Fourier Intelligence
康复商业化真实落地,但人形机器人上行空间仍跑在公开证据前面
Fourier 是可信的康复机器人公司,已有真实装机基础,也有技术上认真的人形机器人项目;但经济性、客户质量披露和具名付费人形部署都缺失,当前 RMB 8 billion 锚点难以支撑。按公开估值投入资本前,应继续研究。
封面要素
公司概况
Fourier Intelligence 对外将通用机器人产品线品牌化为 Fourier,是一家 2015 年由 Alex Gu 创立、总部在上海的机器人公司。公司从康复机器人起步,已拓展至全尺寸人形机器人;公开材料覆盖 Fourier Rehab 装机业务、RehabHub 工作流,以及 GR-1、GR-2、GR-3 和 N1 人形机器人产品线。公司公开材料称,Fourier 服务超过 40 个国家和地区的 2,000 多家机构;2024 年 GR-2 发布公告称,公司全球员工超过 500 人。资本基础可信:2022 年 Series D 融资 RMB 400 million,2025 年 Series E 融资近 RMB 800 million,Runyang 的战略投资披露则指向 RMB 8 billion 投前估值。但在公开层面,公司财务仍不透明。
- 成立时间
- 2015-01-01
- 创始人
- Alex Gu, Zen Koh
- 创立地点
- Shanghai, China
- 总部
- Pudong, Shanghai, China
- 产品
- Fourier 销售智能康复系统和 RehabHub 等工作流平台,同时推出 GRx 人形机器人家族及相关开发者工具,包括执行器、遥操作、仿真、数据集和机器人模型。公开证据在康复部署上最强,在付费人形机器人量产部署上明显更弱。
- 客户
- 康复服务机构、医院、神经康复中心、研究伙伴、渠道驱动诊所,以及部分企业或照护人形机器人试点。公开客户记录在康复领域最深,对具名付费人形机器人客户仍然单薄。
- 商业模式
- 以报价制机构机器人销售为主,覆盖硬件、工作流软件、服务和平台附加。Fourier 的公开材料支持“康复设备 + 服务”混合模式,但经常性收入占比、续约质量、实际成交价和单位经济仍未披露。
- 阶段
- Private (Series E)
- 融资情况
- 公开披露轮次包括 2022 年 1 月 RMB 400 million 的 Series D,以及 2025 年 1 月近 RMB 800 million 的 Series E。中文媒体还称公司成立以来累计融资约 RMB 1 billion;2025 年 6 月 Runyang 文件指向 RMB 8 billion 投前估值。完整股权结构表和融资条款未公开。
执行摘要
主要优势
- 康复业务商业化真实存在:覆盖 40+ 国家、2,000+ 家机构,对一家私人机器人公司来说是有意义的装机基础证明
- Fourier 不只是单个人形机器人 demo,产品和工具栈更宽,覆盖康复工作流、人形机器人、执行器、遥操作和仿真
- 融资通道可信:Series D、Series E 和 Runyang 披露显示投资者仍有兴趣,不只是宣传故事
- 公开客户证据在康复业务最强,多家具名服务商和医院描述了实际设备部署和工作流
主要风险
- 收入、毛利率、烧钱速度、现金余额和现金跑道仍未披露,投资逻辑无法靠经营基本面支撑
- 未找到 GR-1 或 GR-3 的公开可验证付费部署,人形机器人变现和 ROI 仍未证明
- 客户集中度、续约率、合同期限和渠道结构都不透明,因为许多可见关系是合作伙伴关系或经销商撮合的部署
- 护理人形机器人规模化会增加监管、隐私、安全和支持负担,但没有公开的 Fourier 专属医院级部署资料包
- 中国人形机器人竞争、出口管制摩擦和价格压力,可能在人形产品线证明持久需求前先压缩利润率
未决问题
- 经审计收入、毛利率、现金余额和现金跑道披露
- 具名付费人形机器人部署,以及 KPI、ROI 和续约证据
- 客户集中度、合同期限、GRR/NRR,以及经销商与直营结构
- 护理部署按型号拆分的监管、隐私和安全文件
- Series E 条款、投资者优先权,以及 RMB 8 billion 锚点背后的完整股权结构表
目录
01公司概况
1.1 身份、足迹与商业模式
Fourier 把自己呈现为全栈机器人公司,公开身份同时覆盖通用机器人品牌(“Fourier”)和聚焦康复的子品牌(“Fourier Rehab”)。官方 About 与首页材料称,公司成立于 2015 年,从外骨骼康复机器人起步,此后扩展出更广产品组合,包括 GRx 人形机器人系列、RehabHub、MetaMotus Galileo 系统和开源工具。公司联系页面把运营足迹锚定在上海、新加坡和马来西亚;多个官方页面则称 Fourier 服务 40 多个国家和地区的 2,000 多家机构。这个组合具有战略意义:Fourier 不是纯人形机器人创业公司,已经在康复和临床场景有装机基础,可用于验证硬件、收集数据,并交叉销售更新的人形机器人产品。公开层面,公司仍以产品叙事而非细颗粒商业指标来讲业务;所审阅来源均未披露收入、毛利率或年经常性收入(ARR)数字。[CO001, CO002, CO003, CO004, CO005, CO006]
| 指标 | 数值 / 状态 | 日期或期间 | 置信度 | 缺口 / 尽调备注 |
|---|---|---|---|---|
| 成立 | 2015 | 历史 | 高 | 成立日期由官方 About 页面和 2022 年融资报道交叉印证 |
| 总部 | 中国上海浦东 | 当前 | 高 | 官方联系页列出上海办公室地址;公开网站之外的法律实体细节仍有限 |
| 国际办公室 | 新加坡和马来西亚已公开列示 | 当前 | 高 | 官方联系页未发现欧洲或美国办公室地址 |
| 员工 | 全球 500+ | 2024-09 | 中 | 官方 GR-2 公告给出带日期的披露;未发现更新的官方员工数 |
| 机构 / 企业触达 | 2,000+ 家机构 | 当前 / 2025-2026 | 高 | 公司官方材料和 2026 年报道在核心触达数字上相互一致 |
| 地域覆盖 | 40+ 个国家和地区 | 当前 / 2025-2026 | 高 | 公司声称的装机基础数字;确切国家列表未公布 |
| 最近一轮融资 | Series E,接近 RMB800M | 2025-01 | 中 | 多家独立中文报道在规模和投资者阵容上相互一致 |
| 最新估值锚点 | 投前 RMB8B | 2025-05 | 中 | 锚点来自 Runyang 战略投资披露;公开投后估值未披露 |
| 累计融资 | 据报道约 RMB1B | 截至 2025 年 | 中 | 仅为中文媒体共识;未发现官方汇总股权结构表 |
| 公开收入 / 利润率 | 未披露 | 当前 | 高 | 已审阅公开来源均未披露收入、ARR 或毛利率 |
机构数、国家覆盖和员工数是公司报告或媒体报告的运营指标;估值是最新公开投前锚点,而非已披露的定价轮投后估值。
[CO001, CO002, CO006, CO007, CO014, CO015]Fourier 的康复装机基础、全球办公室、产品架构和资本结构如何串成当前经营逻辑。
[CO004, CO005, CO007, CO014, CO015, CO033]1.2 创始人、领导层与机构网络
公开领导层图谱以创始人 Alex Gu 为中心,他多次被标识为创始人兼 CEO(或创始人兼集团 CEO);2022 年 Series D 报道则称 Zen Koh 是联合创始人兼集团副 CEO。公开证据没有露出完整董事会或更广治理架构,因此公司概况只能依赖经营层,而不是投资人治理细节。可见的是 Fourier 围绕康复业务搭起的国际机构网络:2022 年融资报道提到 1,000 多家机构医疗客户和 17 个国际联合研究实验室;2025 年官方公告显示 Fourier Rehab 与佛罗里达的 Brooks Rehabilitation、法国的 Kurage 签署 MOU。这些合作强化了一个判断:Fourier 不只是实验室阶段机器人公司;它已经与医院、康复研究者和治疗系统合作伙伴一起工作。不过,已审阅的公开记录没有披露董事会席位、债务融资或老股交易,这仍是后续章节的重要尽调限制。[CO009, CO010, CO011, CO027, CO028, CO029]
| 人物 / 群体 | 公开角色 | 背景或可见度 | 职能覆盖 / 创始人-市场匹配 | 关键人物依赖 |
|---|---|---|---|---|
| Alex Gu | 创始人兼 CEO / 集团 CEO | 官方 About 页面列名创始人;GR-2 发布中以 CEO 身份被引用;长期深耕康复机器人 | 很高 — 推动康复根基、人形机器人转向、融资和公开叙事 | 关键 |
| Zen Koh | 联合创始人兼集团副 CEO | 2022 年 Series D 报道列为联合创始人和运营副手 | 高 — 联合创始人延续性和国际商业化领导力 | 高 |
| Owen Teoh | Fourier Rehab 国际事业部董事总经理 | 在 2025 年 Brooks 和 Kurage 合作公告中被引用 | 中 — 与全球合作伙伴绑定的可见国际康复业务负责人 | 中等 |
| 公开董事会构成 | 未公开披露 | 已审阅公开来源均未列名完整董事会、独立董事或观察员结构 | 公开可见度低 — 无法仅凭开源材料完整评估治理 | 重大尽调缺口 |
公司公布运营负责人,但不公布完整董事会或治理图,因此本表把具名高管和一条明确披露缺口行放在一起。
[CO009, CO010, CO011, CO037]1.3 资本基础、估值与披露
Fourier 的公开融资脉络已经足够清晰,可把它定位为成长阶段中国机器人公司,尽管股权结构表仍不透明。人形机器人浪潮前最强的已确认轮次是 2022 年 1 月 Series D:SoftBank Vision Fund 2 领投 RMB400 million,Prosperity7 Ventures 和 Yuanjing Capital 参投。三年后,36Kr、FL Cube、Sango Automation 和 NetEase 均报道 Fourier 在 2025 年 1 月完成近 RMB800 million 的 Series E,投资方包括 Guoxin Investment、Pudong Venture Capital、Zhangjiang 相关基金、Prosperity7 和 Junshan Capital。2025 年 5 月下旬,Runyang 计划最高 RMB300 million 的战略投资指向 RMB8 billion 投前估值,给市场提供了新的独角兽锚点。但披露质量仍是创投式,而非上市公司级:中文媒体报道累计融资约 RMB1 billion,但官方股权结构拆分、清算优先权、债务时间表或当前收入指标均未公开。[CO012, CO013, CO014, CO015, CO016, CO036]
| 利益相关方 | 角色 | 控制权或经济重要性 | 尽调问题 |
|---|---|---|---|
| SoftBank Vision Fund 2 | 2022 年 Series D 领投方 | 2022 年轮次中公开识别的最大投资者;为成长期机构背书提供强信号 | 确认当前持股、董事会权利,以及 2025 年是否行使跟投权 |
| Prosperity7 Ventures | 2022 年 Series D 投资者,2025 年 Series E 参与方 | 跨越康复到人形机器人转型的战略延续投资者 | 确认累计投入资本,以及与沙特医疗健康或工业的任何商业联系 |
| Yuanjing Capital | Series D 参与方 | 在人形机器人加速前支持成长期康复扩张 | 确认 Series E 后持股是否仍活跃 |
| Guoxin Investment / Pudong VC / Zhangjiang 基金 | Series E 中的国资相关 / 本地资本集团 | 为 2025 年融资团增加上海政策和产业生态支持 | 澄清该轮是否绑定任何政策里程碑、本地制造承诺或治理权 |
| Junshan Capital 与 PEAKVEST | Series E 参与方 | 2025 年轮次中的补充财务投资者 | 确认持股规模和投资条款 |
| Runyang Technology | 战略投资意向伙伴 | 计划最高投资 RMB300M,创造最新公开估值锚点和潜在材料协同 | 确认交易是否交割、价格如何,以及股权是否伴随供应协议 |
| Brooks Rehabilitation / Kurage | 战略生态伙伴 | 扩大临床验证和治疗栈集成,而不是纯融资 | 确认合作是否产生付费部署、联合 IP,还是只有研究 / 发布可见度 |
开源材料识别了轮次参与方和战略伙伴,但没有披露精确持股比例、清算优先权或董事会代表权。
[CO012, CO014, CO015, CO016, CO028, CO029]截至报告日期,Fourier 最清晰的公开指标和状态标记。
估值是投前披露锚点,不是完全披露的投后定价轮价值。
[CO006, CO014, CO015, CO017]1.4 产品路径与商业定位
Fourier 的公开里程碑显示,公司有意把康复机器人基础延伸到通用人形机器人。官方 About 页面把人形机器人项目启动放在 2019 年,把第一代 GR-1 原型里程碑放在 2022 年,正式推出 GR-1 则在 2023 年。官方 GRx 材料和 2024 年 GR-2 发布公告将 GR-1 定位为行业首款量产人形机器人,随后 GR-2 把平台升级到 175 cm、63 kg、最高 53 个自由度,并可选 12-DoF 双手、触觉传感和更友好的开发者控制栈。到 2025 年 8 月,Fourier 又用 GR-3 切换叙事:这是其首款明确面向照护的人形机器人,以更柔和的工业设计和多模态情感交互作为差异化。支持站点文档还显示,Fourier 发布面向开发者的机器人文档和 Fourier ActionNet 数据集,凸显其“made for AI”姿态。相较纯概念阶段人形机器人创业公司,这是有意义的差异化;不过,已审阅公开记录对单位经济和客户级部署规模仍只给出部分可见度。[CO018, CO019, CO020, CO021, CO022, CO023]
1.5 里程碑快照与反向框架
截至报告日期,Fourier 的公开形象支撑很强的动量叙事:公司已有十年历史、康复装机基础、2024 年披露 500 多名员工、2025 年新融资、新康复合作,以及 GR-3 在 CES 2026 展示。但同一组记录也要求保持谨慎。唯一公开估值锚点来自 Runyang 投前估值披露,而不是完整谈判后的定价轮投后估值;常被引用的 GR-1 “交付 100 多台”数字依赖中文二手报道,而非经审计公司台账;The Robot Report 的 2026 年中国展望还明确称,随着投资人要求商业上有意义的应用,竞争压力和整合会持续。换句话说,Fourier 因为起步于康复业务和机构网络,比许多人形机器人创业公司更可信;但它仍处在一个由商业化、成本压力和披露纪律决定高溢价机器人叙事能否转化为持久价值的市场。[CO015, CO017, CO025, CO026, CO031, CO036]
| 日期 | 事件 | 类型 | 金额 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2015-01-01 | Fourier 在上海成立 | 创立 | 公司成立 | Alex Gu | 确立公司的康复机器人起点 |
| 2017-01-01 | Fourier X1 下肢外骨骼在中国大陆商业化 | 产品 | 首个康复产品里程碑 | Fourier | 显示其在力反馈康复系统上的早期技术深度 |
| 2018-01-01 | 首个国际办公室在新加坡设立 | 规模扩张 | 新加坡办公室启动 | Fourier | 开启东南亚布局和国际商业化 |
| 2019-01-01 | 人形机器人项目启动 | 产品 | 人形机器人 R&D 启动 | Fourier | 标志着从垂直康复硬件转向通用机器人 |
| 2020-01-01 | RehabHub 发布 | 产品 | 集成康复平台推出 | Fourier | 强化超越单设备的平台叙事 |
| 2022-01-26 | Series D 融资宣布 | 融资 | RMB400M | SoftBank Vision Fund 2、Prosperity7、Yuanjing 等投资方 | 在人形机器人加速前为康复扩张和全球增长提供资金 |
| 2022-01-01 | 第一代 GR-1 里程碑达成 | 产品 | 原型代诞生 | Fourier | 把康复经验接入人形机器人开发 |
| 2023-01-01 | GR-1 作为量产人形机器人推出 | 产品 | 商业发布里程碑 | Fourier | 确立 Fourier 公开进入人形机器人领域 |
| 2024-09-26 | GR-2 发布 | 产品 | 53 DoF、12-DoF 手可选、披露 500+ 员工 | Fourier | 升级人形机器人线的开发者和工业定位 |
| 2025-01-07 | Series E 融资完成 | 融资 | 接近 RMB800M | Guoxin、PDVC、Zhangjiang 基金、Prosperity7、Junshan 等 | 为人形机器人推进补强资产负债表支持 |
| 2025-05-20 至 2025-06-03 | Kurage 和 Brooks MOU 公布 | 合作 | 战略康复合作 | Fourier Rehab、Kurage、Brooks | 在人形机器人扩张的同时,显示康复应用仍有深度 |
| 2025-08-06 | GR-3 照护导向人形机器人亮相 | 产品 | 全尺寸照护机器人发布 | Fourier | 把叙事转向具备情绪智能的服务机器人 |
| 2026-01-07 | GR-3 在 CES 国际展示 | 规模扩张 | 全球展示 | Fourier | 传递中国之外的国际市场野心和品牌建设 |
这条时间线覆盖与公司身份、融资、产品演进和国际化最相关的公开里程碑;它不是详尽的内部运营史。
[CO001, CO008, CO012, CO014, CO018, CO019]这条精选时间线梳理最能解释 Fourier 从康复机器人转向成长阶段人形机器人公司的里程碑。
[CO001, CO012, CO014, CO020, CO021, CO023]1.6 图表
02市场分析
2.1 市场边界与范围定义
Fourier Intelligence 的可服务市场位于两个结构上不同、但正在加速重叠的机器人板块交汇处。第一是康复机器人:动力外骨骼、治疗机器人和辅助步态设备,主要卖给康复中心、医院,并通过新的 Medicare 路径直接卖给家庭用户。第二是人形与具身 AI 机器人:拟人化、全身机器人,面向工业、医疗和消费终端市场。Fourier 通过 GRx 人形机器人平台和成熟的 Fourier Rehab 组合同时参与这两类市场。 Fourier 的主要总可用市场(TAM)不包括:手术机器人(报销路径和买方群体不同);非穿戴式固定康复设备(持续被动运动、超声治疗);以及覆盖物流、酒店和清洁机器人的更广义专业服务机器人市场(Fortune 数据为 2026 年 USD 31.11B,由运输和物流主导)。把广义服务机器人作为叠加 TAM 会误导判断——Fourier 当前并不覆盖物流或酒店应用。 International Federation of Robotics 正式把动力外骨骼归入医疗机器人分类下的可穿戴机器人,区别于工业机器人。康复和非侵入治疗机器人被单独跟踪为医疗机器人子类。这种定义精度很重要:IFR 医疗机器人品类 2024 年单位销量激增 91%,其中康复和非侵入治疗增长 106%,显示 Fourier 核心细分正在急剧加速。 机器人康复的现状替代品包括:持证物理治疗师和作业治疗师提供手工治疗;体重支持跑台系统(非机器人);以及标准矫形器具,包括非动力支具。从这些替代品迁移到机器人系统,取决于证据、报销覆盖、前期成本接受度和面向治疗师的工作流整合——这些结构性障碍也定义了 Fourier 可服务市场的采纳天花板。[CM001, CM002, CM003, CM008, CM044]
| 分析机构 / 来源 | 2026 年市场规模(USD B) | 复合年增长率(CAGR) | 预测终点年份 | 主要纳入范围 | 区域领先市场 |
|---|---|---|---|---|---|
| Fortune Business Insights | 0.58 | 16.36% | 2034 | 治疗机器人 + 外骨骼 + 辅助机器人 | 北美(2025 年约 75% 份额) |
| Mordor Intelligence | 1.77 | 17.10% | 2031 | 外骨骼 + 治疗机器人 + 服务 | 北美(39.6% 份额);亚太最快 |
| IMARC Group(2025 基准) | 1.82(2025 基准) | 16.53% | 2034 | 治疗机器人 + 外骨骼 + 辅助机器人 + 其他 | 北美;亚太增长中 |
| TBRC(Business Research Co.) | 2.22 | 22.70% | 2030 | 外骨骼 + 治疗机器人 + 相关服务 | 北美;亚太增长最快 |
所有数字都是出厂口径或分析机构估算的市场价值,以 USD B 计。未审阅底层方法论前,各估算无法调和;3.8x 差距(USD 0.58B– USD 2.22B)反映范围定义差异(是否纳入辅助机器人、假肢和打包服务)。IMARC 使用 2025 基准年。保守规划采用中段区间(USD 1.77B– USD 2.22B)。
[CM013, CM014, CM015, CM016, CM019]三层金字塔展示 Fourier 的可服务市场:从广义康复 + 医疗健康人形机器人 TAM,下沉到保守的 2026 年 SOM 估计。
TAM 由中档康复机器人市场(~USD 2B)与 FMI 人形机器人医疗健康细分(USD 10.69B 的 28% = ~USD 3B)相加,再扣除重叠(~USD 0.9B),约为 USD 4.1B;金字塔的 USD 12.91B 数值把 FMI + TBRC 人形机器人总市场中点全额纳入,作为最宽的合理 TAM 边界。SAM 仅采用 Mordor/TBRC 对康复机器人的中档估计(USD 2.0B)。考虑到 Fourier 披露的 2,000+ 家机构基础,SOM 是对 SAM 中 <10% 机构份额的定性估计。
[CM001, CM013, CM014, CM022]2.2 需求驱动:流行病学、人口结构与医疗系统压力
康复机器人最核心的结构性需求驱动,是全球神经和肌骨疾病流行,与老龄化人口叠加,推高越来越大的急性期后康复需求。World Stroke Organization 的 2025 Global Stroke Fact Sheet 基于 GBD 2021 数据,报告 2021 年全球现存卒中病例 93.8 million;卒中是全球第二大死因,也是死亡与残疾合并负担的第三大原因。1990 至 2021 年,卒中新发病例增长 70%;87% 的死亡和 89% 的 DALYs 集中在低收入及中低收入国家。卒中的全球经济成本每年超过 USD 890 billion,并预计到 2050 年接近翻倍。 PAHO 对美洲康复需求的估计显示,2019 年有 366 million 人存在适合康复介入的健康状况,较 1990 年的 231 million 增长 58%,主要由人口增长和老龄化推动。需要康复的概率随年龄呈指数上升。这为任何能扩大卒中和神经康复吞吐量的设备品类创造了长期需求顺风。 International Federation of Robotics 将医疗劳动力短缺和老年人口增长列为专业医疗服务机器人最主要的两个需求驱动。Oliver Wyman 从系统层面描述宏观逻辑:如果没有生产率干预,全球医疗支出可能从当前 USD 11.8 trillion 增至 2040 年 USD 23.1 trillion;在突破性采纳情景下,AI 和专用机器人每年最多可削减 USD 5.1 trillion 的预期成本增量。全球照护人员缺口估计达 15 million,进一步解释了医院系统为何评估用机器人增强治疗人员产能。上述流行病学和劳动力动态共同定义了康复机器人及相邻照护机器人的结构性扩张需求基础。[CM004, CM005, CM006, CM007, CM009, CM010]
康复机器人采购路径的每一阶段都会因证据、监管、报销、成本和工作流障碍而缩窄可服务人群。
阶段值是说明性百分比,代表 TAM 中大致通过各连续采用关口的份额;它们不是来自实证销售漏斗转化数据。系列数据为每个漏斗阶段提供定性障碍标签。
[CM034, CM036, CM038, CM039, CM040, CM041]2.3 核心市场:康复机器人规模与分层
四家独立分析机构发布了 2026 年康复机器人市场估计,区间很宽但有信息量。TBRC 估计市场为 USD 2.22 billion,2030 年增至 USD 4.87 billion,CAGR 22.7%。Mordor Intelligence 估计 2026 年 USD 1.77 billion,2031 年增至 USD 3.89 billion,CAGR 17.1%。IMARC Group 以 2025 年 USD 1.82 billion 市场为基数,预计 CAGR 16.53%。Fortune Business Insights 的估计明显更低:2026 年 USD 579.5 million,CAGR 16.36%;原因是范围更窄(仅含外骨骼和治疗设备,排除部分服务与实验室自动化品类)。不审阅底层方法,这些估计无法完全调和;USD 579.5M–USD 2.22B 的跨度(3.8×)反映了是否纳入辅助机器人、假肢和相关服务的定义差异。 中位共识——Mordor 的 USD 1.77B 到 TBRC 的 USD 2.22B——是 Fourier 康复业务用于 2026 规划最可辩护的 TAM。 在这个市场内,外骨骼机器人约占按类型划分收入的 47.35%(Mordor 2025 数据),确认其相对治疗和辅助设备子类的主导地位。老年用户约占需求的 61.4%。康复中心占终端用户收入的 53.2%;医院是第二大机构买方。居家照护是增速最快的采纳渠道,CAGR 27.1%,部分由 2024 年 CMS 外骨骼重新分类推动。亚太是增长最快的地区,CAGR 21.2%,由卒中发病率上升和人口老龄化驱动——这对总部在上海的 Fourier 是重要动态。按 Fortune 估计,北美仍保有最大区域份额,约占 2025 年收入的 74.76%。 Fourier Intelligence 被 TBRC 的康复机器人竞争格局明确列为关键竞争者,确认公司位于商业市场,而非研究阶段参与者。IFR 数据显示,2024 年单位级激增(康复机器人销量增长 106%)由临床场景专业终端用户推动,这与第 1 章所述 Fourier 2,000 多家机构客户基础一致。 相邻的老年照护机器人市场——包括陪伴机器人、移动服务机器人,以及专为养老机构设计的机器人外骨骼——2026 年规模为 USD 0.39 billion,2031 年增至 USD 0.64 billion,CAGR 10.4%(Knowledge Sourcing)。更广义的 AI 驱动老年照护解决方案市场(包含非机器人 AI)预计到 2031 年达到 USD 7.38 billion,CAGR 31.6%。这些相邻市场可由 Fourier 的 GR-3 照护人形机器人切入,但买方群体不同(养老机构、居家医疗机构),并不是支撑核心康复机器人可服务市场的康复中心和医院买方。[CM013, CM014, CM015, CM016, CM017, CM018]
| 来源 | 2026 年市场规模(USD B) | 复合年增长率(CAGR) | 终端年份 / 数值 | 医疗健康份额 / 备注 |
|---|---|---|---|---|
| Future Market Insights(FMI)研究 | 10.69 | 37.0% (2026–2036) | 到 2036 年 USD 248.9B | 医疗健康以 2026 年市场 28% 份额领先 |
| TBRC(Business Research Co.) | 8.32 | 52.9% (2026–2030) | 到 2030 年 USD 39B | 包含个人辅助 / 照护 |
| Roland Berger(OEM 口径,长期) | N/A (2026) | N/A | 到 2035 年 USD 300B–USD 750B | 运营成本 USD 2/hr;当前高产出用途很少 |
| BusinessWire / ResearchAndMarkets | N/A (2026) | N/A | 2026–2028 年拐点 | 成本同比下降 40%;当前单台 USD 30K–USD 150K |
FMI 和 TBRC 的估计口径和方法不同;二者都没有拆出 Fourier 可触达的医疗健康细分板块总可用市场(TAM)。Roland Berger 给的是战略层面预测,不是短期市场规模估算。制造成本和单价数据来自 BusinessWire/Goldman Sachs。人形机器人市场仍在快速成形,估计值只能作为方向性判断。
[CM022, CM023, CM024, CM025, CM028, CM029]分析师对 2026 年全球康复机器人市场的估计从 USD 0.58B 到 USD 2.22B;中档 USD 1.77B–2.22B 是规划时更相关的走廊。
区间由分析师给出的点估计按 ±10% 扩展,以反映常见抽样和方法不确定性;它们不是分析机构发布的置信区间。养老护理机器人只是单独的相邻市场参考。
[CM013, CM014, CM015, CM016, CM019, CM043]2.4 相邻市场:人形与具身 AI 机器人
人形机器人市场是 Fourier 更长期的增长向量,也是其最雄心勃勃的市场进入主张。两个大样本分析估计把 2026 年全球人形机器人市场分别放在 USD 10.69 billion(FMI)和 USD 8.32 billion(TBRC),未来五年 CAGR 分别为 37% 和 47–53%。医疗是领先应用场景;FMI 将 2026 年人形机器人收入的 28% 分配给医疗相关用例,意味着约 USD 3.0 billion 的可服务医疗人形机器人支出。 Roland Berger 2026 年 4 月报告描绘了更长周期:预计运营成本为每小时 USD 2,使人形机器人在面临结构性劳动力短缺的地区相对人力具备经济吸引力;到 2035 年,OEM 层面市场可能达到 USD 300 billion(基准)至 USD 750 billion(乐观)。TrendForce 预计中国人形机器人产出将在 2026 年增长 94%,Unitree 和 AgiBot 合计约占出货量 80%,行业正从演示转向交付。BusinessWire 引用 Goldman Sachs 数据称,人形机器人成本同比下降 40%,远快于此前模型中的 15–20% 降幅;当前机器人成本区间为 USD 30,000 到 USD 150,000。Unitree 的 R1 于 2025 年 7 月以 USD 5,900 发布,确立了低于 USD 10,000 的消费级价格点。 FMI 和 TBRC 估计相差约 USD 2.37 billion,反映范围边界、地理覆盖和假设时间点不同。两者也高于一些独立行业观点——Roland Berger 指出,今天除狭窄工业试点外,人形机器人几乎没有能真正产出的用例,生态系统(软件、集成、安全认证)落后硬件成熟度三到五年。 对 Fourier 而言,人形机器人市场里真正相关的可服务可获取市场(SAM)不是完整 USD 8–11 billion,而是 Fourier 的 GR-3 人形机器人所定位的医疗、康复和照护辅助子细分。按 FMI 28% 医疗份额套用 USD 10.69B 总量,意味着 2026 年医疗人形机器人 TAM 约 USD ~3B;但 Fourier 将面对 Figure、Agility、Unitree、Boston Dynamics 等全球平台领导者,它们资本和制造规模明显更强。GR-3 通过照护中心设计和临床装机基础形成差异化,这是一个独特角度,但近期市场份额渗透会受限。保守看,2027 年前 Fourier 在人形机器人市场的可服务份额远低于 5%。[CM022, CM023, CM024, CM025, CM026, CM027]
| 买方 / 支付方类型 | 角色 | 预算负责人 | 报销 / 支付路径 | 采用触发点 |
|---|---|---|---|---|
| 康复中心 | 主要机构买方 | 临床运营 / 财务委员会 | 按疗程计费代码;多数市场没有直接设备报销 | 证明在治疗师负荷更低时提升吞吐量,或疗效等同 |
| 医院门诊部门(OPDT) | 次要机构买方 | 医院资本开支 / 价值分析委员会 | OPPS/ASC 报销(2026 年更新 +2.6%);设备打包进 DRG/APC | 监管背书;同类医院采用 |
| 家庭用户(SCI / 卒中后) | 新兴直接买方 | 个人(Medicare 受益人) | Medicare 支具福利类别;CMS 支付约 USD 94,617;20% 共付 | 医生处方 + 完成认证站点培训 |
| VA / 工伤赔偿 | 补充支付方 | 政府 / 保险方项目预算 | 逐案报销决策;计入 Lifeward 约 40% 覆盖率数字 | 医疗必要性认定;FDA 许可文件在档 |
| 研究 / 学术机构 | 面向研发平台的小众买方 | 资助金 / 研究预算 | 无报销路径;直接采购 | 产出可发表数据;参与开发者生态 |
该表反映美国市场 2026 年状态;美国主导北美市场,约贡献全球收入的 75%。美国以外支付方覆盖数据不完整(见证据缺口)。截至 runDate,人形护理机器人(如 Fourier GR-3)在任何市场都没有成熟报销路径。
[CM030, CM031, CM032, CM033]矩阵把买方类型映射到预算来源、报销状态和采用准备度,适用于 2026 年康复机器人;图中标出增长最快的买方(通过 Medicare 的家庭用户)和最大买方(康复中心,53.2% 份额)。
[CM017, CM018, CM030, CM031, CM032, CM033]2.5 支付方、买方与预算负责人图谱
康复机器人采购分成三类买方和支付方原型,各自预算负责人、决策周期和采纳触发点不同。机构买方——康复中心和医院门诊部门——是主导渠道(按 Mordor 收入占市场 53.2%),由机构资本预算出资,采购决策通常由临床运营负责人和财务委员会作出。政府支付方和商业保险的报销通过锚定治疗课程计费代码,间接支持这些购买,并为设备 ROI 提供资金来源。 居家使用买方是增长最快但最小的细分。Medicare 的 2024 Home Health Rule(2024 年 1 月 1 日生效)将动力外骨骼重新划入支具福利类别,为个人外骨骼设备创造报销路径。CMS 拟定每台设备支付约 USD 94,617,Medicare 受益人承担标准 20% 自付。用户必须先在认证临床场地完成训练,才能居家使用。Lifeward 报告称,合并 Medicare、VA 福利和工伤赔偿后,美国脊髓损伤患者约 40% 现在由已经报销个人外骨骼申请的支付方覆盖,而 2024 年前几乎为零。 CY2026 OPPS/ASC final rule 将医院门诊支付费率上调 +2.6%,影响约 4,000 家医院门诊部门和 6,000 家流动手术中心。这一增量费率更新支持机构整体为康复技术采购提供资金的能力,但对机器人专项预算的边际影响是间接的。 研究和学术机构是开发平台与数据生成合作的小众买方渠道——与 Fourier 的开源和开发者生态策略相关,但对近期收入并不重要。 2026 年的人形机器人采购几乎完全由试点驱动,并由企业出资。BMW(Figure)、Amazon 和 GXO(Agility Digit),以及 UBTECH 的汽车 OEM 合作伙伴部署,都是企业 capex 决策,评估周期为一到三年。医疗人形机器人试点更早期;GR-3 这类照护人形机器人没有成熟报销路径,必须靠直接替代人工工时来创造 ROI。这把近期医疗系统机构采纳限制在创新或生产率预算线,而不是主流 capex。[CM030, CM031, CM032, CM033]
| 机器人子类别 | 2024 年销量 | 同比增长 | 主要应用 |
|---|---|---|---|
| 医疗机器人(总计) | ~16,700 | +91% | 所有医疗应用合计 |
| 康复与非侵入式治疗 | ~8,100(估计) | +106% | 步态康复、外骨骼治疗、上肢设备 |
| 手术机器人 | 总量子集 | +41% | 微创手术辅助 |
| 诊断与医学实验室 | 总量子集 | +610% | 实验室自动化、诊断分析 |
IFR World Robotics 2025 Service Robots 使用样本数据(294 家供应商),没有外推到全行业;同比比较只在样本内成立。康复 / 非侵入式治疗销量是根据披露的 106% 增长指数和相对权重估算;IFR 没有在新闻稿中公布各子类别精确销量。手术和诊断销量只被定性描述为 16,700 总量中的子集。
[CM008]2.6 约束视角:逆风与采纳障碍
康复和人形机器人的乐观逻辑建立在人口结构顺风、单位成本下降和政策催化之上。负责任的市场分析必须把这些与实质约束力量一起加权。 临床证据方面:2026 年 JMIR 关于卒中后上肢康复机器人辅助治疗的伞状综述发现运动功能改善,但日常生活活动(ADL)提升通常与常规治疗相当。该发现直接限制了最强支付方审批论点——机器人辅助治疗能产生显著更好的功能结局。2026 年 PubMed RCT 确实发现,对卒中后步态康复,外骨骼辅助康复相较常规训练改善了 Fugl-Meyer 下肢评分和 Barthel Index,比上肢 ADL 更清楚地支持下肢 / 步态康复场景。因此,证据图景是混合且分层的:下肢 / 步态外骨骼用例更强,广义 ADL 结果主张更弱。 人形机器人的商业就绪度方面:IFR 2026 年趋势报告明确表示,人形机器人必须在节拍时间、能耗和维护成本上满足高工业要求,才能与传统自动化竞争;截至 2026 年,关键缺口仍在。Robotics & Automation News 将操作能力、电力与续航、通用自主性识别为三大技术瓶颈;大多数 2026 年部署仍是受监督试点或协同操作应用,而非自主运行。Roland Berger 也确认,生态系统(软件集成、安全认证框架、监管清晰度)落后硬件三到五年。 成本障碍方面:售价 USD 30,000–USD 150,000 的康复机器人设备,其采购上限由报销费率(Medicare 路径约 USD 94,617)、机构资本预算和机构风险容忍度共同决定。长期结果数据混合叠加高前期成本,会压制新兴市场和儿科场景采购。Fortune Business Insights 和 Mordor Intelligence 都把高初始资本开支(capex) 和集成成本列为主要市场约束。隐私和数据安全担忧,加上照护场景中对替代人际连接的焦虑,则构成 AI 机器人老年照护采纳的软性障碍。 综上,这些约束说明:虽然市场规模估计很大、增速也有吸引力,但在 2026–2028 年窗口内,标称 TAM 中真正可触达的比例——尤其对 Fourier 这样的中型玩家——明显小于标题数字暗示。[CM034, CM035, CM036, CM037, CM038, CM039]
2.7 图表
03竞争格局
3.1 Fourier 的竞争图谱是双线分叉,不是单一市场
Fourier 的公开产品表面清楚显示,它试图桥接两个市场,而不是防守一个狭窄利基。公司仍通过 Fourier Rehab 销售密集康复栈——覆盖 40+ 国家 2,000 多家机构、整合 30+ 种机器人类型的 RehabHub 层,以及多种肢体专用系统——同时又把 GR-1 作为量产人形助手、把 GR-3 作为面向照护的社交机器人来营销。这种双重身份意味着直接竞争集合拆成两组,且证明要求不同。康复侧,相关公开同行是 Wandercraft、Ekso Bionics、CYBERDYNE 和 Lifeward,它们围绕临床证据、步态训练工作流,有时也围绕报销准入竞争。人形机器人侧,Fourier 遇到 UBTECH、Unitree、AgiBot、Figure 和 1X;公开讨论集中在出货规模、价格底线、产能和试点部署。 实际含义是,不能拿 Fourier 对标单一 “robotics” 平均值。医院和康复买方关心治疗师效率、监管状态、设备是否适配报销或照护流程。工厂、物流、研究和早期家庭买方更看重自主性、操作能力、硬件可靠性和标价。这让 Fourier 的照护人形机器人策略有意思,也很脆弱:它继承了比工业优先人形机器人同行更强的照护相邻性,也继承了比消费或工厂机器人供应商更慢的医疗式证据标准。[CP001, CP002, CP003, CP004, CP006, CP007]
| 公司 | 核心范围 | 主要买方 | 部署成熟度 | 规模 / 融资信号 | 公开价格信号 | 竞争判断 |
|---|---|---|---|---|---|---|
| Fourier Intelligence | 康复机器人 + GR-1/GR-3 人形机器人 | 医院、康复中心、学术机构、护理试点 | 康复已商业化;人形机器人早期商业化 | >2,000 家机构,覆盖 40+ 个国家;RehabHub 中 30+ 台机器人 | 未公开列价 | 护理渠道衔接最好;同业中公开人形机器人规模证明最弱 |
| Wandercraft | Atalante X 康复外骨骼、Eve 家用外骨骼、Calvin-40 人形机器人 | 当前是康复医院;下一步面向家庭和工厂买方 | Atalante 已商业化;Eve 处于关键试验;Calvin 投产前 | €64.3M D 轮;100+ 家医院;Renault 是投资方 / 客户 | 未公开列价 | 在康复和人形机器人扩张两条线上与 Fourier 重叠最接近 |
