China can build humanoids at scale. The hard part is finding enough buyers | Most humanoid robots are still performative rather than functional, falling short of working in messy, unpredictable environments |
| SM009 | CNBC | CNBC's The China Connection newsletter: China ships more humanoid robots than the U.S. as investors diverge on AI bets | Chinese humanoid startups took the top six spots in Omdia's rankings of global robot shipments in 2025. |
| SP001 | Startup Fundraising | Galaxy Bot Funding: $1.1B Raised, Growth (2026) | Galaxy Bot has raised $1.1B in total disclosed funding. |
| SP002 | Associated Press via TechXplore | China can build humanoids at scale. The hard part is finding enough buyers | Demand for robots may lag behind |
| SP003 | CNBC | China ships more humanoid robots than the U.S. as investors diverge on AI bets | Chinese humanoid startups took the top six spots in Omdia's rankings of global robot shipments in 2025. |
| SP004 | U.S.-China Economic and Security Review Commission | Humanoid Robots | Humanoid robots are undergoing rapid development, led by firms in the United States and China. |
| SP005 | Xinhua | China rolls out digital ID system to regulate booming humanoid robot sector | The new standard enforces a strict "no code, no market access" rule. |
| SP006 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The robot has formally entered CATL's smart production line. |
| SP007 | Mercator Institute for China Studies | Embodied AI: China's ambitious path to transform its robotics industry | China's humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SP008 | Yicai Global | China's UBTECH Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year |
| SP009 | UBTECH Robotics Corp Ltd via HKEX | Annual Results Announcement for the year ended December 31, 2025 | Our revenue increased by 53.3% to RMB2,001.0 million for the year ended December 31, 2025 |
| SI001 | Associated Press | China can build humanoids at scale. The hard part is finding enough buyers | "The economics are tough: humanoid robots remain expensive to produce, fragile in operation, and dependent on highly structured environments to function," |
| SI002 | CNBC | China ships more humanoid robots than the U.S. as investors diverge on AI bets | Chinese humanoid startups are already shipping robots to factories and malls, while their U.S. rivals remain focused on development — and far higher valuations. |
| SI003 | Morgan Stanley | Humanoids: A $5 Trillion Market | The complexity of humanoids, which require sophisticated robot software models and tight integration with hardware, makes them an expensive product. |
| SI004 | Sacra | Fireworks AI revenue, valuation & funding | The gross margin profile, approximately 50%, with a stated target of 60%, reflects the reality that Fireworks is not a pure software business. |
| SI005 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The robot has formally entered CATL's smart production line. It carries out long-running autonomous work in module and battery pack production, replacing manual labor in tasks such as material handling and picking. |
| SI006 | Mercator Institute for China Studies | Embodied AI: China's ambitious path to transform its robotics industry | China’s humanoids are cheaper than those of western competitors but still far too expensive for widespread deployment. For commercial viability, the costs would have to fall by at least half. |
| SI007 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech's overall revenue jumped 53 percent to CNY2 billion (USD290.5 million) in the 12 months ended Dec. 31 from the previous year, while its net loss narrowed 32 percent to CNY789.8 million, the company noted. The gross profit margin rose to 38 percent from 29 percent. |
| SI008 | UBTECH Robotics Corp Ltd | Annual Results Announcement for the year ended December 31, 2025 | Cash used in operation in 2025 was approximately RMB784.1 million (RMB833.6 million in 2024). |
| SE001 | Peking University | Peking University official English site | Transforming academic innovation into real-world solutions |
| SE002 | International Organization for Standardization | ISO - International Organization for Standardization | Standards define what great looks like, setting consistent benchmarks for businesses and consumers alike — ensuring reliability, building trust, and simplifying choices. |
| SE003 | State Council of the People's Republic of China | China's retail sales up 3.5 pct in 2024 | |
| SE004 | Robotics & Automation News | China sets national standards for humanoid robots | As humanoid robot developers race to move their machines from demonstrations and pilot projects into factories and other working environments, another challenge is rapidly coming into focus |
| SE005 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The Galbot S1 features a 50-kilogram dual-arm payload capacity, along with vision-only centimeter-level positioning and 360-degree omnidirectional obstacle avoidance. |
| SE006 | Mercator Institute for China Studies | Embodied AI: China's ambitious path to transform its robotics industry | Galbot has an intelligence-first approach and aspires to reliable, fully autonomous mobile pick-and-place tasks. Its G1 robot’s three models have narrow remits that illustrate the difficulty of building fully autonomous robots: GraspVLA for dexterous robot hands, TrackVLA for navigation and GroceryVLA for retail scenarios. |
| SE007 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million (USD119.2 million), or 41 percent of its total revenue |
| SE008 | UBTECH Robotics Corp Ltd | Annual Results Announcement for the year ended December 31, 2025 | revenue from full-size embodied intelligent humanoid robot products and services grew rapidly by approximately 2,203.7%, from RMB35.6 million for the year ended December 31, 2024, to RMB820.6 million for the year ended December 31, 2025 |
| SU001 | Associated Press | China can build humanoids at scale. The hard part is finding enough buyers | Most humanoid robots are still performative rather than functional, falling short of working in messy, unpredictable environments, said Samm Sacks, a senior fellow at the New America think tank focused on Chinese technology. |