| Ekso Bionics | EksoNR、Indego Therapy、Indego Personal 等产品 | 康复医院、门诊中心、Medicare/VA 患者 | FDA 许可的临床 / 家用路径 | 美国上市医疗科技公司;声称有 200+ 篇已发表文章 | Medicare 挂钩的公开报销锚点;标价未披露 | 直接康复同业中,美国证据和准入姿态最强 |
| CYBERDYNE | HAL 医疗、健康、居家和劳动支持系统 | 医院、诊所、受监督居家康复、劳动支持 | 长期商业化的小众市场 | 上市公司投资者关系与创始人控制治理 | 未公开列价 | 生物信号控制有差异化;近期全球规模披露较薄 |
| Lifeward | ReWalk Personal、ReStore、AlterG、MYOLYN 等产品 | SCI 患者、临床医生、退伍军人、医院 | 临床与居家的连续护理部署 | Nasdaq 上市;业务覆盖美国、以色列、德国 | Medicare 报销路径公开;标价只能间接推断 | 在美国,家用报销路径比 Fourier 更清晰 |
| UBTECH | Walker S 人形机器人 + 养老护理 / 服务组合 | 汽车、物流、商业服务、养老护理项目 | 规模交付已启动 | HKEX 上市;2025 年初以来订单 >RMB 800M | 未公开列价 | 中国工业规模化威胁明显,但公开临床证据有限 |
| Unitree | G1/R1/H1 人形机器人 | 开发者、学校、实验室、早期采用者、部分工业客户 | 已有商业出货,产能爬坡 | 人形机器人占 2025 年收入 >51%;年产 75,000 台人形机器人目标 | G1 起价 US$13.5K;R1 据报 US$5.9K | 价格领先;直接医疗专业化最弱 |
| AgiBot | A2/X/G/D 具身机器人组合 | 工业、服务、酒店 / 接待、研究、物流 | 声称已规模部署;第三方验证仍不均衡 | 声称已出货 5,000 台机器人;据报达到第 10,000 台机器人里程碑 | 未公开列价 | 中国快速放量者;覆盖范围宽,但护理专项证据很少 |
| 1X | NEO Gamma 家用人形机器人 | 家庭与居家测试项目 | 内部居家测试 / 家庭使用推进 | 私有公司;有官方投资者关系页面,但公开规模细节有限 | 保留来源中无公开标价 | 家庭安全叙事成立,但公开部署证据远少于 Figure 或 UBTECH |
| Figure | Figure 03 通用人形机器人 | 家庭帮手愿景 + 工业伙伴 | 工业试点有强伙伴验证,更大范围铺开仍在爬坡 | >US$1B C 轮,投后估值 US$39B(据官方新闻索引) | 无公开标价 | 西方同业中工业证据最可信,但没有公开医疗切入点 |
由于同业披露指标不一,规模 / 融资信号混合了上市状态、融资、订单簿、部署数量和已披露产能。公开价格信号只使用保留公开来源支持的标价、报销参考或明确未披露信息。
[CP001, CP002, CP003, CP007, CP009, CP010]基于证据的序数定位,比较 Fourier 和九个直接竞争对手的两个轴:部署成熟度(1 = 早期试点 / 研究,5 = 规模化商业交付)和照护 / 临床契合度(1 = 公开照护证据很少,5 = 照护或康复是核心产品身份)。Fourier 在照护契合度上高于工业人形机器人同行,但在报销上低于最强康复既有玩家,在公开规模证据上低于 UBTECH/Unitree/AgiBot/Figure。
评分是序数估计,来自公开产品页面、合作伙伴证据、分析师评论和部署公告。x = 部署成熟度,y = 照护 / 临床契合度。该图用于相对定位,不是经审计的基准测试。
[CP007, CP020, CP022, CP024, CP025, CP029]3.2 康复竞争者胜在证据、报销或专门步态工作流
Fourier 当前最强的直接竞争仍在康复机器人,而不是通用人形机器人。Wandercraft 是战略重叠最接近者,因为它现在同时覆盖医院步态训练、通过 Eve 进入居家外骨骼的野心,以及经 Calvin-40 和 Renault 延伸到工业人形机器人。Ekso Bionics 和 Lifeward 是更直接的康复对手:两者都围绕个人外骨骼准入给出明确美国报销信号,也把产品覆盖到临床和家庭或社区场景。CYBERDYNE 在技术上更差异化,使用生物电信号控制,并提供医疗、健康管理和劳动支持变体,但其英文公开材料对近期全球商业规模的透明度较低。 Fourier 在这一组里的优势,是机构康复分销广度。官方材料显示,它拥有比任何单一外骨骼同行更大的多机器人康复平台,应该有助于治疗师关系、打包销售和交叉销售。弱点在于,美国可见的报销和公开支付方证据,目前 Lifeward 和 Ekso 比 Fourier 更清楚。公开价格透明度也薄:除了与 Medicare 相关的个人外骨骼报销参考,大多数供应商仍是报价制。这种不透明降低了纯价格性能比较的置信度,并推动买方转向供应商信任、证据和服务足迹。切换成本存在——临床医生培训、流程重设计、软件工作流和支付方熟悉度——但它们不是手术机器人那种硬平台锁定,因此没有哪个康复在位者不可攻破。[CP008, CP009, CP010, CP011, CP012, CP013]
| 采购标准 | Fourier | 康复龙头(Wandercraft / Ekso / Lifeward / CYBERDYNE) | 人形机器人龙头(UBTECH / Unitree / AgiBot / Figure / 1X) | 证据支持的推论 |
|---|---|---|---|---|
| 临床工作流适配 | 高 — 康复组合和数字化 RehabHub 是核心产品 | 高 — 步态康复和神经康复是核心 | 低到中 — 多数是工业、研究或家庭叙事 | Fourier 最大优势是既有治疗师 / 渠道契合度,不是单纯人形机器人规模 |
| 家用报销路径 | 低 — 保留来源中没有美国公开路径 | 中到高 — Lifeward 和 Ekso 有明确 Medicare 挂钩证明;Wandercraft 和 CYBERDYNE 不太清晰 | 低 — 人形机器人没有成熟报销路径 | 报销仍偏向既有康复外骨骼厂商,而不是人形机器人新进入者 |
| 护理 / 养老互动 | 高 — GR-3 明确针对护理互动优化 | 中 — 靠近康复场景,但产品不那么强调社交表达 | 中 — UBTECH 和 1X 谈护理 / 家庭;公开证据仍少 | 相比工厂优先的人形机器人厂商,Fourier 的护理叙事更清晰 |
| 工业操作 / 工厂证明 | 低 — 没有同等公开产线证明 | 低 — 以康复平台为主 | 高 — UBTECH、AgiBot、Figure 和 Unitree 披露工厂、数据采集或开发者部署 | 工业证明目前不在 Fourier 最强赛道内 |
| 公开价格透明度 | 低 — 报价驱动 | 低到中 — Ekso/Lifeward 个人系统有 Medicare 参考价 | Unitree 高,多数其他厂商低 | Unitree 在重设入门价格预期,多数同业仍不透明 |
| 分销与服务覆盖 | 康复机构内高;公开可证实的人形机器人规模低 | 中到高,取决于地区和报销 | UBTECH/Unitree/AgiBot 在中国工业端高;Figure 中等;1X 目前低 | Fourier 的分销优势在临床端,还不是通用人形机器人 |
单元格只概括公开证据。“低”通常意味着商业证明有限或没有可比公开披露,并不证明技术劣势。缺乏支持的实际价格或 ROI 单元格保留为不可比叙述,不做估算。
[CP003, CP006, CP008, CP011, CP013, CP014]3.3 人形机器人对手用价格、销量或工业验证重置市场,而不是临床适配
在人形机器人上,Fourier 不是规模领导者,也不是价格领导者。UBTECH 已把自己呈现为上市的全栈人形机器人公司,具备量产、汽车合作伙伴,以及对养老和商业服务的明确敞口。Unitree 是公开价格破坏者:G1 在其官网起价 US$13.5K,外部市场摘要则指向 R1 为 US$5.9K,远低于机构康复硬件所隐含的定价逻辑。TrendForce 的 2026 年观点认为 Unitree 和 AgiBot 合计可能拿下中国出货近 80%,凸显中国厂商正在多快把这个品类工业化。AgiBot 自身材料和 CES 新闻稿主张显示,其产品家族很宽,并已出货数千台机器人,尽管独立验证仍不均衡。 西方同行姿态不同。Figure 首页现在为 Figure 03 销售居家帮助叙事,但公开记录里最强的证明仍是 BMW:日常班次、装载 90,000+ 个零件,并影响 30,000+ 辆车,场景在汽车制造。1X 更明确地优先家庭,强调内部居家测试、更安静硬件、被动安全和遥操作。这意味着 Fourier 的 GR-1 和 GR-3 差异化不在原始自主性或制造规模,而在照护相邻性。GR-3 的软壳、强调触感的设计,以及 GR-1 的类人助手框架,比大多数工业同行更适合照护或社会陪伴叙事。但这也是 Fourier 的弱点:GR 系列规模化部署、公开价格和量化 ROI 的公开证据,明显薄于 Unitree 的价格底线、UBTECH 的订单簿或 Figure 经合作伙伴验证的生产学习。[CP020, CP021, CP022, CP023, CP024, CP025]
| 公司 | 代表性产品 | 公开价格或报销锚点 | 合同模式 | 公开包含内容 | 含义 |
|---|---|---|---|---|---|
| Fourier | 康复组合 / GR 人形机器人 | 未公开列价 | 机构报价 / 联系销售 | 披露了产品组合宽度和护理互动,但没有实际 ASP 或服务包 | 仅靠公开数据很难比较性价比 |
| Wandercraft | Atalante X / Eve | 未公开列价 | 机构报价;家用产品尚未大范围商业化 | 临床康复 + 未来家庭和工业延伸 | 竞争更多靠功能和融资势头,而不是公开价格 |
| Ekso | Indego Personal / Therapy;EksoNR | 据报 Medicare 挂钩个人外骨骼费率为 US$91,032 | 临床销售 + 个人使用报销路径 | 公开描述临床和家庭 / 社区用例,但准确标价未披露 | 在康复组中,Ekso 的美国支付锚点最清楚 |
| CYBERDYNE | HAL 医疗 / 健康 / 劳动支持 | 未公开列价 | 机构报价 / 服务项目 | 多个 HAL 版本和居家服务,但没有公开可比 ASP | 差异化技术没有转化为公开定价清晰度 |
| Lifeward | ReWalk Personal / ReStore / AlterG | 据报 Medicare 挂钩个人外骨骼费率为 US$91,032 | 个人报销 + 机构销售 | 公开描述家庭 / 社区和诊所使用 | Lifeward 能卖准入确定性,不只是硬件 |
| UBTECH | Walker S2 工业人形机器人 | 无公开标价 | 项目 / 解决方案销售 | 批量交付、培训和场景部署支持 | 竞争靠规模和运营能力,而不是公开标价 |
| Unitree | G1 人形机器人 | 官方起价 US$13.5K | 直接销售 / 开发者级配置 | 机器人硬件,可选计算和灵巧手升级 | 给通用人形机器人设下激进低端价格锚 |
| Unitree | R1 人形机器人 | 分析师摘要称 US$5.9K | 消费级 / 入门级销售 | 市场叙事中的低成本人形机器人价格地板 | 即使能力不同,也压迫高端人形机器人定位 |
| AgiBot | A2/X/G 组合 | 无公开标价 | 平台与定制化部署 | 多场景组合,硬件 / 软件可配置 | 覆盖面宽,但性价比仍由报价决定 |
| 1X / Figure | NEO Gamma / Figure 03 | 保留官方来源中无公开标价 | 等候名单、试点或伙伴部署模式 | 有家庭帮手和工业叙事,但公开商务条款缺失 | 仅靠公开 ASP 仍难以对西方同业做投资判断 |
机器人行业公开价格数据异常稀薄。因此,该表混合了标价、报销锚点和明确未披露信息。应把它当作透明度地图,而不是严格可比的 ASP 表。
[CP013, CP020, CP022, CP024, CP025, CP028]定性热力图结合面向买方的能力与商业化证据。Fourier 的优势在临床分销和照护 UX;Unitree 领先于公开价格锚点,UBTECH 与 Figure 领先于工业证据,Lifeward / Ekso 领先于公开报销可见度。该图有意比 TP002 的买方标准表更偏商业化权重。
评级仅来自保留的公开来源。高 = 明确且反复出现的公开证据,中 = 已公告或间接证据,低 = 公开证据有限,未知 = 保留材料中没有可比公开披露。
[CP006, CP011, CP013, CP014, CP017, CP020]3.4 Fourier 的护城河是分销和照护适配,但商品化压力正在快速上升
Fourier 最可防守的差异点,不是其人形机器人今天能力最强;而是 Fourier 已经是少数嵌入康复服务机构工作流的人形机器人进入者之一。这一点重要,因为医院和康复采购仍然更依赖信任、培训、服务和临床舒适度,而不是炫技。把照护导向人形机器人叠加到现有康复渠道上,比通用消费或工厂人形机器人突然转向养老照护,更像一个可信的商业化楔子。这也是 Fourier 双重身份的战略价值:康复业务可以播下关系、数据集和渠道准入,这些是纯人形机器人厂商尚未拥有的。 这条边的耐久性仍有限。公开竞争者证据表明,人形机器人品类仍处在试点阶段且受安全约束;灵巧操作、续航、触觉感知和最低风险行为,在真实环境中仍未解决。这些限制帮到 Fourier,因为它们拖慢了所有人。但它们没有解决更大的商业风险:一旦人形机器人硬件更便宜、更模块化,照护相邻软件、服务和临床工作流整合,可能比真正的报销护城河更容易复制。手工治疗、传统辅助设备、内部自动化项目,以及 Wandercraft 的 Calvin-40 或中国工业平台等相邻进入者,仍是可信替代品或未来威胁。因此,近期判断需要细分:Fourier 拥有比多数人形机器人同行更强的照护叙事,但公开展示的规模弱于对手,对美国报销的证明也弱于最佳康复在位者。[CP036, CP037, CP038, CP039, CP040, CP041]
| 护城河主张 | 支撑证据 | 威胁 | 严重性 | 缓释措施 / 尽调问题 |
|---|---|---|---|---|
| 面向康复服务方的临床分销 | Fourier 声称覆盖 40+ 个国家 2,000+ 家机构,RehabHub 中有 30+ 台机器人 | 一些市场里,康复既有厂商已经掌握报销和治疗师信任 | 高 | 获取队列级留存、附加率,以及相对 Ekso、Lifeward、Wandercraft 的赢单 / 输单数据 |
| 适合护理的人形机器人设计 | GR-3 明确为护理互动打造,具备柔软外壳和触觉反馈 | 硬件商品化后,工业优先厂商可以加上更柔性的 HRI 层 | 中 | 要求提供养老护理或医院场景试点结果,并衡量社交设计是否改变采用率或 ROI |
| 从康复向护理人形机器人交叉销售 | Fourier 已经卖进康复科室,多数人形机器人同业还没有 | 医院可能仍会购买单独的专用工具,而不是宽平台 | 中 | 索取康复客户群转化为 GR 系列试点和付费部署的数据 |
| 抵御低成本人形机器人竞争 | 相比消费级或开发者机器人,Fourier 可以主张临床适配度 | Unitree 可见价格地板和中国规模会压低买方预期 | 高 | 需要清晰 ROI 逻辑、服务包定义和买方分层,才能支撑溢价定位 |
| 抵御西方资金充足的人形机器人新进入者 | Figure 和 1X 在护理上更早期,但能大举投入自主能力和 HRI | Figure 已有伙伴验证的工业部署和庞大资本基础 | 高 | 跟踪 Figure、1X 或 UBTECH 是否会在 Fourier 跑通 GR 经济性前进入医疗专项试点 |
| 人形机器人市场时点风险 | IFR 和 Robotics & Automation News 都提示炒作、安全和自主能力约束 | 如果品类更久停留在试点阶段,Fourier 的 GR 产品可能在需求成熟前消耗资源 | 中 | 用已签试点、安全 / 合规里程碑和付费服务采用率对未来 GR 投资设阶段门 |
严重性是基于披露部署规模、报销杠杆和可见价格锚点的定性评估。由于公开证据仍不完整,缓释措施以尽调问题呈现。
[CP036, CP037, CP038, CP039, CP040, CP041]六个公开指标最能解释截至 runDate 的 Fourier 竞争姿态:一个明确的康复渠道优势、两个由竞争对手掌握的报销和价格锚点,以及三个有利于工业人形机器人同行的规模信号。
KPI 组合有意混合分销、报销、定价、订单积压和部署运行时长,因为没有单一公开指标能把这个赛道标准化。数值是当前公开信号,不是经审计的市场份额。
[CP002, CP003, CP013, CP024, CP028, CP034]3.5 图表
04财务情况
4.1 资本结构已足够可见,可锚定估值;但还不足以承销现金生成
Fourier 的公开融资轨迹现在给出了可信的资本地图,尽管公司仍未披露经营财务。2022 年 Series D 从 SoftBank Vision Fund 2、Prosperity7 和 Yuanjing 引入 RMB400 million;管理层称资金将加速商业化拓展、医疗机器人创新和全球增长。2025 年 1 月 Series E 据报道接近 RMB800 million,并明确绑定进一步人形机器人开发和更广商业部署。信息量最大的文件不是 Fourier 新闻稿,而是 Runyang Technology 2025 年 6 月文件,因为它增加了 RMB8 billion 投前估值锚点,显示计划出资 RMB300 million,换取稀释后 3.5060% 所有权,并披露同步 3.0% ESOP 增补。该文件还显示投资人化董事会结构,创始人保留 11 个提名权中的 6 个。这些都是强资本结构信号,但不能替代资产负债表:公开记录仍更多说明谁给 Fourier 投了钱,而不是现有业务赚多少、烧多少或留存多少。[CI001, CI002, CI003, CI005, CI006, CI007]
| 项目 | 公开证据 | 财务含义 | 置信度 | 确切尽调索取项 |
|---|---|---|---|---|
| 2022 年 D 轮融资 | RMB400m,由 SoftBank Vision Fund 2 领投,Prosperity7 和 Yuanjing 参投 | 在人形机器人推进加速前,提供了第一笔清晰可见的后期成长资本 | 高 | 确认到 2025 年初仍剩多少资金 |
| 2025 年 E 轮融资 | 近 RMB800m,投资方包括国资相关和战略成长型投资人 | 为人形机器人和更大范围商业化实质性补强了资产负债表 | 高 | 需要交割日期、投后股权,以及募集资金是否按业务线隔离 |
| Runyang 战略投资锚点 | 计划最高 RMB300m,投前估值 RMB8bn | 增加一个新的独角兽式估值锚点,但没有披露现金余额或交割后现金跑道 | 高 | 确认交易是否交割,以及实际到账现金 |
| 同步发行 ESOP | 在 Runyang 持有 3.5060% 股权的同时,新增 3.0% 完全稀释后 ESOP | 表明公司扩张时仍在为留住人才继续承受稀释 | 高 | 需要交易前后的完全稀释股权结构表 |
| 治理信号 | 交易后,创始人提名 11 名董事中的 6 名 | 创始人保留控制权,但治理仍更像风险投资支持公司,而非上市公司式透明 | 中 | 需要股东协议、否决权和董事会观察员名单 |
| 公开在手现金 / 现金跑道 | 未披露 | 外部读者无法判断当前资本能覆盖 12、18 还是 24 个月扩张 | 高 | 需要当前现金、烧钱速度、现金跑道,以及基准 / 乐观 / 悲观计划 |
| 债务或项目融资义务 | 已审阅公开来源未披露 | 公开来源无法排除潜在杠杆、供应商融资或担保风险 | 中 | 需要债务明细、约束条款包,以及任何表外义务 |
融资时间线是公开的,但实际流动性位置并不公开。本表把可见资本事件与仍缺失的现金和负债图景分开。
[CI001, CI003, CI005, CI007, CI009, CI010]业务不同部分的可见度、服务负担和融资依赖各不相同,但公开模型都不足以推断自筹资金规模化。
该矩阵是方向性、由证据支撑的判断,不是打分热力图。它把可见度和经济吸引力分开,因为在 Fourier 案例中二者是不同问题。
[CI009, CI010, CI011, CI019, CI025, CI027]4.2 可见商业模式是报价制机构机器人,公开价格只有碎片
官方材料显示,Fourier 是混合机器人栈,而不是单一产品 SKU。Fourier 将 RehabHub 定位为整合 30 多种机器人类型的平台,通过留资联系页面销售康复设备,并把 GR-3 描述为面向企业场景、以照护为中心的人形机器人,而非零售买家产品。这一产品呈现意味着至少三层经济:资本性设备、实施或服务,以及某种软件或平台附加;但公司从未披露这些层之间的实际收入结构。公开可查价格同样碎片化。TMTPost 称 GR-3 定价高于 RMB200,000;当前 Europa Satellite 挂牌页则暴露了 EUR106,344.61 的结构化产品元数据。康复侧,一个中国采购追踪器显示,上肢机器人具体中标价为 RMB280,000 和 RMB347,000,更大的康复系统中标价超过 RMB1 million。这些信号足以说明 Fourier 卖进机构预算,而不是消费冲动购买。它们不足以推断实际 ASP、折扣、附加率或利润率,因为每个可见价格点都反映不同套装、地理或渠道。[CI013, CI014, CI015, CI016, CI017, CI018]
| 收入流 | 公开变现证据 | 公开价格或预算锚点 | 收入质量判断 | 具体尽调问题 |
|---|---|---|---|---|
| 康复资本设备 | 官方康复页面和医院采购记录显示,公司向机构护理场景销售设备和系统 | 单台上肢机器人 RMB280k-RMB347k;较大型康复系统 RMB1.07m-RMB1.368m | 企业硬件收入可能较不均衡,受招标节奏和打包差异影响 | 需要按康复 SKU 拆分的实际 ASP、安装收入和产品级毛利率 |
| 人形机器人企业系统 | GR-3 通过联系销售的企业页面和第三方价格信号营销,而不是 Fourier 开放公开购买入口 | TMTPost 称 >RMB200k;Europa 结构化数据列出 EUR106,344.61 | 早期企业人形机器人收入可能由报价驱动,且对试点敏感 | 需要已成交出货量、实际售价,以及价格是否包含服务还是仅硬件 |
| 实施、培训与维护 | 同业外骨骼部署需要 40-50 小时培训和持续支持,意味着既有服务收入也有服务成本 | 没有 Fourier 直接价格;同业显示支持强度的经济影响不小 | 服务可能扩大合同价值,但定价过低也会吃掉毛利 | 需要每套已安装系统的培训附加率、服务合同附加率和年度维护收入 |
| 平台或软件层 | RehabHub 和开发者材料暗示存在数字化编排,以及数据或软件附加销售 | 未发现公开的独立软件价格 | 可能存在质量更高的经常性收入,但收入占比未披露 | 需要软件许可、订阅或数据平台收入占比 |
| 报销驱动准入类比 | 同业外骨骼报销表明,支付方支持可以释放超高价设备的支付意愿 | 既定 Medicare 费率 US$91,032,以及约 US$100k 的设备成本类比 | 准入顺风确实存在,但需逐案判断,不能替代 Fourier 专属毛利披露 | 需要按市场确认报销路径,并确认 Fourier 哪些产品符合条件(如有) |
本表混用 Fourier 自身证据、支付方和同业参照,因为 Fourier 披露了产品和渠道,但没有披露实际收入结构、附加率或经常性软件占比。
[CI013, CI014, CI015, CI017, CI018, CI037]| 产品或基准 | 公开价格信号 | 范围或渠道 | 尚未说明什么 | 为什么重要 |
|---|---|---|---|---|
| Fourier GR-3(公司披露的媒体信号) | 超过 RMB200,000 | TMTPost 关于 B2B 医疗 / 企业目标客群的报道 | 没有定义套餐、服务包含项或实际折扣 | 表明 Fourier 并未把 GR-3 放在低端消费级人形机器人价格带 |
| Fourier GR-3(分销商元数据) | EUR106,344.61 | Europa Satellite 结构化产品列表,有效期至 2026-06-20 | 第三方列表可能包含经销商利润和物流假设 | 确认欧洲存在六位数的外部企业级价格锚点 |
| Fourier M2D02SP 上肢康复机器人 | RMB280,000-RMB347,000 | 120Shop 上的医院采购记录 | 不清楚哪些招标包含配件、培训或本地化 | 说明较小的康复单元仍能占用可观机构预算 |
| Fourier 大型康复系统中标 | RMB1.07m-RMB1.368m;跟踪均值 RMB1.28m | 覆盖更大康复系统的医院采购跟踪 | 产品包差异很大,削弱了同口径 ASP 推断 | 暗示完整康复部署可达到人民币七位数采购额 |
| Unitree G1 人形机器人 | US$13.5K 起 | Unitree 官方直销价格 | 能力、用例和支持包都不同于 Fourier | 形成可见的低端人形机器人价格下限,可能锚定市场预期 |
| 同业外骨骼报销 / 价格锚点 | US$91,032 报销费率;约 US$100k 设备成本参照 | Medicare 加患者 / 服务方报道 | 适用于特定 FDA 获批个人外骨骼,目前不适用于 Fourier 产品 | 说明报销跑通后,康复买方可以支撑高客单价设备 |
币种和套餐口径未标准化。本表只是公开价格锚点的透明度地图,不是实际 ASP 测算表。
[CI018, CI020, CI021, CI022, CI023, CI037]没有公开标价的自助漏斗时,机构需求如何转化为硬件、服务和平台收入。
这是结构性桥接,不是量化收入拆分。公开来源显示了步骤,但没有给出每一步的转化率或利润率。
[CI013, CI014, CI015, CI017, CI018, CI034]公开价格和利润率锚点框定 Fourier 经济性的合理量级,但不假装能解出完整模型。
各项使用不同币种和视角,只作为边界锚点展示。它们没有按打包内容、地域或是否包含服务作标准化。
[CI020, CI021, CI022, CI037, CI038, CI043]4.3 同行类比显示可用毛利率有可能,但服务和商业化仍会把毛利吃掉
由于 Fourier 不发布自己的 P&L,推理经济性的最佳公开办法,是用受证据约束的可比公司。Ekso Bionics 报告 2024 年收入 US$17.9 million,毛利率约 53%,年末现金加受限现金 US$6.5 million,经营现金消耗 US$9.8 million。Lifeward 报告 2025 年收入 US$22.0 million,毛利率 38.2%,经营亏损 US$19.7 million,并明确称销量较低、关税压力、运费和制造费用吸收不足拖累了部分利润率。CYBERDYNE 的 FY2026 收入为 ¥3,846 million,在资产负债表已建立的情况下仍有 ¥601 million 经营亏损。这些可比公司重要,因为它们显示了一个反复出现的模式:康复机器人可以做到体面的硬件毛利率,却仍需要大量资本来支持销售周期、服务、报销工作和产品开发。Fourier 自身公开信号也指向同一方向。TMTPost 称制造成本仍高;患者使用外骨骼的类比需要 40 到 50 小时训练,且仍依赖支付方批准或机构预算。含义不是 Fourier 天然不盈利,而是这个品类运营足够重,单看毛利率无法回答现金消耗问题。[CI017, CI019, CI024, CI025, CI026, CI027]
| 指标 | 公开数值或状态 | 置信度 | 为什么重要 | 确切尽调索取项 |
|---|---|---|---|---|
| Fourier 当前收入 / ARR | 未披露 | 高 | 没有绝对收入规模,估值和烧钱速度就无法按销售额标准化 | 需要 2024-2025 年收入、若适用的 ARR,以及分部结构 |
| 海外康复业务增长线索 | 2023 年海外康复收入增长 >50%,约占总收入 10% | 低 | 这是最清晰的公开牵引力线索,但已经滞后,也不足以支撑趋势判断 | 需要经审计的 2024-2025 年地域收入拆分 |
| Fourier 实际毛利率 | 未披露 | 高 | 毛利率决定硬件定价能否转成可规模化经济模型 | 需要按康复、人形机器人、服务和软件线拆分的毛利率 |
| 部署 / 培训负担 | 同业参照显示,居家使用需 40-50 小时培训,并需要照护者支持 | 中 | 服务强度高,会吃掉毛利或拖慢部署速度 | 需要 Fourier 安装工时、培训义务和现场服务人员配比 |
| 同业毛利率区间 | Ekso 约 53%;Lifeward 2025 全年 38.2%;Lifeward 调整后 Q4 32.6% | 中 | 说明该品类硬件可以跑出正毛利,但差异很大 | 需要 Fourier 相对同业基准的毛利率桥接 |
| 同业年度亏损 / 烧钱区间 | Ekso 经营现金消耗 US$9.8m;Lifeward 经营亏损 US$19.7m;CYBERDYNE 经营亏损 ¥601m | 中 | 说明商业化和支持体系会让该品类在上市后仍然吃资本 | 需要 Fourier 月度烧钱、季度 EBITDA 和营运资本节奏 |
| 同业毛利拖累因素 | 销量偏低、固定开销吸收不足、关税、运费和报销摩擦 | 中 | 标出 Fourier 自身毛利在扩张期可能被压缩的位置 | 需要按产品线拆分的单机 BOM、运费敞口、关税和服务成本吸收情况 |
Fourier 不公布自身 P&L,因此同业参照只能作为有边界的代理。它们说明的是品类经济性,不是 Fourier 的实际结果。
[CI017, CI019, CI029, CI030, CI031, CI032]部署、关税和商业化投入仍然沉重时,看似强劲的硬件定价仍可能烧钱。
同行毛利率区间来自公开可比公司,不是 Fourier 自身数据。由于 Fourier 不披露 BOM、服务成本或分项 opex,该桥接是定性的。
[CI019, CI028, CI029, CI031, CI033, CI038]4.4 财务结论是细分需求和定价权有吸引力,但不透明度太高,难以清晰承销
证据支持一个可信的乐观情景:Fourier 多次完成大额融资,仍能拿到机构级价格点,在海外康复收入上有一个可信公开增长线索,并处于相邻报销和康复市场,预算可以支撑高价值硬件。但反向情景比管理层话术承认的更强。公开层面仍没有收入基数、ARR、毛利率、营运资本画像、月度烧钱速度或现金跑道。最新估值信号来自战略投资文件,而非经审计经营表现。公开价格集合内部也不均质:一些康复设备以几十万人民币成交,更大的系统和人形机器人分销商挂牌价则高得多,使按套装归一化的利润率推断不可能。同行类比显示,38% 到 53% 毛利率并不能阻止每年数百万美元亏损或烧钱。对承销而言,正确态度不是“Fourier 没有经济性”,而是“Fourier 披露的经济性不足以证明当前资本基础能转化为持久、内部融资的增长”。在形成硬财务确信前,管理层仍必须打开账本,披露实际成交价、分部结构、积压订单、现金、现金跑道和客户集中度。[CI016, CI024, CI029, CI031, CI035, CI040]
| 缺失指标 | 为什么重要 | 已有公开证据 | 确切尽调路径 | 不应从公开来源推断什么 |
|---|---|---|---|---|
| 2024-2025 年收入和分部结构 | 需要它来锚定估值、烧钱速度和产品市场匹配 | 公开信息只有 2023 年海外增长线索和装机基数说法 | 索取经审计管理账,并拆出康复硬件、人形机器人、服务和软件的分部桥接 | 不要仅凭融资规模或机构数量推断收入规模 |
| 分业务线毛利率 | 区分好的硬件定价和可持续经济模型 | 同业参照显示 38%-53% 毛利率可行,但还不够 | 索取产品级毛利率和服务成本吸收分析 | 不要把同业毛利率直接套到 Fourier 身上 |
| 现金余额、烧钱速度和现金跑道 | 决定融资依赖和下一轮紧迫度 | 融资轮次和估值公开,但当前现金不公开 | 索取最新资产负债表、月度现金消耗,以及基准 / 乐观 / 悲观情景下的现金跑道 | 不要把 E 轮加 Runyang 新闻当成当前现金跑道的证明 |
| 实际 ASP、折扣和服务附加 | 需要它把公开标价或中标价转成贡献毛利 | 公开来源只显示碎片化报价、分销商和采购锚点 | 索取按产品、地域和渠道拆分且附带服务的已成交交易清单 | 不要把挂牌价或招标价等同于实际毛利 |
| 债务、优先权和担保 | 可能彻底改变有效股权价值和下行风险 | 已审阅公开材料显示治理和股权稀释,不显示债务条款 | 索取债务明细、清算优先权、附属协议和担保 | 不要因为公开信息未提及债务,就假设股权结构表干净 |
| 客户集中度和地域收入敞口 | 需要它判断复购性、渠道权力和地缘敏感度 | 公开来源显示装机基数覆盖广,但不显示收入集中度 | 索取前 10 大客户、分销商结构和地域收入占比 | 不要把 40+ 个国家 2,000 家机构当成收入多元化证明 |
这里每一行都是财务测算的真实卡点,不是锦上添花的 KPI。
[CI016, CI017, CI040, CI041, CI042, CI045]4.5 图表
05产品与技术
5.1 Fourier 发布的架构足以证明平台雏形,但 GR-3 透明度仍不及 GR-2 或 N1
公开证据显示,Fourier 不是只靠单一人形机器人样机做展示的公司。留存材料覆盖三代可见人形机器人及底层部件:GR-1 是公司叙事中的首个量产平台,GR-2 有细到支持中心的架构说明,N1 则是更小、面向开放开发者的人形机器人。GR-2 文档尤其具体,把机器人拆成头部、躯干、手臂、手、腰部和腿部,列出最高 53 个自由度、自研 FSA 2.0 执行器,以及可选 12 自由度灵巧手。DexHand 部件页给出私人机器人厂商少见的底层细节,包括执行器数量、Ethernet 控制、负载规格和 SDK 支持。相比之下,GR-3 的公开定位更多靠照护交互和 CES 演示,而不是同等深度的工程手册。留存公开证据支撑其多模态感知与交互架构,但电池、载荷、续航、可靠性、现场服务等披露没有同样深度,外部买家还不能把它视为规格已完全验证的产品。[CE001, CE002, CE003, CE005, CE006, CE007]
| 资产或层级 | 主要用户 / 买方 | 公开成熟度状态 | 有证据支撑的差异化 | 主要尽调缺口 |
|---|---|---|---|---|
| RehabHub 平台 | 康复中心、医院、治疗师 | 已商业化且广泛部署 | 在数字训练环境中连接 30+ 类机器人,带数据采集和工作流编排 | 未公开模块级 ROI、留存或附加率 |
| 上肢康复模块(ArmMotus、HandyRehab、OTParvos、WristMotus) | 治疗卒中、神经康复和上肢恢复患者的治疗师 | 已商业化,且有具名客户场景 | 结合缆索驱动柔顺控制、EMG 或精细运动训练,以及认知 / 互动治疗 | 未公开按模块拆分的比较结果数据集 |
| 下肢和移动能力模块(AnkleMotus、CycleMotus、BalanceMotus、ExoMotus) | 步态、平衡、移动能力和早期康复团队 | 已商业化,有手册和客户证据 | 从早期下床活动到步态辅助均有覆盖,并加入力反馈和传感训练 | 支付方路径、使用率和服务负担仍不透明 |
| GR-1 人形机器人 | 研究人员、企业试点、创新团队 | 2023 年以来公开的产品平台 | 量产定位,加上集成 FSA 和纯视觉感知叙事 | 量产买家名称和支持条款未公开 |
| GR-2 人形机器人 | 研究、工业实验、医疗康复探索 | 架构信息丰富,但应用证据仍偏实验性 | 53 DOF、FSA 2.0、可选 12-DOF 双手,以及更开放的开发表面 | 未公开付费部署案例或可靠性 KPI |
| GR-3 照护机器人 | 照护、陪伴、接待和辅助交互试点 | 处于展示阶段的国际商业化 | 55 DOF,以及围绕照护打造的多模态交互栈,强调温度感和触觉 | 没有具名量产客户、认证包或详细现场规格 |
| N1 开源人形机器人 | 全球开发者、研究人员和实验室 | 面向开发者,商业化仍早期 | 更小的 23-DOF 机身,运动范围有文档说明,也开放传感器扩展 | 公开部署证据远薄于工具面 |
| 组件和 API 层(DexHand、FSA、SDKs、客户端) | 内部工程师和外部开发者 | 有公开文档,但部分工具仍处 alpha 阶段 | 自研手部和执行器栈开放到底层 SDK | 发布纪律、兼容性稳定性和安全姿态披露不足 |
本矩阵混合了商业部署、产品页、支持文档和开发者仓库。「公开成熟度状态」反映留存公开材料能证明什么,不代表内部出货量、正常运行时间或收入排名。
[CE001, CE005, CE009, CE011, CE024, CE030]分层展示 Fourier 的公开产品架构如何从照护和康复工作流,一路下探到机器人本体、组件、数据和开发者工具。
这张图是结构图,不是覆盖每个内部子系统的网络图。它只反映留存公开来源中明确可见的模块。
[CE002, CE003, CE007, CE010, CE012, CE024]5.2 开发者接口真实且少见地开放,但仍像早期集成栈
Fourier 在 GitHub 和支持中心留下的公开足迹相当扎实。公司明确把自己定位为向开源社区贡献软件、硬件设计和研究成果;留存仓库基本拼出一个完整的具身 AI 闭环:用遥操作采集数据,用 ActionNet 管示范数据,配套模仿学习框架、URDF 和 MJCF 机器人模型、Isaac Gym 训练、MuJoCo 验证,以及面向外部控制的客户端库。NVIDIA 的合作伙伴博客进一步强化了这一点:Fourier 在 GR-2 上使用仿真到现实工作流,并报告了地面起身任务在 3,000 次迭代后的迁移结果。但这套栈还谈不上无摩擦。遥操作仓库提醒,首次硬件设置很棘手,且超出教程范围;公开 GRx 客户端也明确处于 alpha 阶段且不稳定。合起来看,公司内部确实有技术成熟度,外部开发者体验也强过纯营销展示,但还不是成熟机器人平台那种打磨好、按版本发布管理的生态。[CE013, CE014, CE015, CE016, CE017, CE018]
| 层级 / 流程 / 组件 | 在运营模型中的作用 | 公开证据 | 关键依赖 | 主要风险 |
|---|---|---|---|---|
| 驱动与具身 | 借助 FSA、DexHand、关节和传感器,把控制策略转成运动 | 支持文档和 SDK 页面披露执行器、手部和网络细节 | 48V 电源、Ethernet、Linux 托管的 SDK 工具和精密机械集成 | 外部用户仍看不到公开硬件验证、MTBF 或备件视图 |
| 感知与交互 | 支持语音、触摸、视觉和面向人的行为 | GR-2 和 GR-3 文档描述了视觉、音频、触觉和语音模块 | 传感器校准、嵌入式计算和面向具体应用的交互设计 | GR-3 交互主张比公开工程细节更丰富 |
| 数据采集与学习 | 采集示教,并为模仿学习和 VLA 式任务准备模型 | 遥操作概览和仓库提到 ActionNet 及数据流水线 | VR 头显、摄像头、手部追踪和现场设置质量 | 硬件设置复杂,会拖慢第三方团队上手 |
| 仿真与机器人描述 | 真实部署前,让策略先跑过训练和验证 | GitHub 仓库覆盖 URDF、MJCF、Isaac Gym、MuJoCo 和真机部署流程 | NVIDIA GPUs、Isaac 或 MuJoCo 环境,以及模型文件一致性 | 公开栈覆盖面广,但还没有封装成成熟企业平台 |
| 运行时与外部控制 | 向开发者和集成商开放机器人控制 | GRx client 和 SDK 仓库暴露安装路径和 API | 机器人固件、客户端库和支持文档之间的版本兼容性 | 客户端库仍处 alpha 阶段且明确不稳定,引出支持和发布纪律问题 |
本表把 Fourier 的技术栈拆成买方或开发者关心的层级:具身、传感、学习、仿真和运行时。它刻意强调依赖和失效模式,而不是重复产品营销。
[CE011, CE012, CE015, CE016, CE017, CE019]Fourier 的公开产品栈依赖一条链:从组件、机器人型号和仿真,到运行支持和客户部署。
这张图把多个代码仓库和文档界面压缩成一条依赖链。它展示外部采用依赖什么,不是复刻每个代码包。
[CE015, CE016, CE017, CE019, CE020, CE021]5.3 康复系统拥有最清晰的公开部署证据,也最贴合工作流
如果人形机器人故事仍有一部分面向未来,康复业务已经进入商业化。Fourier 的康复页面声称覆盖 40 多个国家的 2,000 多家机构;留存的 RehabHub 宣传册和客户网站页面给本章提供了具名部署证据,而不是抽象营销话术。Ascot 描述了由顾问主导、现场配备十多台设备的多学科路径;STEPS 细列上肢、踝部、循环训练和外骨骼模块;Hobbs 则列出门诊场景中 ArmMotus EMU、ArmMotus M2 Pro 和 ExoMotus M4 的实际安装。宣传册还说明这些部署背后的系统逻辑:设备互联、数字化追踪训练、力和位置感知、自动生成报告,以及一名治疗师可监测多名患者的人员配置模型。这是本章最强证据,说明 Fourier 已经越过演示品阶段。公开记录支撑真实康复产品化、具体治疗师工作流和有意义的实施经验,尽管产品级疗效基准和单位经济性仍披露不足。[CE030, CE031, CE032, CE033, CE034, CE035]
| 用户任务 | 当前工作流问题 | Fourier 方案 | 留存来源中的可观察收益 | 限制 / 未验证点 |
|---|---|---|---|---|
| 上肢治疗师 | 手工带患者反复够物和精细运动练习,劳动强度高,也难标准化 | RehabHub 中的 ArmMotus、HandyRehab、OTParvos 和 WristMotus | 客户场景和手册明确描述了互动任务、可衡量的进展报告和更高重复性 | 没有外部随机对照证据证明这些特定模块能带来更好的长期结果 |
| 步态 / 移动能力临床人员 | 站立、步态和踝部训练需要高强度监督和谨慎递进 | ExoMotus、AnkleMotus、BalanceMotus 和 CycleMotus | 公开材料描述了按需辅助的步态练习、力或位置传感,以及更早期的移动能力支持 | 报销、使用率和人员配置经济性没有公开拆分 |
| 康复运营负责人 | 多设备路径会切碎数据,也分散治疗师注意力 | 作为互联数字环境的 RehabHub | 手册称一名治疗师可监测多名患者,并支持快速设置和自动生成报告 | 这些效率说法来自手册和客户证据,不是经审计运营数据 |
| 机器人开发者 | 采集具身示教并把它们转成策略,在技术上很碎片化 | 遥操作系统、ActionNet、模仿学习框架、URDF/MJCF 模型、Isaac 和 MuJoCo 仓库 | 留存文档显示,从数据采集到仿真再到策略部署有一条连续路径 | 硬件设置仍高度依赖支持,公开客户端 / 运行时尚未稳定 |
| 照护 / 接待运营方 | 面向人的交互需要把视觉、音频、触觉和行为规划打包在一起 | GR-3 照护机器人演示和交互栈 | CES 演示把语音、触摸、下棋和运动交互整合成一个连贯的公开包 | 公开证据仍以演示为主,而不是具名量产运营 |
各行映射留存公开证据中最可见的待办任务。本表把可观察工作流收益,与已发布 ROI、报销或续约数据等买方级证据分开。
[CE014, CE015, CE031, CE033, CE034, CE035]Fourier 最有公开验证的工作流,如何从治疗师目标进入机器人训练、数据采集和照护计划迭代。
该流程抽象自客户页面和 RehabHub 手册中记录最充分的康复运营模式。GR-3 交互试点可能也遵循类似监督闭环,但公开资料少得多。
[CE032, CE033, CE034, CE035, CE038]5.4 Fourier 在康复业务上已经商业化,人形机器人技术也认真,但公开证据还没有把照护机器人规模风险降下来
留存证据支持一个分裂判断。一方面,Fourier 的可信技术护城河来自康复渠道触达、部件级垂直化和真实开发者栈。另一方面,与康复业务相比,其人形机器人商业化公开证据仍偏薄。留存材料中最强的人形机器人商业化来源,是 36Kr 关于数百台 GR 系列出货和医院示范基地合作的报道;最强的 GR-3 来源仍是 CES 发布材料和展示报道。这足以说明 Fourier 不是空气产品,但还不足以说明 GR-3 已经在医院、养老或大型付费企业队列里完成部署验证。最大的公开缺口正是承销时最关键的几项:具名生产客户、照护场景下的详细合规或医疗许可证据,以及联网机器人运行中的安全或隐私控制文档。因此,本章判断应当不对称:Fourier 在康复上看起来商业化真实,在人形机器人上技术可信,但照护人形机器人侧仍更接近试点阶段商业化,而不是已有规模化运营证明。[CE041, CE042, CE043, CE044, CE045, CE046]
| 控制项或信号 | 公开内容 | 覆盖范围 | 仍未验证什么 | 投资判断 |
|---|---|---|---|---|
| 临床质量叙事 | 客户页面和 36Kr 反复强调有效性、治疗师监督和医疗级质量 | 康复部署和未来面向医疗的人形机器人场景 | 留存公开材料没有确切认证、质量体系范围和审计追踪 | 是有用信号,但不能替代合规材料包 |
| 具名站点治理 | Ascot、STEPS 和 Hobbs 将这项技术放进由治疗师主导的路径和受监督照护场景 | 康复用途,而非自主通用操作 | 没有按站点公开的正常运行时间、事件或续约数据集 | 有力证明它适合受监督工作流 |
| 开发者发布纪律 | 公开客户端仍处 alpha 阶段且不稳定;遥操作设置说明把首次硬件问题导向支持团队 | 外部开发者和集成商使用 | 没有公开 SLA、发布节奏承诺或长期兼容政策 | 透明,但仍处早期 |
| 网络和控制接口 | DexHand 和 FSA 文档暴露 Ethernet、静态 IP、电源和调试细节;遥操作文档要求 HTTPS 证书和设备专属设置 | 底层机器人和外设集成 | 未留存安全架构、认证、日志或补丁管理文档 | 集成商可见性不错,企业安全可见性偏弱 |
| 隐私 / 监管证据 | 留存材料包含照护和医疗使用叙事,但没有 GR-3 的详细隐私或监管文档 | 照护或康复场景中面向人的机器人 | 具名医疗许可、数据治理控制和按司法辖区划分的审批仍未核实 | 这是本章最大的尽调卡点之一 |
本章的信任证据来自工作流位置、公开文档习惯,以及工具链中的明确警示。证据足以把康复部署和纯演示秀区分开, 但不足以通过认证、网络安全或隐私尽调。
[CE017, CE020, CE026, CE038, CE045, CE046]| 日期 / 阶段 | 功能或里程碑 | 公开状态 | 含义 | 来源依据 |
|---|---|---|---|---|
| 2023 产品页 | 公司将 GR-1 定位为首款量产人形机器人 | 公开产品化发布状态 | 说明在 GR-2 和 GR-3 之前,Fourier 已有可销售的人形机器人叙事,而不只是一次性的 CES 亮相 | 官方 GR-1 页面 |
| 2024 年末架构阶段 | GR-2 发布,硬件设计、更灵巧的操作能力和开放开发叙事更完整 | 详细支持文档加公开文档仓库 | 这是公开可检查架构和开发者包装上最清晰的一次升级 | 支持中心文档和 GR-2 文档仓库 |
| 2024 年末学习栈阶段 | NVIDIA 记录了 GR-2 的 sim-to-real 工作,以及从 Isaac Gym 向 Isaac Lab 迁移 | 经合作伙伴验证的技术工作流 | 支撑真实的训练与迁移叙事,而不是静态动作演示 | NVIDIA 技术博客 |
| 2025 开发者触点阶段 | N1 文档、模型、仿真仓库和开放开发者叙事已经公开 | 面向开发者,但商业化仍早期 | 扩大了访问和试验面,尽管客户证据仍稀疏 | N1 文档、GitHub 组织和训练仓库 |
| 2025 护理机器人阶段 | GR-3 以聚焦护理、具备多模态交互的人形机器人身份推出 | 公开发布和定位 | 把 Fourier 从康复相邻的工具场景延伸到陪伴和交互场景 | PR Newswire 和 Gasgoo |
| 2026 展示与医疗探索阶段 | CES 首秀加医院示范基地合作 | 国际展示加国内医疗共创 | 释放商业化意图,但距离具名的大规模生产部署仍有差距 | PR Newswire、Gasgoo 和 36Kr |
本路线图有意混合产品页、支持文档、合作伙伴验证和商业化事件。各行是有证据支撑的成熟度标记, 并不等同于出货规模或盈利能力证明。
[CE005, CE009, CE027, CE029, CE041, CE042]公开证据在 RehabHub 上最强,在 GR-3 上更弱;N1 对开发者异常开放,但客户证据仍偏少。
该矩阵是定性判断,按证据加权。“高”表示留存材料中有反复且具体的公开证据;“低”通常表示公开证据稀疏,不等于技术失败证据。
[CE004, CE026, CE043, CE044, CE047, CE048]5.5 证据要点
06客户情况
6.1 康复买家清晰可见;付款人与预算权仍嵌在服务提供方工作流里
Fourier 的公开客户地图不是单一同质买方。文档最充分的用户,是康复服务提供商和医院康复科;它们采购技术,是为了嵌入治疗师主导的照护交付,而不是买一台独立消费设备。在这些场景里,机构是买方,治疗师和康复团队是日常用户,患者则是既有照护项目中的受益者。公开网站页面反复围绕可测量训练、工作流贴合、治疗师效率和数据采集讲价值,而不是只讲硬件新奇。这很关键,因为支付逻辑位于机器人之上一层:CMS 文档规则和 NIH 报销指南都显示,医疗设备采用仍要嵌入可计费治疗路径、留档照护计划和服务提供方内部 ROI 案例。由此形成的买方栈里,康复主任、创新负责人、采购或财务委员会、临床文档负责人都重要。在这套栈的边缘,Fourier 还展示了研究实验室买家、分销商主导的私人诊所以及早期具身照护试点,但这些细分场景的公开付款和续约可见度,远弱于康复核心业务。[CU005, CU006, CU007, CU008, CU009, CU043]
| 客群 | 典型买方 | 日常用户 | 预算 / 付款方逻辑 | 公开证据质量 | 主要披露缺口 |
|---|---|---|---|---|---|
| 机构康复中心和神经康复诊所 | 诊所运营方、康复负责人或业主 | 治疗师和治疗助理 | 服务方资本开支,加可计费治疗吞吐量 | 强 — 多个具名站点披露具体设备和工作流细节 | 各站点未公开续约或价格信息 |
| 公共或医院康复服务 | 医院创新、康复医学或采购负责人 | 医院治疗师和物理治疗师 | 医院预算,加有文档记录的治疗路径和合规要求 | 中到强 — NHG/TTSH、Brooks、Saifee 和 Ren Ci 显示真实活动 | 合同规模和采购路径通常未披露 |
| 研究机构和联合实验室 | 教授、PI、创新办公室或研究中心 | 研究人员、临床医生和研究参与者 | 课题经费、实验室预算或机构创新预算 | 中 — 有具名实验室和研究,但收入状态常常不清 | 付费客户状态和商业转化很少披露 |
| 分销商主导的私营服务方 | 通过区域伙伴合作的诊所运营方 | 治疗师和本地服务团队 | 服务方采购预算,加渠道支持和服务叠加 | 中 — 渠道页和客户页显示设备已进入市场 | 终端客户名单、利润率和服务 SLA 不透明 |
| 护理 / 人形机器人试点 | 创新赞助方、试点负责人或 R&D 预算持有人 | 受监管场景中的员工、患者或访客 | 试点或创新资金,而非已验证运营预算 | 低 — 公开证据主要是演示、MOU 或预售表述 | 没有公开披露的具名付费 GR-1 或 GR-3 客户 |
| 居家或社区康复项目 | 医院或康复网络护理经理 | 患者和社区康复团队 | 服务方 / 社区护理预算,加远程护理经济性 | 低到中 — NHG 和 PRLog 显示意向和居家化抱负 | 几乎没有公开证据显示已规模化部署队列或实现报销转化 |
本表区分谁签支票、谁使用设备,以及谁获得临床收益。“预算 / 付款方逻辑”反映可见的公开路径, 而非已确认的合同机制。
[CU005, CU006, CU007, CU008, CU009, CU043]典型机构客户从康复需求出发,走向部署、工作流集成,以及可能扩张的路径。
阶段根据具名站点页面、NHG/Brooks 合作材料和报销指南推断。它们展示可观察路径,不代表实测的全公司转化率。
[CU006, CU007, CU008, CU014, CU022, CU030]公开证据从宽泛关系信号到可持续收入证明,收窄得很快。
数值是对留存公开材料按证据加权后的示意,不是内部销售管线转化率。100 代表来源中最宽的可见关系池,0 代表没有披露续约经济性或具名付费人形机器人客户。
[CU015, CU031, CU035, CU036, CU040]6.2 具名康复现场证据真实且可跨地域复现,但仍是样本式证据
最可信的客户证据来自具名康复现场和医院集群,它们说明了设备和用例。NHG Health 称 Fourier 技术已在 Tan Tock Seng Hospital 部署并评估,并把关系扩展为新的五年期 MOU 和 Joint Rehabilitation Innovation Hub。Brooks Rehabilitation 称 ArmMotus EMU 和 ExoMotus M4 已部署在佛罗里达的 Center for Innovation。英国的 STEPS、Ascot 和 Hobbs 都给出超越标识墙的现场级证明:STEPS 列出广泛的 RehabHub 套件和结果报告,Ascot 描述由顾问主导、包含 10 多台设备的路径,Hobbs 则点名 Laverstock Care Centre 已安装的设备栈。印度方面有 Saifee Hospital 的 ExoMotus M4 安装;新加坡 Ren Ci 通过 ExoMotus 辅助试点研究,给出了最强的量化现场级结果证据。因此,跨多个国家和机构类型的复现性真实存在。但公开记录仍像样本,而不是核对过的客户清单;若干重要引用——例如 Nagoya 或 Shirley Ryan——更接近研究和生态证明,而不是已披露的经常性商业账户。[CU010, CU011, CU012, CU013, CU014, CU016]
| 日期 / 期间 | 标记 | 证明内容 | 置信度 | 重要限制 |
|---|---|---|---|---|
| 2021 | NHG 和 Fourier 签署初始 MOU | 大型公共医疗集团的兴趣,以及早期居家康复抱负 | 中 | MOU 本身不能证明收入或部署规模 |
| 2023 版宣传册 | 1,000+ 次安装、30+ 个国家、17 个联合实验室 | 2026 刷新前,早期平台规模和研究触达已经存在 | 中 | 公司宣传册口径;未披露活跃账户分母 |
| 2023 NR Times 专访 | 200+ 个 RehabHub 在运营 | RehabHub 概念已扩展到全球多个具名站点 | 中 | 第三方文章;未发布完整站点清单 |
| 2025 | Brooks MOU 和设备部署 | 具名美国康复合作伙伴,披露站点级设备 | 高 | 以研究合作呈现,而非披露合同金额 |
| 2025 | CMC Ludhiana 的 IFNRCON | ArmMotus 和 ExoMotus 在印度市场及临床医生中的可见度 | 中 | 会议露出不等于医院采购 |
| 2026 | NHG 将合作续约五年,并提及 TTSH 部署 / 评估 | 医院集团合作延续,并从研究推进到服务整合 | 高 | 收入贡献和设备数量仍未披露 |
| 2026 | Ren Ci 试点研究,47 人招募 / 30 人完成 | 量化的新加坡站点级使用和早期结果证据 | 中 | 临床海报,而非商业披露 |
| 2026 | Saifee Hospital 新增 ExoMotus M4 | 印度新增具名三级医院安装 | 中 | 证据来自行业媒体,而非医院一手披露 |
这些标记有意混合公司、客户和第三方来源,用来呈现哪些公开信息可以打上时间戳。 它们不是完整销售历史重建。
[CU003, CU010, CU011, CU014, CU023, CU025]| 机构 / 账户 | 客群 | 部署或关系 | 证据状态 | 明确披露内容 | 主要限制 |
|---|---|---|---|---|---|
| Brooks Rehabilitation (Florida) | 康复服务方 / 研究中心 | Center for Innovation 部署 ArmMotus EMU 和 ExoMotus M4 | 已部署研究环境 | 具名站点、具名设备、明确已部署,并有协议开发意图 | 公开材料未披露合同金额、利用率或续约 |
| NHG Health / Tan Tock Seng Hospital(Singapore 医院) | 公共医疗集团 / 医院康复 | 精选 Fourier 康复技术已部署并评估;2026 年五年期 MOU 和 Joint RehabHub | 真实部署加扩大合作 | 具名医疗集团、具名医院、既往协议和护理场景范围 | 采购规模、产品组合和付款方经济性仍未公开 |
| STEPS Rehabilitation (UK) | 私营神经康复服务方 | RehabHub 套件,包含 ArmMotus EMU、ArmMotus M2、AnkleMotus、CycleMotus、ExoMotus M4 | 运营中客户站点 | 具名设备、英国稀缺性表述、患者使用和疗程报告证据 | 私营诊所合同经济性和续约未公开 |
| Saifee Hospital (Mumbai) | 三级医院 | ExoMotus M4 步态训练安装 | 近期具名安装 | 多家行业来源提及同一设备,并说明它适配高流量医院工作流 | 缺少医院一手来源和多年耐用性数据 |
| Ren Ci Community Hospital and Day Rehab Centre(Singapore 机构) | 医院 / 日间康复服务 | ExoMotus M4 辅助步态康复项目 | 试点研究 / 运营项目 | 47 人招募、30 人完成,行动能力显著改善,无不良事件,并节省工时 | 海报不能证明采购金额或复购 |
| Nagoya University / NHG 生态 | 研究伙伴 / 医疗集团 | ArmMotus EMU 肌肉协同分析和 RehabHub 合作 | 生态合作 | 显示亚洲市场具备跨境复制性和研究采用 | 不应等同计入已披露付费终端客户 |
本表是保留来源集中可公开记录的具名引用样本。它有意保持保守,排除了许多缺乏支撑的标识 或只有一个弱来源的引用。
[CU010, CU012, CU014, CU016, CU023, CU025]康复服务商站点的明确证据最强;人形机器人和渠道证据远没有那么贴近收入。
评级是定性判断,反映留存来源中反复出现的具体程度,不是内部账户评分模型。
[CU015, CU018, CU026, CU032, CU033, CU034]6.3 公开证据质量在康复工作流上最高,在收入耐久性上最弱
本章证据质量并不均衡,这对承销很重要。康复现场页面和医院研究可以证明真实使用、设备组合和部分运营适配。NHG 和 Brooks 强于纯新闻稿标识,因为它们描述了部署或现场评估。STEPS 和 Ren Ci 更进一步,说明了使用情况、疗程报告或测量结果。但即便最好的公开引用,通常也停在投资者判断耐久性所需信息之前:合同期限、复购历史、续约时点、NRR、GRR、流失、附加率和现场级利用率曲线均未披露。这意味着公开记录支持“真实采用”的判断,却不必然支持“收入粘性强”。离开核心康复业务后,质量缺口进一步扩大。Nagoya、Kurage 和 HTPG 是有战略意义的生态信号,但不应与终端客户证明混为一谈。它们更多展示渠道开发、R&D 合作和打包方案野心,而不是已预订需求。因此,公开客户证据能确立 Fourier 在康复上商业化真实,却仍未回答这个装机基础能否以有吸引力的单位经济性可靠扩张、续约和交叉销售。[CU015, CU018, CU025, CU026, CU032, CU033]
| 指标 | 公开数值 | 客群 | 置信度 | 尽调要求 |
|---|---|---|---|---|
| 安装基数规模 | 2,000+ 家机构 / 医院,覆盖 40+ 个国家 | 整体康复基数 | 中 | 要求提供活跃账户数,并与历史累计安装数对照 |
| RehabHub 数量 | 200+ 个在运营 | RehabHub 站点 | 中 | 要求提供站点清单、上线日期和退役站点历史 |
| 站点级量化结果 | Ren Ci:47 人招募,30 人完成;平衡和行动能力显著改善;无不良事件 | 具名新加坡项目 | 中 | 要求提供所有具名站点的标准化结果看板 |
| 疗程报告和参与度 | STEPS 提及疗程报告和游戏化联网使用;Saifee 提及数据驱动步态训练 | 选定站点子集 | 中 | 要求提供利用率、节省治疗师工时和患者完成曲线 |
| 续约 / 复购 | 所有客群 | 低 | 要求提供重订购历史、合同年限和按站点扩张情况 | |
| NRR / GRR / 流失 / 满意度 | 所有客群 | 低 | 要求提供 NRR、流失原因、患者和治疗师满意度调查,以及可供访谈客户 |
空值表示保留的客户证据集中未公开该指标,不代表公司内部一定没有该指标。
[CU001, CU025, CU026, CU035, CU036]Fourier 披露了可观的康复规模指标,但这些指标本身不能说明客户关系是否耐久。
柱状项混合了不同公开指标和年份:2026 年当前页面数字、2023 年手册和 Neuro Rehab Times。它们说明覆盖广度,不说明经常性收入质量。
[CU001, CU003, CU029, CU042]6.4 采用摩擦来自资本审批、工作流集成、渠道不透明,以及缺失的人形机器人客户名单
主要客户风险并不是 Fourier 没有任何现实采用,而是公开采用数据还不能让外部人把耐久收入与一个漂亮但混合的生态故事区分开。康复买家仍要过文档、编码、人员配置和资本开支逻辑,因此销售周期很可能锚定在服务提供方经济性和临床验证上,而不是纯功能兴奋。Thor、HTPG、Robocare 和 Healthlink 等渠道伙伴可以加速触达,但也会降低终端客户身份、服务义务和利润率捕获的可见度。当公开具名客户披露本来就相对 2,000+ 家机构的宣传口径偏薄时,这种不透明更严重。最大的证明缺口在人形机器人和照护侧。独立反向分析称,没有公开确认的付费 GR-1 或 GR-3 部署具备具名客户、KPI 或 ROI;公开照护场景引用仍更像演示、预售或创新合作,而不是规模化付费生产。在 Fourier 披露续约、客户集中度,以及有可测量结果的具名人形机器人试点之前,客户故事应被视为康复主导,同时叠加重大转化和集中度不确定性。[CU007, CU008, CU022, CU037, CU038, CU039]
| 扩张驱动因素或风险 | 公开记录显示 | 潜在影响 | 尽调路径 |
|---|---|---|---|
| RehabHub 在 30+ 款机器人上的交叉销售 | 平台可把多台设备打包进同一站点,并标准化数据采集 | 康复路径成熟的场景下,可能加深账户价值 | 询问附加率、单站平均设备数和服务收入结构 |
| 研究和联合实验室生态 | 17 个联合实验室和多个学术伙伴提升可见度,并带来产品反馈回路 | 能加快证据生成;若计作客户,可能夸大已变现需求 | 在账户台账中区分付费客户和 R&D 伙伴 |
| 分销商主导渠道 | Thor、HTPG、Robocare 和 Healthlink 拓宽市场触达 | 渠道杠杆可能削弱直接账户可见度和利润率控制 | 审阅分销合同、现场服务义务和终端客户报告节奏 |
| 医院创新中心和 MOU | NHG、Brooks 和 Nagoya 展示由合作带来的采用入口 | 试点若不转化,管线可能比已签约经常性收入更好看 | 要求提供从 MOU / 演示到采购订单或多年合同的转化率 |
| 人形机器人护理叙事 | 未公开具名付费 GR-1 或 GR-3 客户 | 若按近期收入看待,可能拉长销售周期,并分散康复主业焦点 | 要求提供护理用例的具名付费试点、ROI 数据和监管路径里程碑 |
本表区分可能成为真实增长驱动的扩张机制,以及那些让公开证据难以支撑投资判断的披露缺口。
[CU022, CU032, CU033, CU037, CU038, CU039]6.5 证据要点
07风险
7.1 康复医疗科技风险在报销、合规和上市后纪律,不在产品是否存在
Fourier 的康复业务已经不是从 0 到 1 的证明故事。Brooks 和 Fourier 自身材料支撑一套真实装机产品,公司也公开把叙事锚定在 2,000+ 家机构、40 个国家的足迹上。相较许多人形机器人初创公司,这降低了产品存在性风险,但没有消除医疗科技式监管和报销风险。在欧洲,AI 医疗设备今天仍经由 MDR 进入市场;随着制度分阶段落地,还要吸收 AI Act 对数据治理、透明度、网络安全、人类监督和上市后监测的义务。在美国,FDA 指南把网络安全纳入设备合规;一旦患者邻近系统在现场失效,召回、纠正、移除和警戒制度都会变得重要。需求侧,CMS 仍把治疗经济性绑在有记录的医疗必要性和正确编码上,而不是机器人新奇感上。RehabHub 的工作流和数据承诺可能帮助服务提供方,但投资者仍需要证明这些好处会转化为可报销使用、续约质量和耐久现场经济性。[CR002, CR003, CR004, CR005, CR006, CR010]
| 规则 / 案例 / 制度 | 司法辖区 | 当前状态 | 可能性 | 严重性 | 缓释措施 | 剩余敞口 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| 未披露的患者相邻人形机器人认证路径 | 美国 / 欧盟 / 重点护理市场 | 保留来源中未公开 GR-3 或 GR-1 护理部署的 FDA / CE / 患者使用路径 | 高 | 严重 | 可借助康复关系和未来申报资料工作 | 在具名路径和时间表公布前,敞口极高 | 要求提供逐型号监管矩阵、预期用途边界,以及任何监管机构或公告机构往来 |
| 欧盟 MDR + AI Act 医疗器械 AI 栈 | 欧盟 | MDR 仍是当前路径;AI Act Annex I 义务预计自 2027 年 8 月起适用,指南已指向数据治理、 透明度、监督、网络安全和上市后监测 | 中 | 高 | 把 AI 控制纳入现有 MDR QMS 和技术文件 | 高,因为未披露公开的 AI 专项资料包 | 审阅技术文件、数据集治理、人类监督设计和上市后监测计划 |
| FDA 对联网设备的网络安全和上市前预期 | 美国 | FDA 将网络安全和第 524B 条提交内容纳入设备合规 | 中 | 高 | 复用医疗科技式网络安全设计控制和提交纪律 | 若没有公开 SBOM、威胁模型或补丁流程证据,敞口为高 | 要求提供 524B 准备包、SBOM、威胁模型和协同漏洞流程 |
| 安全事件后的召回 / 纠正 / 警戒制度 | 美国 / 欧盟 | 一旦出现健康风险事件,FDA 召回、纠正和移除规则,以及欧盟警戒和现场安全纠正措施制度都会适用 | 中 | 高 | 若系统已建立,安装基数反馈可支撑投诉处理和 CAPA | 中高,因为公开事件管理证据较薄 | 要求提供投诉处理 SOP、CAPA 看板、召回模拟结果,以及 NCAR / FSCA 历史 |
| CMS 医疗必要性和治疗编码依赖 | 美国服务方市场 | 超过 2026 阈值的治疗需要 KX 支撑的医疗必要性;RTM 扩展有帮助,但计费仍由规则驱动 | 高 | 高 | 工作流和数据功能可能帮助服务方记录并监测护理 | 站点 ROI 层面敞口高,因为利用率和拒付数据未披露 | 要求提供付款方组合、KX 拒付率、RTM 使用、按设备利用率,以及旗舰站点续约经济性 |
| 护理交互中的生物识别和隐私合规 | 欧盟 / 英国及隐私敏感市场 | 面部、语音、触摸和情绪线索处理会给护理机器人带来 GDPR 和问责问题 | 中高 | 高 | 隐私内建和清晰角色分配可降低敞口 | 在 DPA 和数据流架构可用前,敞口为高 | 要求提供 DPA、留存地图、生物识别处理清单、云 / 数据驻留,以及事件通知条款 |
各行按剩余投资严重性排序,并有意区分康复医疗科技义务和人形机器人平台义务。
[CR010, CR011, CR012, CR013, CR014, CR015]| 失效模式 | 可能性 | 严重性 | 缓释成熟度 | 剩余敞口 | 未解决缺口 |
|---|---|---|---|---|---|
| 联网康复设备或临床数据通道遭网络入侵 | 中 | 高 | 低到中 | 高 | 未保留公开 SBOM、漏洞披露或事件响应资料包 |
| 康复使用中贴近患者的运动、感知或控制失效 | 中 | 高 | 低至中 | 高 | 未保留公开 CAPA 摘要、现场故障率或召回演练证据 |
| 非结构化护理或公共环境中的人形机器人行为失效 | 高 | 极高 | 低 | 极高 | 没有公开经验证的生产运行数据或护理场景可靠性指标 |
| 全球装机基础和多机器人工作流带来的服务与支持负担 | 中 | 中高 | 中 | 中高 | 未见公开可用的正常运行时间、SLA、备件或现场服务人员披露 |
| 削弱临床医生信任和文档记录的数据日志或集成失效 | 中高 | 高 | 中 | 高 | 未保留面向人的护理部署中的公开审计追踪或医院 IT 案例研究 |
运营行聚焦可能传导至患者安全、客户信任和收入韧性的失效模式,而不是泛泛的制造风险。
[CR003, CR004, CR015, CR016, CR017, CR020]康复业务风险更偏报销 / 合规;人形机器人风险更偏认证、责任和商业化。
矩阵分数来自留存来源的定性、证据加权判断,不是内部风险模型输出。
[CR015, CR020, CR022, CR026, CR036, CR043]7.2 人形机器人逻辑受缺失运营证明和隐私 / 责任暴露约束,远甚于认知度
人形机器人的风险栈不同于康复医疗科技栈。公开来源可以展示 GR-3 的设计意图、CES 可见度、照护导向交互功能,以及 Fourier 从研究和康复走向更以人为中心环境的野心。但它们还没有展示具名付费部署、已披露监管路径或现场可靠性记录——这些才足以让投资者承销养老或医院邻近场景的规模化。独立追踪器和反向报道都把人形机器人线描述为商业化前或未经验证:没有公开确认的付费 GR-1 或 GR-3 部署,没有具名 ROI 案例,高端 GR-2 的渠道触达也有限。这很重要,因为面向人的照护机器人把物理安全、自治、隐私和责任揉成一个包。面部、语音、触摸和情绪线索处理可能带来 GDPR 和生物识别暴露;任何从演示空间进入辅助场景的动作,也会让制造商、运营方、集成商和云 / 软件提供商之间的责任问题更尖锐。正确读法不是“没有丑闻浮出,所以低风险”,而是“后果很重的风险,因为公开营销界面眼下跑在披露的合规包前面”。[CR007, CR008, CR026, CR027, CR028, CR029]
| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| 法规事务与质量负责人 | 需要打通康复医疗技术规则与人形机器人护理用例 | 中 | 高 | 以现有医疗技术能力作为底座 | 索取组织架构图、公告机构关系,以及质量 / 法规岗位空缺 |
| 安全验证与现场可靠性工程 | 非结构化环境中的人形机器人使用,需要不同于监督式康复的证据 | 高 | 极高 | 约束早期用例,并记录所有事件 | 索取验证方案、现场故障分类和分阶段推出计划 |
| 医疗企业销售和报销专家 | 医疗服务方 ROI 取决于编码、利用率和采购证明 | 高 | 高 | 把设备销售与工作流、结果证据绑定 | 索取分客群赢单 / 输单分析、付款方结构和续约动作 |
| 具身 AI 软件与数据工程 | 人形机器人扩规模,需要的软件、云和非结构化环境能力,不是康复机电一项就够 | 中高 | 高 | 围绕仿真、安全和模型运营招聘并建立合作 | 索取模型治理、遥操作 / 数据闭环和云架构责任归属图 |
| 全球服务与客户成功运营 | 覆盖 40 个国家意味着支持、培训和备件复杂度 | 中 | 中高 | 标准化现场服务和远程诊断 | 索取按区域划分的支持覆盖图、SLA 条款和升级指标 |
执行行聚焦可能在监管或需求之前打断看似可信产品故事的组织瓶颈。
[CR002, CR015, CR017, CR025, CR032, CR039]监管、安全和隐私失误会如何传导为采用变慢、利润率变弱和融资压力。
这张图展示传导逻辑,不是实测弹性。
[CR022, CR026, CR027, CR039, CR044, CR045]7.3 出口管制、成本和竞争强度可能压缩高端医疗人形机器人策略的窗口
即便 Fourier 解决了产品就绪度,仍要穿过一个越来越奖励规模、低价格和合规深度的地缘政治与竞争环境。美国出口管制继续把涉华先进算力置于筛查、许可和最终用途纪律之下;即便 2026 年 BIS 对特定芯片有所放松,也仍以安全控制、合规程序和第三方测试为条件。中国自身的人形机器人标准体系进一步抬高本地门槛,把安全、伦理、应用和数据 / 模型治理要求延伸到全生命周期。与此同时,商业窗口正在收紧。TrendForce 预计中国 2026 年产量增长 94%,Unitree 和 AgiBot 份额接近 80%;TechCrunch 和 CNBC 则描述了一个在高成本、软件、安全和泡沫压力下,从演示驱动的热度转向运营驱动采用的市场。这对 Fourier 不利,因为其高端角度依赖医疗健康可信度,而公开证据仍显示人形机器人客户证明有限,估计价格点也远高于供应更广的低成本替代品。[CR030, CR031, CR032, CR033, CR034, CR035]
| 依赖项 | 交易对手 / 监管制度 | 角色 | 集中度 | 失效场景 | 严重性 | 缓释措施 | 剩余暴露 |
|---|---|---|---|---|---|---|---|
| 先进 AI 算力 | NVIDIA / BIS 管制芯片栈 | 支撑高端具身 AI 训练和部署工作流 | 中高 | 许可证延迟或管制收紧拖慢路线图并抬高成本 | 高 | 备选供应商和本地算力栈可能逐步降低暴露 | 高,因公开缓释细节很薄 |
| 中国相关出口筛查 | EAR、筛查清单、最终用途管制 | 为跨境发货、服务和交易对手把关 | 高 | 交易筛查失效会造成发货延迟、许可证被拒或合规行动 | 高 | 强出口方控制、客户筛查和最终用途尽调 | 中高 |
| 医院 / 医疗服务方工作流负责人 | 治疗师、文档负责人、采购团队 | 批准、记录并把康复使用落地 | 高 | 临床热情无法转化为可报销、可复用的使用 | 高 | 工作流匹配和数据采集可能改善采用 | 高,直到站点经济性披露 |
| 临床与研究合作伙伴 | Brooks、NHG Health、Nagoya、峰会生态 | 提供证明点、验证和市场路径可信度 | 中 | 合作停留在试点形态,无法转成规模化付费生产 | 中高 | 用装机和联合研究撬动具名商业案例 | 中高 |
| 经销商和服务合作伙伴 | 区域渠道和支持组织 | 提供本地覆盖、服务和客户入口 | 中高 | 服务质量或利润截留不透明,会削弱续约质量和客户可见度 | 中高 | 收紧服务治理并披露直接客户 | 中高 |
| 资本市场基准重置 | 人形机器人同行融资和市场情绪 | 影响融资条款和估值支撑 | 中 | 同行规模和资金更足的对手压缩 Fourier 的容错空间 | 高 | 靠医疗细分定位和真实康复经济性做出差异 | 人形机器人证据仍薄时为高 |
本登记表把监管方、医院、芯片和资本市场都视为依赖项,因为每一项都能直接改变商业化的节奏和成本。
[CR022, CR025, CR030, CR031, CR036, CR037]Fourier 的高溢价风险路径穿过监管方、芯片、服务商、合作伙伴,以及一组快速扩张的竞争对手。
依赖项按运营卡点分组,不覆盖每个供应商或客户关系。
[CR030, CR031, CR036, CR037, CR041, CR042]7.4 正确监测点不是“还没有争议”,而是 Fourier 能否把康复可信度转成披露的合规部署
监测集应聚焦 Fourier 能否把康复可信度转成已披露、合规的运营证明。一个具名付费人形机器人部署,且任务指标清楚,比又一次会议亮相更重要。逐型号公布 FDA、CE 或 NMPA 策略,比又一次生活方式照护机器人演示更重要。披露安全包、DPA、事件响应流程和上市后计划,比泛泛而谈 AI 更重要。康复侧,旗舰现场的硬利用率和报销证据,也比漏斗顶端的机构数量更重要。这一框架还保留了围绕公开争议的重要细微差别:本章没有留存点名 Fourier 的铁证式执法行动或公开召回,但没有浮出水面的争议,并不等于安全或合规画像已经去风险。公共监测系统和召回规则存在;缺的,是 Fourier 具体披露其质量、网络、隐私和监管控制在规模化时会如何运转。在这个披露包出现前,风险应通过具体触发器监测,而不是从沉默中获得安慰。[CR020, CR021, CR025, CR027, CR041, CR042]
| 风险 | 可监控触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 人形机器人商业化证明缺口 | 具名付费部署披露 | 至少一个具名 GR-1 或 GR-3 客户,确认付费状态,并披露 KPI 或 ROI 细节 | 明显改善承保判断;若缺席,人形机器人仍只是期权,而非基准情景增长 |
| 护理用途监管不透明 | 按型号公开的档案或路径 | 发布 CE / FDA / NMPA 分类、预期用途边界,或贴近患者使用的时间表 | 降低监管尾部风险;若持续缺席,应采用更慢的采用假设 |
| 安全和产品责任冲击 | 召回、NCAR、FSCA 或严重事件通知 | 任何需要公开纠正措施或广泛现场通知的患者安全事件 | 立即触发下行;重审部署假设,并为服务 / 补救成本预留 |
| 康复报销转化风险 | 医疗服务方利用率和拒付证据 | 标杆站点显示稳定利用率、可接受的 KX 拒付率,以及复购或续约行为 | 支撑康复业务韧性;若失败,说明工作流收益没有转化为持久经济性 |
| 出口管制与算力依赖 | 芯片或许可证导致发货或训练中断 | 新限制或供应商中断让路线图或发货延迟超过一个季度 | 重设发布节奏、利润率假设和资金需求 |
| 竞争性价格压缩 | 可比对手在成本和可及性上超过 Fourier | 低成本对手广泛可得,而 Fourier 仍缺少溢价证明 | 高端医疗人形机器人逻辑削弱,估值支撑应压缩 |
触发项按可观察性选择:可通过公开披露、采购证据、监管通知或明确客户案例跟踪。
[CR022, CR027, CR036, CR043, CR044, CR045]7.5 证据要点
08估值
8.1 公开锚点有用,但单靠它不足以支撑高确信度入场
对 Fourier 的正确解读,不是 RMB8 billion 投前锚点显然错误,而是公开证据太薄,还无法高确信度判断它便宜、公允还是昂贵。Runyang 的文件给了本章一个许多私人机器人故事缺少的东西:真实备案的价格锚点、支票规模,以及稀释后持股测算。这很重要。2025 年 1 月的 Series E 也显示,背后仍有新资本胃口,投资者范围较宽,包括上海国资相关资本和知名私人支持者。但同一份公开记录也显示仍缺什么。这个估值锚点没有收入、毛利率、烧钱速度、现金余额或现金跑道披露。即便是完全稀释口径的测算也需要解读,因为这笔投资同时伴随 ESOP 补充和战略背景,而不是一笔干净为外部投资者给普通股定价的标准公平交易融资。两者合在一起,得到的纪律性结论是:把该文件视为新钱的天花板测试,而不是证明这家公司已经凭运营基本面配得上 RMB8 billion 存在权。对能做私下尽调的战略内部人来说,当前水平可能仍公允。对只能依赖公开证据的外部投资者来说,建议实质上是继续研究或观察,置信度中等,风险评级高。[CV001, CV002, CV003, CV004, CV005, CV006]
| 维度 | 当前判断 | 判断依据 | 投资含义 |
|---|---|---|---|
| 投资建议 | 继续研究 / 观察 | 公开证据支持其有真实业务和真实融资锚,但经济性不足,难以有把握地承保价格 | 没有私有尽调访问权时,不要追着公开锚点加价 |
| 置信度 | 中 | 核心不确定性来自经营披露缺失,而非公开事实彼此矛盾 | 若收入和利润率数据浮出,保留修正空间 |
| 风险评级 | 高 | 人形机器人期权有价值,但公开经济性不透明,情景分布很宽 | 要求下行保护和明确止损触发项 |
| 估值立场 | 对内部投资人合理到偏高;对新外部投资人偏高 | 战略内部投资人可能比公开投资者更了解收入质量和交割条款 | 外部投资者不应假设自己拥有内部视野 |
| 公开锚点 | 可作为有用的上限测试,但不是干净的公允价值证明 | Runyang 文件是真实的,但背景包含战略资本和同步 ESOP 稀释 | 把 RMB8bn 投前估值当作参考点,而非充分的承保答案 |
| 回报门槛 | 3x 需要约 RMB24bn 退出价值 | 该结果需要重大收入披露和更强的人形机器人商业化证明 | 不要基于当前锚点把 3x 建成基准情景 |
本表有意保留判断色彩:它从公开证据归纳投资结论,而不是复述公司质量卖点。
[CV001, CV003, CV006, CV007, CV035, CV036]| 维度 | 投资逻辑 | 反向逻辑 | 什么会改变判断 |
|---|---|---|---|
| 康复业务 | 真实康复布局很可能给 Fourier 提供商业化底线 | 公开投资者仍看不到收入质量、毛利率或续约韧性 | 披露分部收入、毛利率和头部客户集中度 |
| 人形机器人项目 | 在康复之外增加上行期权和战略相关性 | 公开证明仍远弱于美国叙事龙头,可能还不足以支撑溢价定价 | 披露具名付费部署和任务级运营指标 |
| 资本基础 | Series E 和 Runyang 显示其具备可信融资能力 | 锚点来自战略融资背景,而非透明经营表现 | 展示最终交割条款、投后估值和股权结构瀑布 |
| 可比公司框架 | Fourier 介于康复设备和人形机器人平台两类之间 | 这种中间位置也可能意味着它吃到两类折扣,却拿不到任一类的完整溢价 | 说明 Fourier 凭什么获得可持续的倍数区间,而不只是叙事期权 |
| 入场纪律 | 如果隐藏经济性强,当前锚点可以辩护 | 缺少经济性披露时,新投资者承保的不只是业务质量,也是在承保管理层不透明 | 要么改善披露,要么要求更便宜的入场价格 |
每一行都说明什么才会真正改变判断;没有这些变化,反向逻辑仍然有效。
[CV006, CV014, CV015, CV020, CV021, CV022]建议来自几股力量的抵消:Runyang 文件提供真实价格锚,康复证据也真实存在,但经济性缺失、人形机器人证据有限、可比公司估值跨度很宽。
流程是分析性的,不是字面流程;它把估值决策从证据质量映射到投资建议。
[CV001, CV006, CV023, CV036, CV037, CV040]8.2 公开康复同行框定地板,私人人形机器人轮次框定叙事天花板
Fourier 的估值问题在于,公开类比公司指向两个很不同的方向。一边,Ekso、Lifeward 等上市康复和外骨骼公司显示,当收入规模不大、亏损持续且投资者能直接看到数字时,公开市场会非常不宽容。这些公司证明,真实产品、报销进展和真正商业化,并不会自动换来高估值倍数。另一边,CYBERDYNE,尤其是 UBTECH,又说明机器人公司一旦上市、财务可见,并被视为更广平台而非狭窄康复设备厂商,就能获得明显更高倍数。私人人形机器人轮次进一步拉大差距。Figure 跟踪到的 $39 billion Series C,以及 Apptronik 在 2026 年 2 月约 $5.3 billion 到 $5.5 billion 的估值,显示当投资者承销平台可选性而非当前经济性时,美国具身 AI 故事能被定价得多激进。CNBC 的报道把机制说得很清楚:美国初创公司被按 AI 平台估值,而中国公司更常被按工业硬件企业估值,即使中国公司出货更多硬件。因此,Fourier 的公开锚点夹在两个世界之间。只用承压的公开康复可比公司来解释它,估值太高;把它读成 Figure 式前沿人形机器人叙事估值,又远远太低。最干净的解释不是单一赛道可比捷径,而是“康复商业化 + 期权价值”的混合框架。[CV008, CV009, CV010, CV011, CV012, CV013]