| SU002 | CNBC | China ships more humanoid robots than the U.S. as investors diverge on AI bets | Chinese humanoid startups are already shipping robots to factories and malls, while their U.S. rivals remain focused on development — and far higher valuations. |
| SU003 | International Federation of Robotics | Global Robot Density in Factories Doubled in Seven Years | China took third place in 2023, surpassing Germany and Japan. The country's push to the use of automation technology results in a high robot density of 470 robots per 10,000 employees (2022: 402 units). |
| SU004 | Open Robotics | ROS official homepage | |
| SU005 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The robot has formally entered CATL's smart production line. It carries out long-running autonomous work in module and battery pack production, replacing manual labor in tasks such as material handling and picking. |
| SU006 | Mercator Institute for China Studies | Embodied AI: China's ambitious path to transform its robotics industry | China’s humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SU007 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million (USD119.2 million), or 41 percent of its total revenue, the Shenzhen-based company said in its annual financial report released yesterday. |
| SU008 | UBTECH Robotics Corp Ltd | Annual Results Announcement for the year ended December 31, 2025 | revenue from full-size embodied intelligent humanoid robot products and services grew rapidly by approximately 2,203.7%, from RMB35.6 million for the year ended December 31, 2024, to RMB820.6 million for the year ended December 31, 2025, becoming UBTECH’s largest source of revenue. |
| SR001 | Associated Press via TechXplore | China can build humanoids at scale. The hard part is finding enough buyers | "The economics are tough: humanoid robots remain expensive to produce, fragile in operation, and dependent on highly structured environments to function," Sacks explained. |
| SR002 | CNBC | The China Connection: China ships more humanoid robots than the U.S. as investors diverge on AI bets | Geopolitics has complicated the investment landscape. |
| SR003 | TrendForce | China's Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share | Together, these two companies are projected to account for nearly 80% of total shipments. |
| SR004 | Hill Dickinson | Humanoid robots and the law: preparing for a new legal frontier | |
| SR005 | CnEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The robot has formally entered CATL's smart production line. It carries out long-running autonomous work in module and battery pack production, replacing manual labor in tasks such as material handling and picking. |
| SR006 | Mercator Institute for China Studies | Embodied AI: China's ambitious path to transform its robotics industry | China’s humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. They remain far from being fully autonomous machines capable of perceiving and responding in real time in the physical world. |
| SR007 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech's overall revenue jumped 53 percent to CNY2 billion (USD290.5 million) in the 12 months ended Dec. 31 from the previous year, while its net loss narrowed 32 percent to CNY789.8 million, the company noted. |
| SR008 | UBTECH Robotics via Hong Kong Exchanges and Clearing | Annual Results Announcement for the year ended December 31, 2025 | Credit impairment losses for the year are primarily due to delayed payments from certain government-related customers, and corresponding impairment was provided in 2025 on a prudent basis. |
| SV001 | Associated Press | China can build humanoids at scale. The hard part is finding enough buyers | "The economics are tough: humanoid robots remain expensive to produce, fragile in operation, and dependent on highly structured environments to function," Sacks explained. |
| SV002 | CNBC | China ships more humanoid robots than the U.S. as investors diverge on AI bets | That’s well above the [$3 billion-plus valuation](https://www.prnewswire.com/news-releases/galbot-secures-over-300-million-in-new-funding-breaking-records-with-3-billion-valuation-in-chinas-humanoid-robot-sector-302647204.html) of Chinese startup Galbot, which claims to be [the highest-valued privately-held Chinese company in the sector](https://mp.weixin.qq.com/s/jdQ2SuK1T2tYyvoo23FyVg). |
| SV003 | Robotics & Automation News | Robotics & Automation News homepage | As humanoid robot developers race to move their machines from demonstrations and pilot projects into factories and other working environments, another challenge is rapidly coming into focus |
| SV004 | Robozaps | Figure AI Review: Team, Investors & $39B | It is also nowhere near enough to validate a $39 billion valuation on operating results alone. |
| SV005 | CNEVPost | CATL teams with Galbot to scale humanoid robots in battery factories | The robot has formally entered CATL's smart production line. It carries out long-running autonomous work in module and battery pack production, replacing manual labor in tasks such as material handling and picking. |
| SV006 | Mercator Institute for China Studies | Embodied AI: China’s ambitious path to transform its robotics industry | China’s humanoids are cheaper than those of western competitors but still far too expensive for widespread deployment. For commercial viability, the costs would have to fall by at least half. |
| SV007 | Yicai Global | China’s Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million (USD119.2 million), or 41 percent of its total revenue, the Shenzhen-based company said in its annual financial report released yesterday. |
| SV008 | UBTECH Robotics Corp Ltd | Annual Results Announcement for the year ended December 31, 2025 | Our revenue increased by 53.3% to RMB2,001.0 million for the year ended December 31, 2025 from RMB1,305.4 million for the year ended December 31, 2024. |