| 可比对象 | 可观察估值 | 收入 / 倍数快照 | 与 Fourier 的相关性 | 局限 |
|---|---|---|---|---|
| Fourier 公开锚点 | RMB8bn 投前;隐含 ~RMB8.56bn 完全稀释 | 未披露公开收入或利润率 | 该公司唯一直接公开价格锚点 | 战略背景和 ESOP 稀释削弱其作为市场出清价格的干净度 |
| Ekso Bionics | ~US$42m 市值 | 2025 收入 US$12.8m;~3.9x EV / 收入;53% 毛利率 | 展示规模仍有限时,已披露康复机器人能按什么水平交易 | 公开股权价值很大程度受流动性和二级市场悲观情绪影响 |
| Lifeward | ~US$18m 市值 | FY2025 收入 ~US$22.0m;~0.9x EV / 收入 | 是有可见经济性的康复机器人公司的有用底部参照 | 估值带困境属性,不是更健康资产的干净公允价值目标 |
| CYBERDYNE | ~JPY55bn 市值 | FY2026 收入 JPY3.846bn;~10.5x EV / 收入 | 展示有披露和平台叙事的康复相邻上市机器人公司的上沿 | 日本上市倍数包含市场结构和本地投资者影响 |
| UBTECH | ~HK$55bn 市值 | 2024 收入 ~US$0.17bn;~22.1x EV / 收入 | 对已披露的人形机器人 / 服务机器人叙事而言,这是最佳上市公开标尺 | 产品和公开市场叙事远宽于 Fourier 当前证据集 |
| Apptronik | ~US$5.3–5.5bn 投后 | >$935m Series A;累计融资近 US$1bn | 有用的私人市场人形机器人叙事可比对象,低于 Figure,高于多数公开同行 | 私募市场标记;收入未公开披露 |
| Figure | US$39bn 投后 | US$1bn Series C;累计融资 US$1.75bn | 展示美国具身 AI 叙事定价的当前上沿 | 纯叙事天花板,不是 Fourier 当前现实的基准可比对象 |
可比集合有意混合:公开康复同行框定底部,上市机器人平台框定已披露的溢价区间,私有人形机器人轮次框定叙事天花板。表内数值不做汇率标准化,因为目的在于框定区间,而非合成精度。
[CV001, CV003, CV008, CV009, CV010, CV011]在选定销售额倍数下,为支撑 RMB8bn 股权价值锚点所需的大致年收入;假设净债务有限,并把股权价值粗略当作 EV 的代理。
敏感性分析把 RMB8bn 的投前锚点粗略当作企业价值代理,因为公开材料未披露债务和超额现金。它是估值纪律工具,不是预测。
[CV024, CV025, CV026, CV027, CV028, CV029]8.3 乐观、基准、悲观框架显示,当前锚点接近今天公开证据所能支撑的上沿
在这里,受证据约束的情景模型比堆满表格的 DCF 更诚实。悲观情景假设康复业务真实,但不足以扛住估值;人形机器人项目仍有战略兴趣,却没有转化成具名付费部署或已披露经济性证明。在这种世界里,公开倍数压缩和新一轮折价可能把公允价值推向大约 RMB4.5 billion 至 RMB6.0 billion。基准情景假设公开证据对 Fourier 康复深度的判断方向正确,人形机器人仍保留有意义的期权价值,但仍不给公司拒绝披露的经济性记任何分。这得到大约 RMB6.5 billion 至 RMB8.5 billion 的区间;值得注意的是,当前公开锚点已经接近该区间上沿。乐观情景需要的不只是乐观情绪。它需要已披露收入和利润率质量,需要证明康复需求耐久而非只靠 标识堆砌,还需要具名人形机器人部署来缩小与更高倍数机器人公司的可比性缺口。只有到那时,RMB9.5 billion 至 RMB13.0 billion 区间才合理。这个框架也解释了,为什么从当前起点出发,很难为 3x 风投式回报门槛辩护。大约 RMB24 billion 退出并非不可能,但今天的公开证据不足以把它建模为基准情景。[CV024, CV025, CV026, CV027, CV028, CV029]
| 情景 | 估值范围(RMB bn) | 核心假设 | 关键缺失证据 | 行动含义 | 概率信号 |
|---|---|---|---|---|---|
| 悲观 | 4.5–6.0 | 康复业务仍真实,但规模小于隐含预期;人形机器人证明多停留在宣传层面;新资金重定价风险 | 无公开收入、无具名付费人形机器人部署、无清晰融资交割确认 | 避免按当前锚点付价;等待重置或披露 | 若下一轮融资带结构化条款或折价,该情景信号明显 |
| 基准 | 6.5–8.5 | 康复商业化实质真实,人形机器人保留期权价值,但公开披露没有改善 | 分部经济性、现金跑道、续约质量 | 观察 / 继续研究;不要高于锚点付价 | 与当前公开记录最一致 |
| 当前公开锚点 | 8.0 投前 / ~8.56 完全稀释 | Runyang 文件真实,Series E 历史支持其持续获得资本 | 最终交割条款、优先股堆叠和当前经济性仍缺失 | 只作为价格参考;不是充分承保证明 | 已接近基准情景支撑上沿 |
| 乐观 | 9.5–13.0 | 披露的收入和利润率强度、持久康复续约、具名付费人形机器人部署,支撑更高倍数待遇 | 需要证明上行来自经营,而不只是叙事 | 只有披露或里程碑证明落地后才上调 | 需要多个新事实,而不只是另一条融资标题 |
区间是受证据约束的分析师估计,不是 DCF 输出。Fourier 没有披露收紧模型所需的经营数据,因此有意避免虚假精确。
[CV003, CV031, CV032, CV033, CV034, CV035]| 触发项 | 阈值 / 事件 | 对估值的传导 | 行动含义 |
|---|---|---|---|
| Runyang 锚点未能交割或重定价 | 最终交易低于已披露条款交割、明显延迟,或被结构化资本替代 | 削弱唯一直接公开价格锚点,并暗示隐藏弱点 | 立即下调;重置估值框架 |
| 下一次重大融资前仍无经济性披露 | 又一轮大额融资到来,但仍不披露收入、利润率或现金跑道 | 说明管理层仍预期投资者按叙事、而非经济性定价 | 不要相对当前锚点支付溢价 |
| 人形机器人证明仍不具名 | 下一个重大里程碑窗口内,没有具名付费人形机器人部署或任务级指标 | 期权价值仍属投机,不应承载溢价倍数 | 将估值维持在基准或悲观区间内 |
| 康复经济性披露后一旦令人失望 | 披露的分部利润率、续约质量或客户集中度显著差于隐含预期 | 移除 Fourier 区别于纯叙事同行的商业化底线 | 视为重大下行重置 |
| 出现降价轮或厚重优先权堆叠 | 新融资定价低于锚点,或引入强势高级权益 | 即便企业叙事不变,普通股回报也会被压缩 | 回避,或按更低普通股价值重新定价 |
触发项指应改变投资测算的可观察事件,不是风险章节已覆盖的泛化业务风险。
[CV034, CV035, CV042, CV043, CV044]受证据约束的估值带(RMB bn),展示当前公开锚点相对悲观、基准和乐观情景的位置。
所有数值都是分析师区间,不是管理层指引。当前锚点区间有意同时纳入 Runyang 文件的投前和摊薄后解读。
[CV003, CV031, CV032, CV033, CV034, CV035]IC 风格评分卡显示,Fourier 最强的点是康复商业化和资本可得性;披露质量和人形机器人证据仍是估值支撑的硬约束。
这些分数是分析师在 0–10 尺度上的序位判断,用来概括投资论证质量,而不是假装能把这门生意压成一个确定性模型。
[CV036, CV037, CV038, CV039, CV040]8.4 本章决策规则很简单:缺失运营事实变成披露证据前,不要高于锚点出价
这里最重要的估值纪律是程序性的。Fourier 不需要再拿一份叙事材料来改善判断;它需要一小组运营披露,让投资者能把它放进真实估值框架里。第一道门是财务可见度:收入、毛利率、烧钱速度和现金跑道。第二道门是融资质量:Runyang 交易是否真的按公开描述条款完成,以及普通股之上有什么优先股堆叠或反稀释保护。第三道门是商业化质量:具名付费人形机器人部署、客户集中度、续约行为,以及康复销售耐久而非只是覆盖面广的证据。在这些事实可见之前,建议应保持价格敏感和证据敏感。也就是说,不要在公开锚点之上做高确信度承销,也不要把虚假精确的 DCF 强加到一家不透明私人公司上。上行情景足够真实,公司原则上仍可投资,尤其因为康复业务给了它许多人形机器人初创公司没有的地板。但公开记录仍支持等待更好披露或更有吸引力的入场价格,而不是把当前锚点视为自证有效。[CV036, CV037, CV038, CV039, CV040, CV041]
| 主题 | 缺失证据 | 重要性 | 尽调路径 |
|---|---|---|---|
| 收入和毛利率 | 当前年度收入、康复 / 人形机器人分部拆分、毛利率和贡献利润率 | 用于把 Fourier 放进真实估值倍数区间,剔除伪精确猜测 | 索取经审计或管理层确认的 KPI 包,以及客户 / 队列桥接 |
| 现金跑道和烧钱速度 | E 轮和 Runyang 之后的现金余额、月度烧钱和营运资本状况 | 看不清现金跑道,投资人就无法判断该锚点买到的是时间还是表面观感 | 索取月度现金桥接和 12–18 个月计划 |
| Runyang 交割条款 | 已披露投资是否完全交割、交割日期,以及条款是否有变 | 本章的公开价格锚点取决于该交易真实且可持续 | 索取已签署交割文件或更新后的股权结构表备忘录 |
| 优先股堆叠 | 后续投资人的清算优先权、反稀释条款和高级权益 | 普通股回报可能远差于企业价值测算显示的结果 | 审阅融资文件和律师摘要 |
| 人形机器人商业化证据 | 具名付费部署、合同规模、续约条款和任务 KPI | 用于证明人形机器人项目不只是期权价值 | 访谈参考客户,并核验合同级部署证据 |
| 康复客户质量 | 大客户集中度、复购节奏和续约行为 | 用于判断康复业务是稳固底盘,还是只是 logo 覆盖面广 | 分部队列分析和前 20 大账户审阅 |
这些尽调问题按最快改变估值判断的能力排序。财务披露排在最前,因为其他估值论点都要建立在它之上。
[CV041, CV042, CV043]8.5 证据要点
免责声明
本报告仅供研究与尽调使用,不构成投资建议。本报告基于截至 2026-05-21 可获得的公司公开材料、融资披露、客户案例和独立报道。Fourier 是未上市公司,关键经营指标仍未披露,因此作出任何投资决定前,应在一手尽调中验证估值和风险结论。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Fourier was founded in 2015. | 高 | SO002, SO013 |
| CO002 | Fourier’s publicly listed headquarters address is in Pudong, Shanghai, China. | 高 | SO003, SO005 |
| CO003 | Fourier says its mission is to leverage full-stack robotics technology to enrich people’s lives. | 高 | SO001, SO005 |
| CO004 | The company says it evolved from exoskeleton robots into a portfolio spanning humanoid robots and intelligent rehabilitation systems. | 高 | SO002, SO001 |
| CO005 | Official company materials say the product lineup includes the GRx humanoid series, RehabHub, the Galileo system, and open-source solutions. | 高 | SO002, SO001 |
| CO006 | Fourier says its products serve more than 2,000 institutions across more than 40 countries and regions. | 高 | SO002, SO018 |
| CO007 | Fourier’s public contact page lists offices in Shanghai, Singapore, and Malaysia. | 中 | SO003 |
| CO008 | Fourier’s official company history says it established its first international office in Singapore in 2018. | 中 | SO002 |
| CO009 | Alex Gu is identified in official company materials as the founder of Fourier. | 高 | SO002, SO024 |
| CO010 | Alex Gu is described publicly as founder and CEO or founder and group CEO of Fourier. | 高 | SO005, SO013 |
| CO011 | Zen Koh was publicly described as co-founder and group deputy CEO in 2022 funding coverage. | 中 | SO013 |
| CO012 | Fourier’s January 2022 Series D raised RMB400 million and was led by SoftBank Vision Fund 2 with backing from Prosperity7 and Yuanjing Capital. | 中 | SO013, SO016 |
| CO013 | The 2022 Series D was described as funding go-to-market expansion, healthcare robotics innovation, and global growth. | 中 | SO013 |
| CO014 | Chinese media reported that Fourier completed a nearly RMB800 million Series E financing in January 2025. | 中 | SO014, SO015, SO016, SO023 |
| CO015 | Late-May 2025 coverage of Runyang’s planned strategic investment put Fourier’s pre-money valuation at RMB8 billion. | 中 | SO020, SO022, SO024 |
| CO016 | Chinese media coverage described Fourier’s cumulative financing since founding as roughly RMB1 billion. | 中 | SO020, SO021 |
| CO017 | Fourier said in its GR-2 launch announcement that the company had over 500 employees globally. | 中 | SO005 |
| CO018 | Fourier’s official company history says the humanoid robot project was launched in 2019. | 中 | SO002 |
| CO019 | Fourier’s official company history says the first generation of GR-1 was born in 2022. | 中 | SO002 |
| CO020 | Official company materials describe GR-1 as being introduced in 2023 and as the first mass-produced humanoid robot in the industry. | 高 | SO002, SO004 |
| CO021 | Official GR-2 materials describe the robot as 175 centimeters tall, 63 kilograms, and equipped with up to 53 degrees of freedom. | 高 | SO005, SO010, SO012 |
| CO022 | Fourier says GR-2 adds optional 12-DoF dexterous hands, tactile sensing, new FSA 2.0 actuators, and mainstream developer-framework support. | 高 | SO005, SO010, SO012 |
| CO023 | Fourier unveiled GR-3 in August 2025 as its first care-centric full-size humanoid robot. | 高 | SO007, SO017 |
| CO024 | GR-3 was described as 165 centimeters tall, 71 kilograms, up to 55 degrees of freedom, and equipped with multimodal perception plus hot-swappable batteries. | 高 | SO007, SO017, SO018 |
| CO025 | TMTPost reported that Fourier’s rehabilitation products were deployed in over 2,000 institutions across more than 40 countries and regions. | 中 | SO017 |
| CO026 | TMTPost reported that overseas revenue from Fourier’s rehab division grew more than 50% in 2023 and accounted for about 10% of total revenue. | 低 | SO017 |
| CO027 | The January 2022 funding coverage said Fourier had more than 1,000 institutional healthcare customers and 17 international research joint labs. | 中 | SO013 |
| CO028 | Fourier Rehab’s June 2025 MOU with Brooks Rehabilitation covered deployment of ArmMotus EMU and ExoMotus M4 at Brooks’ Center for Innovation in Florida. | 中 | SO008 |
| CO029 | Fourier Rehab’s May 2025 MOU with Kurage aimed to integrate NeuroSkin with Fourier rehab devices and target a Medica 2025 launch. | 中 | SO009 |
| CO030 | Gasgoo reported in January 2026 that Fourier had partnered with more than 20 leading universities, research institutes, and technology companies worldwide. | 中 | SO018 |
| CO031 | An Eastmoney feature said GR-1 deliveries had exceeded 100 units by September 2024. | 低 | SO024 |
| CO032 | Fourier’s GR-2 support documentation says the platform has experimental applications in guidance, academic research, industrial manufacturing, and medical rehabilitation. | 中 | SO010 |
| CO033 | Fourier’s support site publishes both GR-2 documentation and the Fourier ActionNet dataset, evidencing a developer-facing full-stack tooling strategy. | 中 | SO010, SO011 |
| CO034 | Fourier’s public website architecture separates the general robotics brand 'Fourier' from the rehab-focused sub-brand 'Fourier Rehab'. | 高 | SO001, SO002, SO006 |
| CO035 | Official Fourier pages display the ICP record number 沪ICP备16009987号-1, tying the public site to a Shanghai-registered online presence. | 高 | SO001, SO002, SO003 |
| CO036 | No reviewed public source disclosed Fourier’s current revenue, ARR, or gross margin figures. | 低 | SO002, SO005, SO015, SO020 |
| CO037 | No reviewed public source disclosed Fourier’s board composition, debt facilities, or secondary transaction terms. | 低 | SO002, SO013, SO020 |
| CO038 | Official and media materials together show rehabilitation robotics as Fourier’s installed-base business and humanoids as the newer growth narrative layered on top. | 中 | SO002, SO013, SO017 |
| CO039 | The Robot Report wrote that competitive pressure and consolidation would persist in Chinese robotics in 2026 as investors demanded real commercial applications. | 中 | SO025 |
| CO040 | Gasgoo and PR Newswire coverage shows Fourier used CES 2026 to showcase GR-3 internationally alongside a smaller companion concept. | 中 | SO018, SO019 |
| CM001 | Fourier Intelligence's addressable market spans two primary segments—rehabilitation robotics and humanoid/embodied AI robotics—with adjacent exposure to broader healthcare service robotics but excluding logistics, hospitality, and surgical robotics. | 高 | SM001, SM002, SM003 |
| CM002 | The International Federation of Robotics formally classifies powered exoskeletons (wearable robots) as a distinct sub-category of medical robots—separate from industrial robots and non-wearable service robots. | 中 | SM003 |
| CM003 | The broader professional service robotics market (Fortune estimate USD 31.11B in 2026, dominated by transport/logistics at 102,900 units) is not additive to Fourier's TAM; Fourier does not compete in logistics, hospitality, or cleaning robot segments. | 中 | SM020, SM002 |
| CM004 | In 2019, across the Americas, 366 million people had health conditions amenable to rehabilitation—a 58% increase from 231 million in 1990—driven by population growth and aging. | 中 | SM008 |
| CM005 | Global stroke prevalence reached 93.8 million cases in 2021; stroke is the second leading cause of death and the third leading cause of combined death and disability globally. | 中 | SM009 |
| CM006 | The estimated global annual economic cost of stroke exceeds USD 890 billion (0.66% of global GDP) and is projected to nearly double by 2050. | 中 | SM009 |
| CM007 | Global stroke incident cases rose 70% between 1990 and 2021; 87% of stroke deaths and 89% of DALYs reside in lower- and lower-middle-income countries. | 中 | SM009 |
| CM008 | IFR's World Robotics 2025 report shows global medical robot unit sales rose 91% to approximately 16,700 units in 2024; rehabilitation and non-invasive therapy robot sales specifically rose 106% year-over-year. | 高 | SM002, SM001 |
| CM009 | IFR identifies staff shortages and the growing elderly population as the two key demand drivers for professional service robots in 2025–2026, including medical robots. | 高 | SM001, SM002 |
| CM010 | Oliver Wyman projects global healthcare spending rising from USD 11.8 trillion today to USD 23.1 trillion by 2040 without productivity intervention; AI and robotics could remove up to USD 5.1 trillion of projected annual cost increase in a breakthrough scenario. | 中 | SM022 |
| CM011 | A global caregiver shortage estimated at 15 million workers is cited as a primary structural driver for AI-powered and robotic elderly care solutions. | 中 | SM023 |
| CM012 | The likelihood of rehabilitation need rises exponentially with age; aging global populations are the primary long-run structural demand driver for rehabilitation robotics. | 中 | SM008, SM022 |
| CM013 | TBRC estimates the global rehabilitation robotics market at USD 2.22 billion in 2026, growing at a 22.7% CAGR to USD 4.87 billion by 2030; end-users are rehabilitation centers and hospitals. | 中 | SM016 |
| CM014 | Mordor Intelligence estimates global rehabilitation robots market at USD 1.77 billion in 2026 at 17.1% CAGR to USD 3.89 billion by 2031; exoskeleton robots represent 47.35% of market share by type as of 2025. | 中 | SM017 |
| CM015 | IMARC Group estimates the global rehabilitation robots market at USD 1.82 billion in 2025, growing at a 16.53% CAGR to USD 7.5 billion by 2034; key drivers include aging population, rising disability prevalence, and technological advancement. | 中 | SM018 |
| CM016 | Fortune Business Insights estimates the global rehabilitation robots market at USD 579.5M in 2026 at 16.36% CAGR; North America accounted for approximately 74.76% of global revenue in 2025, making it the dominant regional market. | 中 | SM019 |
| CM017 | Rehabilitation centers account for approximately 53.2% of rehabilitation robots market revenue by end-user; homecare is the fastest-growing channel at 27.1% CAGR through 2031. | 中 | SM017 |
| CM018 | Asia-Pacific is the fastest-growing geography for rehabilitation robotics at 21.2% CAGR driven by rising stroke incidence and aging populations; North America retains the largest revenue share. | 中 | SM017, SM019 |
| CM019 | Analyst estimates for global rehabilitation robotics in 2026 span from USD 579.5M (Fortune) to USD 2.22B (TBRC)—a 3.8× range—reflecting definitional scope differences between exoskeleton-only and full therapeutic device universe including bundled services. | 中 | SM016, SM017, SM018, SM019 |
| CM020 | Geriatric and elderly users account for approximately 61.4% of rehabilitation robots market demand, making the aging demographic the dominant patient cohort by revenue share. | 中 | SM017 |
| CM021 | Fourier Intelligence is explicitly named as a key player in the global rehabilitation robotics competitive landscape by TBRC, confirming the company's commercial market participation. | 中 | SM016 |
| CM022 | Future Market Insights estimates the humanoid robot market at USD 10.69 billion in 2026 growing at 37% CAGR to USD 248.9 billion by 2036; healthcare leads application segments with 28% of 2026 market share. | 中 | SM015 |
| CM023 | TBRC estimates the global humanoid robot market at USD 8.32 billion in 2026 growing at 52.9% CAGR to USD 39 billion by 2030; applications include personal assistance and caregiving. | 中 | SM024 |
| CM024 | Roland Berger projects the OEM-level humanoid robot market at USD 300 billion (baseline) to USD 750 billion (optimistic) by 2035, scaling further to USD 2–4 trillion by 2050, driven by a USD 2 per hour projected operating cost. | 中 | SM012 |
| CM025 | Goldman Sachs data cited by BusinessWire reports humanoid robot manufacturing costs fell 40% year-over-year—far exceeding the 15–20% prior projection—with current unit costs ranging from USD 30,000 to USD 150,000. | 中 | SM021 |
| CM026 | Unitree launched its R1 humanoid at USD 5,900 in July 2025, establishing the first sub-USD 10,000 consumer-grade price point in humanoid robotics and demonstrating a consumer affordability tier. | 中 | SM021 |
| CM027 | TrendForce projects China humanoid robot annual output will grow 94% in 2026, with Unitree and AgiBot accounting for approximately 80% of total shipments as the market shifts from demonstration to delivery. | 中 | SM013 |
| CM028 | Roland Berger notes that very few productive use cases for humanoid robots exist today and that the software/ecosystem layer trails hardware maturity by three to five years, with supply-chain and safety-certification gaps remaining as of 2026. | 中 | SM012, SM014 |
| CM029 | FMI and TBRC estimates for the humanoid robot market in 2026 diverge by approximately USD 2.37 billion (USD 10.69B vs. USD 8.32B), reflecting different scope inclusions, geographic coverage, and methodology vintage. | 中 | SM015, SM024 |
| CM030 | CMS finalized the CY2026 OPPS/ASC payment rule in November 2025 with a +2.6% update rate, affecting approximately 4,000 hospital outpatient departments and 6,000 ambulatory surgical centers, providing incremental budget headroom for rehabilitation technology. | 高 | SM004, SM007 |
| CM031 | CMS' 2024 Home Health Rule (effective January 1, 2024) reclassified powered exoskeletons within the Medicare brace benefit category, creating the first general reimbursement pathway for personal exoskeleton devices for home use. | 高 | SM007, SM006, SM004 |
| CM032 | CMS proposed a payment rate of approximately USD 94,617 for personal exoskeleton devices under the brace benefit category; Medicare beneficiaries are responsible for a 20% copay, and home use requires training completion at a certified clinical site. | 高 | SM007, SM004 |
| CM033 | Lifeward reports approximately 40% of U.S. spinal cord injury patients are now covered by payers that have reimbursed personal exoskeleton claims, combining Medicare, VA benefits, and workers' compensation coverage as of April 2024. | 中 | SM005 |
| CM034 | A 2026 JMIR umbrella review of robot-assisted therapy for upper-limb rehabilitation after stroke found improvements in motor function, but activities of daily living (ADL) gains were generally comparable to conventional therapy—limiting the strongest payer-approval argument. | 中 | SM011 |
| CM035 | A 2026 PubMed RCT found exoskeleton-assisted rehabilitation improved Fugl-Meyer lower- extremity scores and Barthel Index versus conventional training for post-stroke gait rehabilitation, supporting clinical utility for lower-limb exoskeleton applications. | 中 | SM010 |
| CM036 | IFR's 2026 trends report states humanoid robots must match high industrial requirements in cycle times, energy consumption, and maintenance costs—and that key capability gaps remain—before they can compete with traditional automation. | 高 | SM001, SM002 |
| CM037 | Robotics & Automation News identifies manipulation, power and endurance, and general-purpose autonomy as the three dominant technical bottlenecks for humanoid robots; most 2026 deployments remain supervised pilot or co-pilot applications. | 中 | SM014 |
| CM038 | High upfront capital costs (USD 30K–USD 150K for humanoids; USD 1–2B market implies significant per-unit ASP barriers) and mixed long-term outcome evidence temper rehabilitation robotics procurement, especially in pediatric and emerging-market settings. | 中 | SM017, SM021 |
| CM039 | The JMIR umbrella review concludes that ADL gains from robot-assisted upper-limb therapy after stroke are generally comparable to conventional therapy, directly weakening the evidence-based case for superior clinical efficacy. | 中 | SM011 |
| CM040 | Roland Berger's 2026 report confirms the software and ecosystem layer for humanoid robots trails hardware maturity by three to five years; regulatory gaps, supply-chain fragility, and safety-certification frameworks remain incomplete as of April 2026. | 中 | SM012 |
| CM041 | Fortune Business Insights and Mordor Intelligence both cite high initial capital expenditure and maintenance/integration costs as the primary restraints on service and rehabilitation robotics adoption across healthcare settings. | 中 | SM019, SM017 |
| CM042 | Knowledge Sourcing Intelligence identifies data security and privacy concerns, AI accuracy risks, and anxiety about replacing human connection in care settings as material barriers to AI-robotic elderly care adoption. | 中 | SM023 |
| CM043 | Knowledge Sourcing estimates the elderly care robotics market (companion, mobile service, and robotic exoskeleton segments) at USD 0.39 billion in 2026 growing to USD 0.64 billion by 2031 at a 10.4% CAGR. | 中 | SM025 |
| CM044 | The primary status-quo substitutes for rehabilitation robots in clinical settings include licensed physiotherapists and occupational therapists providing manual therapy, non-powered body-weight support treadmill systems, and standard non-powered orthotic devices. | 中 | SM011, SM017 |
| CP001 | Fourier publicly presents itself as both a rehabilitation-robotics company and a GR-series humanoid company rather than a single-product vendor. | 高 | SP001, SP002, SP003 |
| CP002 | Fourier Rehab says it serves more than 2,000 institutions across over 40 countries. | 中 | SP001 |
| CP003 | Fourier says RehabHub integrates more than 30 rehabilitation robot types into one digital workflow. | 中 | SP001 |
| CP004 | Fourier markets GR-1 as a mass-produced humanoid robot and general-purpose assistant. | 中 | SP002 |
| CP005 | Fourier GR-1's public specification page lists a 165 cm height, 55 kg weight, 44 joints, and 5 km/hr top speed. | 中 | SP002 |
| CP006 | Fourier markets GR-3 as a care-oriented robot with soft shell materials, multimodal interaction, touch sensing, and emotional-response features rather than factory-task proof. | 中 | SP003 |
| CP007 | Mordor Intelligence's rehabilitation-robots competitor list names Cyberdyne, Ekso Bionics, ReWalk Robotics, Fourier Intelligence, and Wandercraft among leading companies. | 中 | SP027 |
| CP008 | Wandercraft's public materials and recent coverage show a dual roadmap spanning the Atalante X rehab exoskeleton, the Eve personal exoskeleton, and the Calvin-40 humanoid. | 高 | SP004, SP005 |
| CP009 | EU-Startups reports that Wandercraft's Atalante X is deployed in over 100 hospitals and rehabilitation centres globally. | 中 | SP005 |
| CP010 | Wandercraft closed a €64.3 million Series D and says Renault became both a strategic investor and the first commercial customer for Calvin-40. | 中 | SP005 |
| CP011 | Ekso states that EksoNR is FDA-cleared for stroke, acquired brain injury, multiple sclerosis, and spinal cord injury and is backed by more than 200 published articles. | 中 | SP006 |
| CP012 | Ekso Indego Therapy is marketed as a lower-limb powered exoskeleton that can be fitted in under five minutes and used across outpatient and wellness settings. | 中 | SP007 |
| CP013 | Public reimbursement reporting indicates that ReWalk Personal and Ekso Indego Personal were the two immediate beneficiaries of Medicare's established personal-exoskeleton payment rate, with US$91,032 cited as the established rate. | 高 | SP013, SP011 |
| CP014 | CYBERDYNE's HAL platform differentiates itself through bio-electrical-signal control and spans medical, well-being, labor-support, and at-home services. | 中 | SP008 |
| CP015 | Lifeward publicly positions itself as a continuum-of-care rehabilitation company with ReWalk Personal, ReStore Exo-Suit, AlterG systems, and MYOLYN products. | 高 | SP010, SP012 |
| CP016 | Lifeward says ReWalk was the first personal exoskeleton to receive Medicare coverage for an eligible beneficiary. | 中 | SP011 |
| CP017 | Lifeward says nearly 40% of people with spinal cord injury are covered by payers that have reimbursed medically eligible personal-exoskeleton claims when Medicare, VA, and workers' compensation are combined. | 中 | SP011 |
| CP018 | Lifeward's investor relations page describes the company as Nasdaq-listed and operating in the United States, Israel, and Germany. | 中 | SP012 |
| CP019 | The clearest direct rehabilitation-robotics peer set for Fourier in public sources is Wandercraft, Ekso Bionics, CYBERDYNE, and Lifeward. | 中 | SP027, SP004, SP006, SP008, SP010 |
| CP020 | UBTECH describes itself as an HKEX-listed full-stack humanoid and smart-service-robot company spanning industrial manufacturing, smart elderly care, business service, and consumer applications. | 高 | SP014, SP016 |
| CP021 | UBTECH says Walker is the first commercialized biped life-sized humanoid robot in China and that Walker S has entered multiple automotive production lines. | 中 | SP014 |
| CP022 | UBTECH says Walker S2 entered mass production and delivery, with the first batch consisting of several hundred robots and a target of 500 units delivered within the year. | 中 | SP015 |
| CP023 | UBTECH says it is targeting annual capacity of 5,000 industrial humanoid robots by 2026 and 10,000 by 2027. | 中 | SP015, SP030 |
| CP024 | UBTECH says Walker-series orders accumulated since early 2025 exceed 800 million yuan and name BYD, Geely, FAW-Volkswagen, BAIC, Foxconn, and SF Express among collaborators. | 中 | SP015 |
| CP025 | Unitree publicly prices G1 from US$13.5K and describes it as a humanoid platform for developers and advanced users, with roughly two hours of battery life. | 中 | SP017 |
| CP026 | TrendForce projects that Unitree and AgiBot together will account for nearly 80% of China's humanoid shipments in 2026. | 中 | SP018 |
| CP027 | TrendForce says Unitree's 2025 prospectus shows humanoid robots surpassed quadrupeds as the company's largest revenue segment and that Unitree plans 75,000 humanoid robots of annual capacity. | 中 | SP018 |
| CP028 | BusinessWire's ResearchAndMarkets summary says Unitree's R1 launched at US$5,900, creating a much lower public humanoid price anchor than most peers. | 中 | SP028 |
| CP029 | AGIBOT's official site shows a broad embodied-robot portfolio across A2, X, G, D, and accessory lines for service, industrial, research, and data applications. | 中 | SP019 |
| CP030 | At CES 2026, AGIBOT said it had shipped 5,000 robots to date and already had mass-produced humanoids deployed across eight commercial application categories. | 中 | SP020 |
| CP031 | Independent 2026 reporting says AgiBot hit a 10,000-robot milestone after scaling from 5,000 to 10,000 units in only three months, underscoring unusually fast Chinese ramp speed. | 中 | SP030, SP018 |
| CP032 | 1X positions NEO Gamma as a home humanoid emphasizing internal home testing, passive safety, quieter hardware, and teleoperation-friendly autonomy rather than industrial throughput. | 中 | SP021 |
| CP033 | 1X's official investor-relations surface confirms the company is actively courting investors, but its retained official sources do not disclose the kind of public deployment counts seen at UBTECH or Figure. | 低 | SP021, SP022 |
| CP034 | Figure's homepage frames Figure 03 as a general-purpose home-help robot, and the company news index says it was ramping Figure 03 production in April 2026. | 高 | SP023, SP024 |
| CP035 | Figure's official news index also states that the company exceeded US$1B in Series C funding at a US$39B post-money valuation in September 2025. | 中 | SP024 |
| CP036 | Figure says its BMW deployment ran 10-hour weekday shifts, loaded more than 90,000 parts over 1,250+ hours, and contributed to the production of 30,000+ BMW X3 vehicles. | 高 | SP025, SP026 |
| CP037 | BMW says the Spartanburg humanoid pilot delivered measurable value under real-world conditions and is informing wider Physical AI deployment, but the proof point is automotive production rather than healthcare or eldercare. | 高 | SP026, SP025 |
| CP038 | Robotics & Automation News says humanoid robots still face major commercial gaps in dexterous manipulation, endurance, and general-purpose autonomy outside controlled settings. | 中 | SP029 |
| CP039 | IFR says humanoid enthusiasm is being amplified by hype and social-media demos while real-world adoption timing and regional commercialization paths remain uncertain. | 中 | SP031 |
| CP040 | China's 2026 humanoid standards push reflects rapid scaling but also unresolved safety, tactile-sensing, and minimum-risk-condition challenges as robots move closer to people. | 中 | SP030 |
| CP041 | Fourier's clearest differentiation versus industrial-first humanoid vendors is its existing clinical distribution and care-oriented product narrative. | 中 | SP001, SP003, SP020, SP023, SP025 |
| CP042 | In U.S. home-use exoskeleton reimbursement, Lifeward and Ekso currently have clearer public pathways than Fourier, Wandercraft, or CYBERDYNE. | 中 | SP011, SP013, SP004, SP008 |
| CP043 | Switching costs in rehabilitation robotics are meaningful but not absolute because vendors emphasize clinician training, modular fitting, software analytics, and protocol integration rather than irreversible hardware lock-in. | 中 | SP001, SP007, SP010 |
| CP044 | Fourier's main weakness versus humanoid peers is thinner public evidence on GR-series commercial deployments, pricing, and independently verified ROI. | 中 | SP002, SP003, SP015, SP025, SP026 |
| CP045 | Public competitor data are not directly comparable because companies disclose heterogeneous signals such as institutions, reimbursement rates, order backlog, runtime hours, funding rounds, or shipment counts instead of a common market-share metric. | 中 | SP001, SP005, SP015, SP018, SP024, SP025, SP030 |
| CI001 | Fourier raised RMB400 million in a January 2022 Series D financing. | 中 | SI006, SI009 |
| CI002 | SoftBank Vision Fund 2 led the 2022 Series D, with Prosperity7 Ventures and Yuanjing Capital also named in coverage. | 中 | SI006, SI009 |
| CI003 | Series D proceeds were described as funding go-to-market expansion, advanced healthcare-robotics innovation, and global growth. | 中 | SI006 |
| CI004 | The 2022 Series D coverage said Fourier’s revenue nearly tripled in 2021 despite pandemic conditions. | 中 | SI006 |
| CI005 | Fourier’s January 2025 Series E financing was reported at nearly RMB800 million. | 中 | SI007, SI008, SI009 |
| CI006 | Named Series E backers included Guoxin Investment, Pudong Venture Capital, Zhangjiang-linked funds, Prosperity7, and Junshan or PEAKVEST. | 中 | SI007, SI008, SI009 |
| CI007 | Series E reporting said Fourier would keep advancing humanoid robots and wider commercial scenarios under the GRx roadmap. | 中 | SI007, SI009 |
| CI008 | FL Cube separately framed Series E proceeds as support for continuous robotics innovation and deployment. | 低 | SI008 |
| CI009 | Runyang Technology publicly disclosed a planned investment of up to RMB300 million into Fourier at an RMB8 billion pre-money valuation. | 高 | SI010, SI011, SI012 |
| CI010 | The Runyang filing shows simultaneous ESOP issuance: after the financing, Runyang would hold 3.5060% and the new ESOP vehicle 3.0000% on a fully diluted basis. | 高 | SI010, SI011 |
| CI011 | The Runyang filing says the post-transaction board would have 11 seats, with founders nominating 6 and major investors such as SoftBank, Prosperity7, Yuanjing, Wuxi Bio-linked capital, and SME fund representatives one each. | 中 | SI010 |
| CI012 | The Runyang filing describes Fourier as a 600-plus employee company spanning humanoid robotics and intelligent rehabilitation solutions, with 30-plus product series serving 2,000-plus hospitals or institutions in 40-plus countries. | 中 | SI010 |
| CI013 | Fourier’s official GR-3 and RehabHub pages use contact-led calls to action rather than posted transaction pricing, indicating quote-based institutional selling. | 中 | SI003, SI004 |
| CI014 | Fourier’s official rehab materials position RehabHub as a multi-stage rehabilitation platform integrating more than 30 robot types, implying hardware-plus-platform monetization rather than a single-device model. | 中 | SI002, SI003 |
| CI015 | Fourier’s official GR-3 materials position the product for care-oriented, human-centric enterprise scenarios rather than consumer retail. | 中 | SI001, SI004 |
| CI016 | Fourier’s official product pages and download center expose brochures and product families but no audited revenue, ARR, gross margin, or cash disclosures. | 中 | SI001, SI002, SI003, SI004 |
| CI017 | TMTPost reported that overseas revenue from Fourier’s rehab division grew more than 50% in 2023 and accounted for about 10% of total revenue. | 低 | SI005 |
| CI018 | TMTPost reported that GR-3 is priced above RMB200,000 and is targeted at healthcare institutions and enterprise clients rather than mass consumers. | 中 | SI005 |
| CI019 | Founder commentary in TMTPost said manufacturing costs remain high because humanoid components are complex, even if scale should lower cost over time. | 中 | SI005 |
| CI020 | Europa Satellite’s structured product metadata lists the Fourier GR-3 at EUR106,344.61 with price validity through 2026-06-20. | 中 | SI013 |
| CI021 | A 120Shop procurement page records 2024 winning bids of RMB280,000 and RMB347,000 for Fourier’s M2D02SP upper-limb rehabilitation robot. | 中 | SI014 |
| CI022 | The same 120Shop page also records larger Fourier rehab-system procurement anchors of RMB1.07 million and RMB1.368 million and reports a tracked average of RMB1.28 million, low of RMB0.97 million, and high of RMB1.42 million. | 中 | SI014 |
| CI023 | Unitree publicly lists the G1 humanoid at a starting price of US$13.5K. | 中 | SI015 |
| CI024 | The gap between Unitree’s US$13.5K G1 and Fourier’s six-figure distributor or procurement anchors suggests Fourier must defend a higher-value care and rehab niche rather than compete on broad humanoid commoditization. | 中 | SI013, SI014, SI015 |
| CI025 | The Robot Report’s summary of Interact Analysis says high prices and dexterity gaps are likely to slow humanoid adoption into the next decade. | 高 | SI016, SI027 |
| CI026 | Interact Analysis still forecasts the humanoid market could exceed 40,000 units and about US$2 billion of revenue by 2032, but only after current barriers erode. | 高 | SI016, SI027 |
| CI027 | Bain says most humanoid deployments remain early-stage and heavily reliant on human supervision in structured environments. | 中 | SI017 |
| CI028 | Bain says most humanoids operate for about two hours today and a full eight-hour shift could take up to 10 years or longer to achieve. | 中 | SI017 |
| CI029 | Ekso Bionics reported full-year 2024 revenue of US$17.9 million, gross margin of about 53%, net cash used in operations of US$9.8 million, and year-end cash plus restricted cash of US$6.5 million. | 高 | SI018, SI019 |
| CI030 | Ekso’s Q4 2024 revenue was US$5.1 million with a 53% gross margin and US$1.4 million of net cash used in operations for the quarter. | 中 | SI018 |
| CI031 | Lifeward reported full-year 2025 revenue of US$22.0 million, gross margin of 38.2%, and operating loss of US$19.7 million. | 高 | SI020, SI021 |
| CI032 | Lifeward’s Q4 2025 revenue was US$5.1 million, with GAAP gross margin of 21.4%, adjusted gross margin of 32.6%, and operating loss of US$5.1 million. | 中 | SI021 |
| CI033 | Lifeward attributed margin pressure to lower sales volumes, under-absorption of fixed manufacturing overhead, and higher tariffs and freight expense. | 中 | SI021 |
| CI034 | Lifeward said its hybrid direct-plus-channel sales model and expanded reimbursement capabilities are central to growth but are not yet fully reflected in current revenue. | 中 | SI021 |
| CI035 | CYBERDYNE reported FY2026 revenue of ¥3,846 million, operating loss of ¥601 million, and year-end cash and cash equivalents of ¥8,991 million. | 高 | SI022, SI023 |
| CI036 | CYBERDYNE said it would not issue an FY2027 consolidated forecast because uncertainty in this new market makes accurate forecasting difficult. | 中 | SI023 |
| CI037 | Lifeward says the finalized Medicare reimbursement pathway for personal exoskeletons is still medically eligibility-based and that nearly 40% of spinal-cord-injury patients are insured by reimbursing payers. | 中 | SI024 |
| CI038 | Sheltering Arms and VPM show that personal exoskeletons cost about US$91,000 to US$100,000 and require roughly 40 to 50 hours of training, underscoring service intensity beyond the device bill of materials. | 中 | SI025, SI026 |
| CI039 | VPM reported that Medicare covers about 80% of the just-over-US$91,000 device price, leaving the patient to secondary insurance or out-of-pocket expense. | 中 | SI025 |
| CI040 | Across the reviewed public record, Fourier still does not disclose current revenue, ARR, gross margin, monthly burn, cash balance, or runway. | 中 | SI001, SI002, SI003, SI004, SI005, SI006, SI007, SI010 |
| CI041 | Because the public valuation anchor is clearer than any disclosed earnings metric, Fourier is presently easier to underwrite as a financing story than as a statement-backed operating model. | 中 | SI009, SI010, SI011 |
| CI042 | Public evidence supports a mixed hardware-plus-services-plus-platform model, but realized software attachment, service gross profit, and recurring revenue share remain undisclosed. | 中 | SI003, SI004, SI021 |
| CI043 | Public Fourier price anchors already span from low-six-figure renminbi devices to low-seven-figure institutional rehab systems, reinforcing enterprise procurement economics rather than consumer turnover. | 中 | SI005, SI014 |
| CI044 | Peer evidence shows that even roughly 38% to 53% gross margins can coexist with sizable annual operating losses or operating cash burn in rehabilitation robotics. | 中 | SI018, SI021, SI023 |
| CI045 | Without disclosed cash or runway, Fourier’s next financing trigger is more likely to be tied to humanoid scale-up, working-capital needs, and commercial proof than to visible self-funding from current operations. | 中 | SI005, SI010, SI018, SI021, SI023 |
| CI046 | Debt facilities, liquidation preferences, project-finance obligations, and customer-concentration metrics are not disclosed in the reviewed public set. | 中 | SI010, SI020, SI021 |
| CE001 | GR-2 public support documentation describes the robot as 1.75 meters tall, 63 kilograms, and up to 53 degrees of freedom, using FSA 2.0 actuators and self-developed dexterous hands. | 中 | SE005 |
| CE002 | GR-2 documentation says the head integrates visual perception, voice interaction, and speaker modules. | 中 | SE005 |
| CE003 | GR-2 documentation says each arm has 7 degrees of freedom, each leg has 6 degrees of freedom, and the standard versus enhanced hands expose 6 versus 12 degrees of freedom per hand. | 中 | SE005 |
| CE004 | GR-2 documentation says the robot has already reached experimental applications in guidance and consultation, academic research, industrial manufacturing, and medical rehabilitation. | 中 | SE005 |
| CE005 | Retained 2025–2026 GR-3 launch coverage describes GR-3 as Fourier’s first full-size care-focused humanoid and a central part of its CES 2026 international showcase. | 中 | SE003, SE004 |
| CE006 | Retained GR-3 launch coverage says GR-3 stands about 165 centimeters tall and has 55 degrees of freedom. | 中 | SE003, SE004 |
| CE007 | Retained GR-3 launch coverage says the robot uses a full-perception multimodal interaction system spanning visual, audio, and tactile sensing. | 中 | SE003, SE004 |
| CE008 | Retained GR-3 public sources primarily show CES interaction demos such as conversation, chess, dance, and touch-based interaction rather than named production deployments. | 中 | SE003, SE004 |
| CE009 | N1 documentation describes a 1.3-meter, 38-kilogram humanoid with 23 body degrees of freedom, 3.5 m/s maximum stable running speed, 15°–20° slope handling, and 20-centimeter stair traversal. | 中 | SE006 |
| CE010 | N1 documentation says the head supports expansion for externally installed vision sensors such as stereo cameras. | 中 | SE006 |
| CE011 | The DexHand documentation says Fourier’s five-finger hand has 6 built-in linear actuators, 6 active degrees of freedom, 5 passive degrees of freedom, and 11 movable joints. | 中 | SE008 |
| CE012 | The DexHand documentation says the device supports Ethernet control, 12V–60V supply, 6N average fingertip force, and 40kg maximum whole-hand load. | 中 | SE008 |
| CE013 | Fourier’s teleoperation overview describes a complete upper-limb imitation-learning pipeline covering data collection, training, and an open-source dataset. | 中 | SE007 |
| CE014 | The teleoperation overview says ActionNet contains more than 30,000 teleoperation trajectories for dexterous bimanual tabletop manipulation tasks. | 中 | SE007 |
| CE015 | The teleoperation repo says Fourier’s system enables real-time humanoid control through VR headsets and hand tracking and includes data-collection features for imitation learning and VLA-style work. | 中 | SE012 |
| CE016 | The teleoperation repo lists Apple Vision Pro and Meta Quest 3 headsets, DepthAI Oak and Intel RealSense cameras, and 6-DOF and 12-DOF hands among supported configurations. | 中 | SE012 |
| CE017 | The teleoperation repo warns that first-time hardware setup can be tricky and recommends contacting Fourier support for hardware issues outside the tutorial scope. | 中 | SE012 |
| CE018 | Fourier’s GitHub organization page says the company actively contributes software, hardware designs, and research publications to the open-source community. | 中 | SE009 |
| CE019 | Fourier’s public GRx pipeline repo organizes model preparation, Isaac Gym training, MuJoCo verification, and deployment to real GRx robots as separate stages and repositories. | 中 | SE013 |
| CE020 | The GRx-Gym repo says Fourier uses NVIDIA Isaac Gym with legged_gym and rsl_rl to train N1 walking policies and export the resulting policy for real-robot deployment. | 中 | SE014 |
| CE021 | The GRx-MuJoCo repo says Fourier uses MuJoCo 3.x for N1 policy verification and visualization. | 中 | SE015 |
| CE022 | The GRx URDF model repo says Fourier provides kinematic and dynamic robot models, collision assets, and support for Isaac Gym and Isaac Sim across GR1, GR2, and GR3 variants. | 中 | SE016 |
| CE023 | The GRx MJCF repo provides a URDF-to-MJCF conversion tool with optional sensor configuration and MuJoCo node settings rather than a finished commercial simulator package. | 中 | SE017 |
| CE024 | The FSA actuator SDK exposes Linux and Windows C++ and Python examples over 48V power and Ethernet connectivity. | 中 | SE019 |
| CE025 | The DexHand SDK exposes Python and C++ interfaces over UDP and currently only supports Linux systems. | 中 | SE018 |
| CE026 | The public GRx client says Wiki-GRx-Deploy is deprecated, requires alpha-version compatibility, and is highly unstable and subject to change. | 中 | SE020 |
| CE027 | NVIDIA says Fourier used Isaac Gym to develop and test GR-2 and is porting those workflows to Isaac Lab. | 中 | SE021 |
| CE028 | NVIDIA says Fourier used simulation for complex multi-robot scenarios and pretraining of grasping algorithms before real deployment. | 中 | SE021 |
| CE029 | NVIDIA says a GR-2 floor-to-stand policy ran 3,000 iterations in around 15 hours and transferred to physical controls with 89% success. | 中 | SE021 |
| CE030 | Fourier Rehab’s official materials say the company serves more than 2,000 institutions across more than 40 countries. | 中 | SE002 |
| CE031 | Fourier Rehab’s official materials say RehabHub integrates more than 30 types of robots inside a fully digital training environment. | 中 | SE002 |
| CE032 | The retained RehabHub brochure describes the platform as a one-stop concept spanning robotics therapy, neuromodulation, sensor technology, virtual reality, and functional electrical stimulation. | 中 | SE022 |
| CE033 | The retained RehabHub brochure says linked devices allow one therapist to monitor multiple patients rather than only one-to-one training. | 中 | SE022 |
| CE034 | The retained RehabHub brochure says the system requires about one minute to set up and is efficient enough to train 15 patients daily. | 中 | SE022 |
| CE035 | The retained RehabHub brochure says Fourier devices use an in-house force-feedback algorithm, force and position sensors, and auto-generated reports to adjust therapy. | 中 | SE022 |
| CE036 | STEPS says HandyRehab uses eight individual motors and EMG-sensor integration to support passive, active-assistive, and functional hand training. | 中 | SE024 |
| CE037 | STEPS says ArmMotus EMU uses a back-drivable cable-driven mechanism and lightweight carbon-fiber rods to reduce friction and inertia and improve human-machine compliance. | 中 | SE024 |
| CE038 | Ascot describes RehabHub as part of a consultant-led multidisciplinary rehabilitation pathway with more than ten robotic and interactive devices on site and therapist monitoring of quality, safety, and progression. | 中 | SE023 |
| CE039 | Hobbs says its outpatient center installed ArmMotus EMU, ArmMotus M2 Pro, and ExoMotus M4 as part of becoming a Fourier RehabHub site. | 中 | SE025 |
| CE040 | Hobbs says ExoMotus M4 combines an in-house motion control unit, multidimensional force sensor, and high-performing motors to support walking assistance. | 中 | SE025 |
| CE041 | 36Kr reports Fourier and Shanghai International Medical Center will jointly build a demonstration base for embodied-intelligence rehabilitation and co-create application standards for rehab scenarios. | 中 | SE026 |
| CE042 | 36Kr reports the GRx series has cumulative shipments in the hundreds and downstream application scenarios spanning AI R&D companies, factory manufacturing, reception services, and dangerous-goods protection. | 低 | SE026 |
| CE043 | The retained public record supports commercialization proof for rehabilitation systems but only mixed pilot-stage proof for Fourier’s humanoid line. | 中 | SE002, SE003, SE023, SE024, SE025, SE026 |
| CE044 | Fourier’s most evidence-backed differentiation is care and rehab adjacency built on installed base, multi-device workflow integration, and interaction design rather than publicly proven leadership in humanoid shipment scale or autonomy. | 中 | SE002, SE021, SE023, SE026 |
| CE045 | The retained public set does not verify named production customers, explicit regulatory approvals, or a detailed compliance pack for GR-3 care deployments. | 中 | SE003, SE004, SE026 |
| CE046 | 36Kr reports that healthcare deployment still faces technology, clinical-practice, supervision, and medical-license hurdles, and quotes management emphasizing effectiveness and medical-grade quality as prerequisites. | 中 | SE026 |
| CE047 | GR-3 public launch materials emphasize warmth, companionship, multimodal interaction, and showcase tasks but do not disclose detailed payload, battery life, runtime, or named buyer references in the retained set. | 中 | SE003, SE004 |
| CE048 | N1 docs and public repos show a meaningful developer surface, but the retained public stack still looks more like a research and integrator ecosystem than a mature commercial support platform. | 中 | SE006, SE009, SE011, SE020 |
| CE049 | At access time, the public GR-2 docs repo showed 198 commits, one visible star, and no releases published, suggesting documentation exists but external packaging and community traction remain early. | 低 | SE010 |
| CE050 | The N1 docs repo says it is documentation-only and that example code mainly lives elsewhere, which indicates public tooling breadth but also fragmentation across repos. | 中 | SE011 |
| CE051 | Fourier’s GR-1 product page describes GR-1 as the first mass-produced humanoid robot. | 中 | SE027 |
| CE052 | Fourier’s GR-1 product page lists 165 centimeter height, 55 kilogram weight, 44 joints, 5 km/h speed, and 230 Nm maximum peak torque. | 中 | SE027 |
| CE053 | Fourier’s GR-1 product page says FSA integrates motor, driver, reducer, and encoder into a single module and that the pure-vision stack uses six RGB cameras. | 中 | SE027 |
| CU001 | Fourier says its rehabilitation business serves more than 2,000 organizations and hospitals across more than 40 countries. | 高 | SU001, SU026 |
| CU002 | Fourier’s current rehabilitation page says RehabHub integrates more than 30 robot types inside a digital training environment. | 中 | SU001 |
| CU003 | The retained brochure describes an earlier Fourier Rehab base of more than 1,000 installations in more than 30 countries and 17 international research joint labs. | 中 | SU010 |
| CU004 | The brochure says linked RehabHub devices can cut therapist labor intensity, allow one therapist to monitor multiple patients, require about one minute of setup, and support roughly 15 patients per day. | 中 | SU010 |
| CU005 | The visible customer map spans institutional rehab centers, hospital rehab services, research and joint-lab partners, distributor-enabled clinics, and emerging care-humanoid pilots. | 中 | SU001, SU003, SU010, SU023 |
| CU006 | In the public rehab deployments, the buyer is usually the provider organization, therapists are the daily users, and patients are the beneficiaries inside a clinical pathway rather than the direct purchaser. | 中 | SU004, SU006, SU007, SU010 |
| CU007 | Public payer evidence suggests Fourier sells into provider capex or innovation budgets that still need reimbursable therapy workflows and documentation support rather than a simple device reimbursement path paid to the vendor. | 中 | SU017, SU024, SU025 |
| CU008 | CMS requires physician or NPP plan-of-care certification and documentation for outpatient rehabilitation therapy billing, so adoption has to fit billable therapy compliance rather than only device capability. | 高 | SU024, SU025 |
| CU009 | The NIH reimbursement guide says medical-device coverage, coding, payment, and patient out-of-pocket exposure should be addressed early, making payer strategy part of the buyer business case. | 中 | SU025 |
| CU010 | NHG Health and Fourier Rehab signed a new five-year MOU in May 2026 focused on advancing rehabilitation and robotic technologies. | 高 | SU002, SU013 |
| CU011 | NHG says the 2026 agreement builds on an initial 2021 MOU and a 2022 master research collaboration agreement. | 高 | SU002, SU012 |
| CU012 | NHG says selected Fourier rehabilitation technologies have already been deployed and evaluated for rehabilitation services at Tan Tock Seng Hospital. | 中 | SU002 |
| CU013 | The renewed NHG partnership covers inpatient, outpatient, and transitional care settings through a Joint Rehabilitation Innovation Hub focused on real-world validation and workflow integration. | 中 | SU002 |
| CU014 | Brooks Rehabilitation and Fourier say ArmMotus EMU and ExoMotus M4 are being deployed at Brooks’ Center for Innovation in Florida. | 高 | SU004, SU005 |
| CU015 | The Brooks relationship is framed as a research and protocol-development collaboration, not a disclosed commercial contract with public pricing, renewal, or utilization metrics. | 中 | SU004, SU005 |
| CU016 | STEPS publicly lists a live RehabHub with ArmMotus EMU, ArmMotus M2, AnkleMotus, CycleMotus, ExoMotus M4, and related devices inside a neurorehabilitation suite. | 中 | SU006, SU009 |
| CU017 | INPA says STEPS is home to one of only two ExoMotus machines in the UK and the only ArmMotus EMU. | 中 | SU009 |
| CU018 | STEPS and INPA describe patient use, connected games, and session-end reporting, giving public evidence of active workflow use rather than a logo-only reference. | 中 | SU006, SU009 |
| CU019 | Ascot describes itself as London’s only Fourier Centre of Excellence with more than 10 robotic and interactive rehabilitation devices on site inside a consultant-led multidisciplinary pathway. | 中 | SU007 |
| CU020 | Ascot explicitly positions RehabHub as part of a clinical pathway rather than as a standalone machine session, which clarifies the user and workflow owner structure. | 中 | SU007 |
| CU021 | Hobbs says its Laverstock Care Centre became a Fourier RehabHub via Thor and installed ArmMotus EMU, ArmMotus M2 Pro, and ExoMotus M4. | 中 | SU008 |
| CU022 | Hobbs and the brochure show that some UK deployments run through Thor as distributor and sales partner, so public customer proof can be channel-mediated rather than directly disclosed by Fourier. | 中 | SU008, SU010 |
| CU023 | Three Indian trade publications report that Saifee Hospital in Mumbai added ExoMotus M4 to its rehabilitation services. | 中 | SU017, SU018, SU019 |
| CU024 | Saifee’s Medical Director said ExoMotus M4 supports safe, personalized, data-driven gait training while reducing therapist effort in a high-volume tertiary hospital context. | 中 | SU017, SU018, SU019 |
| CU025 | Ren Ci Community Hospital and Day Rehab Centre ran an eight-session ExoMotus M4 adjunct program in which 47 patients were recruited and 30 completed the programme. | 中 | SU016 |
| CU026 | The Ren Ci pilot reported significant improvements in balance and mobility, no adverse events, and manhour savings when therapy assistants handled lower-complexity sessions. | 中 | SU016 |
| CU027 | Exoskeleton Report says GReAT 2026 delegates visited Xinchang Medical Center and Yong Ci Hospital, where Fourier technologies were presented in real-world clinical settings. | 中 | SU003 |
| CU028 | Exoskeleton Report and Gasgoo say Fourier signed 2026 MOUs with NHG Health and Nagoya University, and Gasgoo adds that the Nagoya work will use ArmMotus EMU for muscle-synergy analysis. | 中 | SU003, SU013 |
| CU029 | Neuro Rehab Times says more than 200 RehabHubs are in operation globally and names Shirley Ryan AbilityLab, STEPS Rehabilitation, and Yongchi Rehabilitation Hospital as examples. | 中 | SU010, SU011 |
| CU030 | The retained public set shows repeatability across Singapore, the US, the UK, India, and at least China/Japan research links, but the proof is uneven between direct customer pages and partnership/news coverage. | 中 | SU002, SU004, SU006, SU011, SU013, SU017 |
| CU031 | The strongest public customer proofs are rehab-site pages that name specific devices and workflow use, such as Brooks, STEPS, Ascot, Hobbs, Saifee, and Ren Ci. | 中 | SU004, SU006, SU007, SU008, SU016, SU017 |
| CU032 | Several visible relationships remain ecosystem or research partnerships rather than clearly disclosed paying customers, including NHG expansion, Nagoya collaboration, Kurage integration, and HTPG distribution. | 中 | SU002, SU013, SU022, SU023 |
| CU033 | HTPG’s 2026 announcement is a channel and demonstration hub that targets rehabilitation providers nationwide, not a named provider contract or end-customer revenue disclosure. | 中 | SU023 |
| CU034 | The Kurage deal is an ecosystem bundle pending clinical protocols and technical documentation for wider adoption, which makes it partnership proof rather than customer conversion proof. | 中 | SU022 |
| CU035 | The retained public set does not disclose NRR, GRR, churn, contract length, or renewal cohorts for Fourier’s customer base. | 低 | SU001, SU002, SU004, SU011, SU026 |
| CU036 | Public evidence is materially stronger on deployment announcements, pilot studies, and workflow anecdotes than on repeat-purchase economics or multiyear renewals. | 中 | SU004, SU009, SU016, SU017 |
| CU037 | Robocare and Healthlink pages show distributor-led regional channels for Fourier rehab devices, but neither page discloses named end-customer rosters or pricing. | 中 | SU020, SU021 |
| CU038 | Channel-led expansion can widen coverage while reducing visibility into who pays, what service obligations attach to each install, and whether a deployment is recurring or one-off. | 中 | SU020, SU021, SU023 |
| CU039 | robotics.press says Fourier’s rehabilitation franchise is credible but its humanoid pivot remains unproven with no verified paid industrial deployments or disclosed shipping volumes. | 中 | SU026, SU027 |
| CU040 | robotics.press says there are zero publicly confirmed paid deployments for either GR-1 or GR-3 and that pre-sale campaigns do not constitute commercial traction. | 中 | SU027 |
| CU041 | The same adverse analysis says healthcare-humanoid adoption faces elongated regulatory clearance and safety-led sales cycles, especially in clinical and eldercare settings. | 中 | SU027 |
| CU042 | Because public named-account disclosure is thin outside rehab while the installed-base claim is broad, outside investors cannot tell from public materials whether revenue is diversified or concentrated in a small number of large institutional buyers. | 中 | SU001, SU026, SU027 |
| CU043 | The 2021 NHG partnership already emphasized home-based rehabilitation, earlier patient care, telerehabilitation, and reduced community manpower, showing that labor efficiency has been part of the value proposition for years. | 中 | SU012 |
| CU044 | Fourier’s CMC Ludhiana visibility at IFNRCON 2025 is conference and market-development proof, not a disclosed production deployment inside the hospital. | 中 | SU014, SU015 |
| CR001 | Fourier’s current public surface presents GR-3 humanoids and RehabHub rehabilitation systems as part of one company mission, implying management is spanning two different regulatory and operating regimes at once. | 中 | SR001 |
| CR002 | Fourier’s rehabilitation business is publicly described as serving more than 2,000 institutions across over 40 countries, a scale claim repeated by both the company and independent robotics coverage. | 高 | SR002, SR027, SR028 |
| CR003 | Fourier says RehabHub integrates more than 30 robot types inside a fully digital training environment. | 中 | SR002 |
| CR004 | Fourier says RehabHub connects equipment, users, and institutions through a medical information system and is intended to reduce manual intervention and costs. | 中 | SR002 |
| CR005 | Fourier and Brooks both say ArmMotus EMU and ExoMotus M4 were part of the Brooks collaboration and intended for Brooks’ Florida innovation environment. | 高 | SR003, SR004 |
| CR006 | Brooks says ArmMotus EMU and ExoMotus M4 are in place and will support joint studies and scalable therapy protocols informed by clinical data. | 高 | SR004, SR003 |
| CR007 | Fourier’s CES launch materials describe GR-3 as a 165 cm, 55-DOF full-size care-focused humanoid using multimodal visual, audio, and tactile sensing. | 中 | SR005 |
| CR008 | Independent CES coverage says GR-3 is aimed at homes, public spaces, commercial environments, and assisted living, and that its interaction stack relies on face, movement, touch, and pressure sensing. | 中 | SR024, SR005 |
| CR009 | The EU AI Act is designed to protect health, safety, and fundamental rights while complementing existing Union data-protection law. | 中 | SR006 |
| CR010 | The joint AIB/MDCG guidance says MDR/IVDR and the AI Act apply simultaneously and complementarily when a medical device contains one or more high-risk AI systems. | 高 | SR009, SR006 |
| CR011 | The joint AIB/MDCG guidance says data governance, record-keeping, transparency, human oversight, accuracy, robustness, cybersecurity, and post-market monitoring all have to be integrated for medical-device AI. | 高 | SR009, SR030 |
| CR012 | As of March 2026, MDR/IVDR remains the sole current certification path for AI-enabled medical devices and no separate AI Act conformity assessment is yet required. | 高 | SR030, SR008 |
| CR013 | The published AI Act timing cited in 2026 guidance places Annex III high-risk obligations on 2 August 2026 and Article 6(1)/Annex I medical-device AI obligations on 2 August 2027. | 高 | SR030, SR009 |
| CR014 | The European Commission says MDR replaced the prior medical-device directives from 26 May 2021 and remains the governing framework for medical-device market access. | 高 | SR008, SR007 |
| CR015 | FDA says connected medical devices can be vulnerable to security breaches and that such breaches can affect device safety and effectiveness. | 中 | SR010 |
| CR016 | FDA says manufacturers and health care delivery organizations must work together to manage medical-device cybersecurity risk. | 中 | SR010 |
| CR017 | FDA’s current cybersecurity page ties section 524B, premarket submission content, threat modeling, and SBOM-related materials directly to device compliance expectations. | 中 | SR010 |
| CR018 | FDA says its recently approved devices page is incomplete by itself and that full diligence must also use PMA, 510(k), HDE, and De Novo databases. | 高 | SR011, SR012 |
| CR019 | FDA’s device databases page lists Devices@FDA, AccessGUDID, 522 surveillance, MAUDE, and MedSun as public surfaces for monitoring approvals and post-market signals. | 中 | SR012 |
| CR020 | FDA says device recalls are usually voluntary under 21 CFR 7, but FDA can order recalls under 21 CFR 810 and manufacturers/importers must report health-risk corrections and removals under 21 CFR 806. | 中 | SR015 |
| CR021 | The European Commission says vigilance systems, NCAR dissemination, field safety corrective actions, and Eudamed are core parts of EU device surveillance when incidents occur. | 中 | SR016 |
| CR022 | CMS says outpatient therapy claims above the 2026 KX thresholds must include the KX modifier to attest medical necessity supported by documentation, and claims over the threshold are denied without it. | 高 | SR013, SR014 |
| CR023 | CMS set the 2026 KX modifier thresholds at $2,480 for occupational therapy and $2,480 for combined physical therapy and speech-language pathology services. | 高 | SR013, SR014 |
| CR024 | CMS added RTM codes 98979, 98984, and 98985 as sometimes-therapy codes in the 2026 therapy code list, so monitoring support exists but still sits inside coding and documentation rules. | 中 | SR013 |
| CR025 | Fourier’s rehab value proposition depends on workflow integration, data capture, therapist planning, and efficiency claims, so demand quality still depends on whether those workflow gains survive reimbursement and utilization scrutiny. | 中 | SR002, SR013 |
| CR026 | Independent robotics coverage describes Fourier’s humanoid business as pre-commercial and lacking publicly verified paid deployments even while the rehab business is real and revenue-backed. | 中 | SR027, SR028 |
| CR027 | Adverse independent coverage says there are zero publicly confirmed paid deployments for GR-1 or GR-3, with no named customers, disclosed KPIs, or measurable ROI. | 中 | SR028 |
| CR028 | Independent product analysis says GR-2 is enterprise-only, priced at roughly $150,000 or more, and lacks a North American sales channel. | 中 | SR025 |
| CR029 | Humanoid.guide lists GR-2 at about $125,000, classifies availability as prototype, and says the product is not currently available for purchase. | 中 | SR026 |
| CR030 | TrendForce says China’s humanoid output is expected to grow 94% in 2026 and that Unitree and AgiBot could capture nearly 80% of shipments. | 中 | SR020 |
| CR031 | TrendForce says Unitree is targeting capacity of 75,000 humanoids annually and AgiBot has already reached its 10,000th embodied robot, signaling scale pressure ahead of most peers. | 中 | SR020 |
| CR032 | TechCrunch says the industry is shifting from demo-driven excitement toward operations-driven adoption and that durable demand depends on reliable repeatable value in production or service environments. | 中 | SR021 |
| CR033 | TechCrunch says current humanoid shipment counts remain tiny and that it is still unclear how many units represent commercial sales versus demos or pilot deployments. | 中 | SR021 |
| CR034 | CNBC reports that analysts still see high costs, reliance on U.S. chips, regulatory barriers, and AI limitations in unpredictable environments as major humanoid bottlenecks. | 中 | SR022, SR021 |
| CR035 | CNBC reports that China’s NDRC warned of a humanoid-robotics bubble risk because the country already has more than 150 companies launching many similar products. | 中 | SR022 |
| CR036 | BIS says exports of H200-, MI325X-, and similar-class chips to China are reviewed case by case and require security controls, compliance procedures, and independent U.S. testing. | 高 | SR017, SR018 |
| CR037 | Trade.gov says China-related EAR compliance requires classification, screening, end-use controls, and can lead to heightened scrutiny, Entity List exposure, or license requirements for advanced computing items. | 中 | SR018 |
| CR038 | China’s 2026 national humanoid standard system covers the full lifecycle with application, safety, ethics, data, and model-governance requirements, raising the local compliance floor as the market scales. | 中 | SR019 |
| CR039 | Legal commentary says responsibility for humanoid harm is still contested among manufacturers, operators, software providers, and end users, making contractual allocation of responsibility increasingly important. | 中 | SR029 |
| CR040 | Legal commentary says face, gesture, emotional-cue, and biometric processing in humanoids can conflict with GDPR and therefore require privacy-by-design, lawful collection, and transparent safeguards. | 中 | SR029, SR006 |
| CR041 | Company and exoskeleton-industry coverage both show Fourier is trying to extend from hospital and lab rehab settings toward community, home, and embodied-care scenarios, which broadens the compliance and support burden. | 中 | SR001, SR023 |
| CR042 | Exoskeleton Report shows NHG Health and Nagoya partnership activity centered on clinical translation and a Joint Rehabilitation Innovation Hub rather than disclosed recurring economics, reinforcing a partnership-led commercialization path. | 中 | SR023, SR002 |
| CR043 | Independent robotics coverage treats Fourier as a company to watch rather than underwrite aggressively because a 42-point Mosaic Score decline coincides with heavier humanoid marketing and still-thin public deployment proof. | 中 | SR028, SR027 |
| CR044 | The retained public evidence set surfaces rules, databases, and general liability expectations, but not a Fourier-specific hospital-grade security packet, DPA, or device-level regulatory dossier for care deployments. | 低 | SR010, SR012, SR029 |
| CR045 | The most decision-useful risk triggers are named paid humanoid deployments with KPI disclosure, model-by-model regulatory dossiers, disclosed safety or recall events, provider-side reimbursement/utilization proof, export-control disruptions, and evidence that premium pricing is still defendable against lower-cost rivals. | 中 | SR028, SR020, SR022, SR017, SR013, SR015 |
| CV001 | Runyang Technology’s June 2025 filing disclosed a planned RMB300 million investment into Fourier at an RMB8 billion pre-money valuation. | 高 | SV001, SV004 |
| CV002 | The same Runyang filing says Runyang would own 3.5060% of Fourier after the financing while a new ESOP vehicle would hold 3.0000% on a fully diluted basis. | 中 | SV001 |
| CV003 | Using Runyang’s disclosed percentage and check size implies a fully diluted post-close equity value of roughly RMB8.56 billion rather than a simple RMB8.30 billion pre-plus-cash figure. | 中 | SV001 |
| CV004 | Fourier, Grandall, Fineline Cube, and 36Kr all reported that the January 2025 Series E financing totaled nearly RMB800 million. | 高 | SV002, SV003, SV004, SV005 |
| CV005 | Public Series E reporting names Guoxin Investment, Pudong Venture Capital, Zhangjiang-affiliated capital, and private investors including Prosperity7 and Junshan Capital among the round’s backers. | 高 | SV002, SV003, SV004, SV005 |
| CV006 | Retained public sources disclose the valuation anchor and financing headlines but still do not disclose Fourier’s current revenue, gross margin, burn, or cash runway. | 中 | SV001, SV002, SV003, SV004, SV005 |
| CV007 | Because current operating economics are undisclosed, a clean DCF is not evidence-supported and the valuation framework has to rely on comparables and milestone probability instead. | 中 | SV001, SV002, SV003, SV004, SV005 |
| CV008 | Ekso’s 2025 Form 10-K reported $12.799 million of revenue, $6.845 million of gross profit, and a 53% gross margin for 2025. | 中 | SV012 |
| CV009 | Yahoo Finance showed Ekso at roughly $42 million market capitalization and 3.93x enterprise-value-to-revenue as of May 20, 2026. | 中 | SV013, SV030 |
| CV010 | Stock Analysis showed Lifeward at roughly $22.0 million of FY2025 revenue, while Yahoo Finance showed only about 0.90x enterprise-value-to-revenue as of May 20, 2026. | 中 | SV015, SV017 |
| CV011 | CYBERDYNE reported FY2026 revenue of ¥3.846 billion, and Yahoo Finance showed about 10.54x enterprise-value-to-revenue and 14.17x price-to-sales in May 2026. | 中 | SV019, SV020 |
| CV012 | UBTECH traded around HK$55–57 billion of market capitalization and about 22.12x enterprise-value-to-revenue in May 2026. | 中 | SV022, SV023 |
| CV013 | CompaniesMarketCap shows UBTECH generated roughly US$0.17 billion of revenue in 2024. | 中 | SV024 |
| CV014 | Ekso and Lifeward together show that real rehabilitation-robotics commercialization can still trade at low-single-digit or sub-1x observable revenue multiples when scale and profitability remain weak. | 中 | SV012, SV013, SV015, SV017 |
| CV015 | CYBERDYNE and UBTECH show that disclosed public robotics platforms can sustain double-digit revenue multiples when investors can observe listed financials and wider platform optionality. | 中 | SV019, SV020, SV022, SV023, SV024 |
| CV016 | Apptronik announced in February 2026 that it had raised more than $935 million of Series A capital and nearly $1 billion of total funding. | 高 | SV008, SV010 |
| CV017 | Sacra put Apptronik at roughly a $5.3–$5.5 billion post-money valuation in February 2026 and described the extension as roughly a 3x multiple to the initial Series A valuation. | 中 | SV010 |
| CV018 | Tracxn lists Figure’s September 16, 2025 Series C as a $1 billion round at a $39 billion post-money valuation and $1.75 billion of cumulative funding. | 高 | SV011, SV007 |
| CV019 | CNBC reported in April 2026 that Figure was valued at at least $39 billion and Apptronik at around $5 billion even though Chinese humanoid startups were shipping more units. | 中 | SV007 |
| CV020 | CNBC argued that U.S. humanoid startups are being priced as broad AI platforms while Chinese startups are being priced more as industrial hardware plays. | 中 | SV007 |
| CV021 | TechCrunch cited Rodney Brooks arguing that current humanoid funding is a bubble and that many expensive training efforts will never scale to mass production. | 中 | SV006 |
| CV022 | Fourier’s RMB8 billion anchor is far below the newest U.S. humanoid narrative rounds but far above distressed listed rehab-exoskeleton equity values. | 中 | SV001, SV007, SV010, SV011, SV013, SV017 |
| CV023 | The public anchor therefore looks more like a blended rehab-commercialization-plus-humanoid-option valuation than a pure frontier-humanoid narrative multiple. | 中 | SV001, SV007, SV010, SV011, SV014, SV015, SV019, SV024 |
| CV024 | If investors treated the RMB8 billion anchor as approximate enterprise value, it would require roughly RMB4.0 billion of annual revenue at 2x sales. | 中 | SV001, SV013, SV017, SV020, SV023 |
| CV025 | The same anchor would require roughly RMB2.0 billion of annual revenue at 4x sales. | 中 | SV001, SV013, SV017, SV020, SV023 |
| CV026 | The same anchor would require roughly RMB1.0 billion of annual revenue at 8x sales. | 中 | SV001, SV013, SV017, SV020, SV023 |
| CV027 | The same anchor would require roughly RMB0.67 billion of annual revenue at 12x sales. | 中 | SV001, SV013, SV017, SV020, SV023 |
| CV028 | The same anchor would require roughly RMB0.40 billion of annual revenue at 20x sales. | 中 | SV001, SV013, SV017, SV020, SV023 |
| CV029 | Because Fourier does not publicly disclose revenue, investors cannot tell whether the company is currently closer to the 1x-to-4x rehab band or the 10x-to-20x disclosed robotics narrative band. | 中 | SV001, SV012, SV013, SV015, SV017, SV019, SV020, SV023, SV024 |
| CV030 | The Runyang anchor is informative but not a perfectly clean price because it was struck in a strategic context and alongside a simultaneous ESOP top-up. | 中 | SV001 |
| CV031 | A bear case around RMB4.5 billion to RMB6.0 billion is consistent with real rehab commercialization but continued humanoid proof gaps and public-multiple compression. | 中 | SV012, SV015, SV019, SV021 |
| CV032 | A base case around RMB6.5 billion to RMB8.5 billion assumes Fourier’s rehab business is materially real and the humanoid program retains option value, but it does not assume new operating disclosure. | 中 | SV001, SV002, SV003, SV004, SV005, SV020, SV021 |
| CV033 | A bull case around RMB9.5 billion to RMB13.0 billion requires disclosed revenue and margin strength plus named paid humanoid deployments that narrow the gap to higher-multiple robotics peers. | 中 | SV007, SV010, SV011, SV020, SV023, SV024 |
| CV034 | On public evidence alone, the current RMB8 billion pre-money anchor already sits near the top of the base-case range rather than in the middle of an obvious upside case. | 中 | SV001, SV020, SV021 |
| CV035 | A 3x venture-style return from an RMB8 billion entry point would require an exit around RMB24 billion, which current public comparables do not support without major disclosure and commercialization progress. | 中 | SV001, SV010, SV011, SV020, SV023 |
| CV036 | The evidence-supported posture is research-more or track rather than buy if the only observable price reference is the Runyang filing. | 中 | SV001, SV006, SV007, SV012, SV015, SV019, SV023 |
| CV037 | For new outside minority investors, the current anchor should not be underwritten as a conviction entry price on public evidence alone. | 中 | SV001, SV006, SV023 |
| CV038 | The valuation stance is fair-to-stretched for strategic insiders who may see private operating data, but stretched for new investors without diligence access. | 中 | SV001, SV006, SV020, SV023 |
| CV039 | Confidence should remain medium because the core debate is driven by missing operating data rather than by multiple public data sets that disagree on disclosed results. | 中 | SV001, SV006, SV015, SV017, SV019, SV020, SV023 |
| CV040 | Risk should remain high because the current price is easier to defend as strategic optionality than as observed cash-generation power. | 中 | SV001, SV006, SV020, SV023 |
| CV041 | A clean upgrade would require audited or management-certified revenue, gross margin, backlog or order-book quality, and cash-runway disclosure. | 低 | |
| CV042 | A clean re-rating would also require confirmation that the Runyang investment actually closed on the publicly described terms and was not repriced or restructured. | 中 | SV001 |
| CV043 | Named paid humanoid deployments, rehab renewal quality, and customer concentration matter more for valuation than additional demo videos or conference appearances. | 中 | SV003, SV006, SV007 |
| CV044 | A new financing below the implied Runyang anchor or with heavy structure would be a thesis-break signal that the public anchor was too generous. | 中 | SV001, SV006, SV007 |
| CV045 | If future disclosure shows revenue scale and margins consistent with the higher-multiple CYBERDYNE or UBTECH bands, today’s anchor could prove conservative rather than stretched. | 低 | SV019, SV020, SV023, SV024 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Fourier | FOURIER-Robotics | |
| SO002 | Fourier | About Us | |
| SO003 | Fourier | Contact Us | |
| SO004 | Fourier | GRx Humanoid Robot Series | |
| SO005 | Fourier | Fourier Unveils the Next-Generation Humanoid Robot GR-2 | |
| SO006 | Fourier | Newsroom | |
| SO007 | Fourier | Fourier Unveils Its First Care-Centric Humanoid GR-3 | |
| SO008 | Fourier | Fourier Rehab and Brooks Rehabilitation Sign Strategic MOU to Advance Research in Robotic Neurorehabilitation | |
| SO009 | Fourier | Fourier Rehab and Kurage Announce Strategic Partnership to Advance Functional Rehabilitation Through Integrated NeuroSkin Technology | |
| SO010 | Fourier Support Center | About GR-2 | Fourier 文档中心 | |
| SO011 | Fourier Support Center | Fourier ActionNet 数据集 | Fourier 文档中心 | |
| SO012 | Fourier | GR2三折页_英文单页 | |
| SO013 | The Rakyat Post | Robotic Healthcare Company, Fourier Intelligence, Secures RM264 Million In Funding | |
| SO014 | Fineline Cube | Fourier Intelligence Secures USD 109 Million in Series E Funding for Robotics Innovation | |
| SO015 | 36Kr Europe | "Fourier" has completed a new round of financing, with the total amount reaching nearly 800 million yuan. | |
| SO016 | Sango Automation | Nearly 800 Million! General Robotics Company Fourier Officially Announced The E Series Of Financing | |
| SO017 | TMTPost | Fourier Launches First Humanoid Care Robot, Advancing Emotional AI in Robotics | |
| SO018 | Gasgoo | CES 2026: Fourier Intelligence unveils full-size humanoid robot GR-3 | |
| SO019 | PR Newswire | Fourier Makes CES Debut With GR-3, a Next-Generation Care-Focused Humanoid Robot | |
| SO020 | Pedaily / 投资界 | 上海跑出80亿超级独角兽 | |
| SO021 | DoNews | 上海傅利叶智能估值80亿,成通用机器人领域独角兽 | |
| SO022 | Tencent News | 人形机器人公司傅利叶成估值80亿元的“独角兽” | |
| SO023 | NetEase | 人形机器人企业傅利叶智能完成新⼀轮近8亿元融资|奥特|澎湃_网易订阅 | |
| SO024 | Eastmoney | 上市公司豪掷3亿,投了张江这家机器人独角兽“傅利叶”! | |
| SO025 | The Robot Report | Chinese robotics outlook for 2026 includes cobot growth, competitive pressure | |
| SM001 | International Federation of Robotics | Top 5 Global Robotics Trends 2026 | The global market value of industrial robot installations has reached an all-time high of US$ 16.7 billion. |
| SM002 | International Federation of Robotics | World Robotics 2025 report – SERVICE ROBOTS – released by IFR | Sales of rehabilitation and non-invasive therapy robots were up 106%. |
| SM003 | International Federation of Robotics | World Robotics 2025 — Definition of Service Robots | |
| SM004 | Centers for Medicare & Medicaid Services | Calendar Year 2026 Hospital Outpatient Prospective Payment System (OPPS) and Ambulatory Surgical Center Final Rule | The final rule will affect approximately 4,000 hospitals and 6,000 ASCs. |
| SM005 | Lifeward (formerly ReWalk Robotics) | Medicare Reimbursement Established for Medically Eligible Beneficiaries | Including individuals with VA benefits, Workers Compensation, and Medicare, nearly 40% of individuals with spinal cord injury are insured by payers that have reimbursed claims. |
| SM006 | VPM (Virginia Public Media) | Medicare begins covering cost of robotic exoskeletons | |
| SM007 | New Mobility | Medicare to Cover Personal Exoskeletons | CMS has proposed a payment level of $94,617 for devices in the brace benefit category. The beneficiary is responsible for Medicare's 20% copay. |
| SM008 | Pan American Health Organization (PAHO) / WHO | Estimates of the need for rehabilitation in the Americas | In 2019, across the Americas, 366 million people had health conditions that would benefit from rehabilitation, a 58% increase from 231 million in 1990. |
| SM009 | World Stroke Organization / PMC (PubMed Central) | WSO Global Stroke Fact Sheet 2025 (GBD 2021) | The estimated global cost of stroke is over US$890 billion (0.66% of the global GDP) per year, and is projected to almost double by 2050. |
| SM010 | PubMed / BMC Neurology | RCT on lower-limb exoskeleton rehabilitation after stroke (Zhang et al. 2026) | |
| SM011 | Journal of Medical Internet Research (JMIR) | Robot-Assisted Therapy for Upper Limb Rehabilitation After Stroke: Umbrella Review | Robot-assisted therapy improves upper-limb motor function after stroke, but ADL gains are generally comparable to conventional therapy. |
| SM012 | Roland Berger | Humanoid robots 2026: The convergence moment | Today there are only very few productive use cases of humanoid robots. |
| SM013 | TrendForce | TrendForce: China Humanoid Robot Output to Surge 94% in 2026 | Annual output growth up to 94% in 2026; Unitree and AgiBot projected to account for nearly 80% of total shipments. |
| SM014 | Robotics & Automation News | The State of Humanoid Robotics: From Research Labs to Real-World Potential | Manipulation, power and endurance, and general-purpose autonomy remain major bottlenecks; most deployments still pilots or co-pilot. |
| SM015 | Future Market Insights (FMI) | Humanoid Robot Market Size, Market Forecast and Outlook | The humanoid robot market was valued at USD 7.80 billion in 2025, projected to reach USD 10.69 billion in 2026, at a 37.0% CAGR to USD 248.90 billion by 2036. |
| SM016 | The Business Research Company (TBRC) | Rehabilitation Robotics Global Market Report 2026 | Rehabilitation robotics market size will grow from $1.81 billion in 2025 to $2.22 billion in 2026 at a CAGR of 22.7%. |
| SM017 | Mordor Intelligence | Rehabilitation Robots Market — Size, Share & Industry Analysis 2026–2031 | Rehabilitation robots market size in 2026 is estimated at USD 1.77 billion; exoskeletons led with 47.35% share. |
| SM018 | IMARC Group | Rehabilitation Robots Market Size, Share, Report 2026–2034 | The global rehabilitation robots market size reached USD 1,819.5 Million in 2025, growing at a CAGR of 16.53% during 2026–2034. |
| SM019 | Fortune Business Insights | Rehabilitation Robots Market Size, Share & Growth — Global Report 2034 | |
| SM020 | Fortune Business Insights | Service Robotics Market Size, Share, Growth — Global Report 2034 | The global service robotics market size was valued at USD 26.35 billion in 2025 and is projected to grow from USD 31.11 billion in 2026. |
| SM021 | BusinessWire / ResearchAndMarkets | Global Humanoid Robots Market Report 2026–2040 — Pricing Drops Below $10,000 | Goldman Sachs reported that manufacturing costs declined 40% year-over-year versus earlier projections of 15–20% annually; current costs ranging from $30,000–$150,000. |
| SM022 | Oliver Wyman | Boosting Healthcare Productivity with AI and Robotics | Without decisive action, global healthcare spending could almost double, from $11.8 trillion today to $23.1 trillion by 2040. |
| SM023 | Knowledge Sourcing Intelligence | Next-Gen Elderly Care: How AI Is Driving Industry Transformation | The AI-Powered Solutions For Elderly Care Market is predicted to reach USD 7.38 billion by 2031 at a high 31.60% CAGR. |
| SM024 | The Business Research Company (TBRC) | Humanoid Robot Global Market Report 2026 | Humanoid robot market size will grow from $5.44 billion in 2025 to $8.32 billion in 2026 at a CAGR of 52.9%. |
| SM025 | Knowledge Sourcing Intelligence | Elderly Care Robotics Market — Strategic Insights and Forecasts 2026–2031 | The elderly care robotics market is set to reach USD 0.64 billion in 2031, growing at a CAGR of 10.4% from USD 0.39 billion in 2026. |
| SP001 | Fourier | The Most Advanced Rehabilitation Robots | Today, Fourier Rehab serves more than 2,000 institutions across over 40 countries. |
| SP002 | Fourier | Fourier GR-1 | The First Mass-Produced Humanoid Robot. |
| SP003 | Fourier | Fourier GR-3 | Fourier GR-3 — Your caring and lovable friend. |
| SP004 | Wandercraft | Wandercraft Home | Atalante X — Exosquelette auto-équilibré pour l'entraînement à la marche en rééducation. |
| SP005 | EU-Startups | French Wandercraft raises €64.3 million to take robotics from hospitals to homes and factory floors | Wandercraft has established itself as an innovator in robotic mobility systems, especially through its clinical exoskeleton Atalante X, now deployed in over 100 hospitals and rehabilitation centres globally. |
| SP006 | Ekso Bionics | EksoNR | As the first exoskeleton FDA-cleared for stroke, acquired brain injury, MS and SCI and backed by more than 200 published articles... |
| SP007 | Ekso Bionics | Ekso Indego Therapy | Ekso Indego Therapy is an adjustable, lower-limb powered exoskeleton that can be custom-sized and perfectly fitted to patients in less than five minutes. |
| SP008 | CYBERDYNE | The world's first Wearable Cyborg HAL | HAL detects these so-called bio-electrical signals to perform the desired movements with the wearer's voluntary commands. |
| SP009 | CYBERDYNE | Investor Relations | This dual class structure reflects the concentration of voting rights with Dr. Yoshiyuki Sankai and the foundations. |
| SP010 | Lifeward | Lifeward - Redefine Possible | A growing portfolio of innovative products... ReWalk Personal Exoskeleton, AlterG Anti-Gravity Systems, ReStore Exo-Suit, MYOLYN FES Cycling. |
| SP011 | Lifeward | Medicare Reimbursement Established for Medically Eligible Beneficiaries | Nearly 40% of individuals with spinal cord injury are insured by payers that have reimbursed claims for medically eligible individuals. |
| SP012 | Lifeward | Investors | Lifeward Ltd. | ReWalk Robotics Ltd. is now doing business as Lifeward (Nasdaq: LFWD). |
| SP013 | Exoskeleton Report | Medicare Sets Reimbursement Rates for Personal Exoskeletons: A Milestone for Accessibility | Home-use powered exoskeletons... will be reimbursed $109,238 to $81,929 at an established rate of $91,032. |
| SP014 | UBTECH | About UBTECH | Our self-developed Walker is the first commercialized biped life-sized humanoid robot in China. |
| SP015 | PR Newswire | UBTECH Humanoid Robot Walker S2 Begins Mass Production and Delivery, with Orders Exceeding 800 Million Yuan | UBTECH has begun mass production and delivery of the first batch of several hundred full-size industrial humanoid robots, Walker S2. |
| SP016 | HKEX News | Listed Company Information Title Search — UBTECH ROBOTICS (09880) | Stock Code: 09880 — UBTECH ROBOTICS. |
| SP017 | Unitree Robotics | Unitree G1 | Price from $13.5K. |
| SP018 | TrendForce | China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share | Unitree Robotics and AgiBot are emerging as clear leaders... projected to account for nearly 80% of total shipments. |
| SP019 | AGIBOT | AGIBOT Home | Products — AGIBOT A2, AGIBOT X, AGIBOT Genie, AGIBOT D1, AGIBOT C5. |
| SP020 | PR Newswire | AGIBOT Makes Its U.S. Market Debut at CES 2026 with Its Full Humanoid Robot Portfolio | AGIBOT... 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. |
| SP021 | 1X | Introducing NEO Gamma | NEO Gamma is the next generation of home humanoids designed and engineered by 1X Technologies. |
| SP022 | 1X | Investor Relations | 1X | |
| SP023 | Figure | Figure | Figure 03 is a general purpose humanoid robot for every day. |
| SP024 | Figure | News | Figure | Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation. |
| SP025 | Figure | F.02 Contributed to the Production of 30,000 Cars at BMW | Contributed to the production of 30,000+ X3 vehicles. |
| SP026 | BMW Group PressClub | BMW Group to deploy humanoid robots in production in Germany for the first time | The results demonstrated that Physical AI can deliver measurable added value under real-world conditions. |
| SP027 | Mordor Intelligence | Rehabilitation Robots Companies - Top Company List | Rehabilitation Robots Company List ... Cyberdyne Inc., Ekso Bionics Holdings Inc., ReWalk Robotics Ltd., Fourier Intelligence, Wandercraft. |
| SP028 | Business Wire / ResearchAndMarkets | Global Humanoid Robots Market Report 2026-2040: Humanoid Robot Pricing Drops Below $10,000 as Market Expansion Accelerates | Unitree shocked the market in July 2025 by launching its R1 humanoid at just $5,900. |
| SP029 | Robotics & Automation News | The state of humanoid robotics: From research labs to real-world potential | These layers make for platforms that are impressively capable in demos and narrow tasks, but far from general-purpose autonomy. |
| SP030 | Robotics & Automation News | Why China’s new humanoid robot standards could change the industry | The emergence of humanoid robots introduces a different set of challenges... ensuring predictable and safe behavior becomes increasingly important. |
| SP031 | International Federation of Robotics | New IFR position paper on humanoid robots published | The latest IFR position paper takes a closer look at humanoid robots and tries to separate vision and reality. |
| SI001 | Fourier | GRx Humanoid Robot Series | The GRx humanoid robot series is a bipedal robot product lineup developed by Fourier. |
| SI002 | Fourier | Download | Fourier GR-3 Series Brochure. |
| SI003 | Fourier Rehab | The_Most_Advanced_Rehabilitation_Robots | RehabHub integrates over 30 types of robots to deliver comprehensive rehabilitation solutions. |
| SI004 | Fourier | GR-3 | Contact Sales. |
| SI005 | TMTPost | Fourier Launches First Humanoid Care Robot, Advancing Emotional AI in Robotics | In 2023, overseas revenue from its rehab division surged more than 50%, accounting for about 10% of total revenue. |
| SI006 | The Rakyat Post | Robotic Healthcare Company, Fourier Intelligence, Secures RM264 Million In Funding | The new funding will accelerate Fourier’s go-to-market expansion, enable innovation across its advanced healthcare robotics solutions and accelerate global growth opportunities. |
| SI007 | 36Kr | "Fourier" has completed a new round of financing, with the total amount reaching nearly 800 million yuan. | Fourier Intelligence will continue to promote the development of humanoid robots. |
| SI008 | Fineline Cube | Fourier Intelligence Secures USD 109 Million in Series E Funding for Robotics Innovation | The proceeds will be used to facilitate Fourier’s continuous innovation in the robotics field. |
| SI009 | Sango Automation | Nearly 800 Million! General Robotics Company Fourier Officially Announced The E Series Of Financing | In the future, Fourier will continue to promote the development of humanoid robots ... to promote the application of robot technology in more commercial scenarios. |
| SI010 | CNINFO / Runyang Technology | 关于对外投资暨增资上海傅利叶智能科技有限公司的进展公告 | 润阳科技按照投前人民币 800,000.0000 万元的估值,以人民币 30,000.0000 万元认购目标公司新增注册资本。 |
| SI011 | China Fund News | 润阳科技:拟3亿元增资傅利叶智能 中国基金报 | 公司拟以自有资金3亿元认购上海傅利叶智能科技新增注册资本141,972元。 |
| SI012 | Sina Finance | 润阳科技拟增资傅利叶智能 金额不超3亿元 | 本次增资的金额不超过3亿元,傅利叶的投前整体估值为80亿元。 |
| SI013 | Europa Satellite | Fourier GR-3 Humanoid: Specs, Uses, Pricing | Europa Satellite | Structured product metadata on the page lists priceCurrency EUR and price 106344.61. |
| SI014 | 120Shop | 傅利叶智能M2D02SP上肢康复机器人价格,傅利叶智能M2D02SP上肢康复机器人中标价,傅利叶智能M2D02SP上肢康复机器人市场价 | 全国平均价 128.00万;最高价 142.00万;最低价 97.00万。 |
| SI015 | Unitree Robotics | Humanoid robot G1_Humanoid Robot Functions_Humanoid Robot Price | Unitree Robotics | Price from $13.5K. |
| SI016 | The Robot Report | Despite the hype, Interact Analysis expects humanoid adoption to remain slow | High prices and dexterity gaps are likely to persist into the next decade. |
| SI017 | Bain & Company | Humanoid Robots: From Demos to Deployment | Most deployments remain early-stage, with heavy reliance on human supervision. |
| SI018 | Ekso Bionics / ChronoScale IR | Ekso Bionics Reports Fourth Quarter and Full-Year 2024 Financial Results | For the full year ended December 31, 2024, the Company used $9.8 million of net cash in operations. |
| SI019 | Ekso Bionics / ChronoScale IR | 10-K - 02/23/2026 - Ekso Bionics Holdings, Inc. | The majority of our sales are generated from our Enterprise Health products. |
| SI020 | Lifeward | Investors | Lifeward Ltd. | Our portfolio of proven solutions features innovative products including the ReWalk Exoskeleton, the ReStore Exo-Suit, the MyoCycle FES bikes, and ... AlterG Anti-Gravity systems. |
| SI021 | Nasdaq / Lifeward | Lifeward Reports Fourth Quarter and Full Year 2025 Financial Results | Revenue for the year ended December 31, 2025 was $22.0 million. |
| SI022 | CYBERDYNE | INVESTOR RELATIONS - CYBERDYNE | Financial Result for the Fiscal Year Ended March 31, 2026. |
| SI023 | CYBERDYNE | Financial Result for the Fiscal Year Ended March 31, 2026 | March 31, 2026 3,846 (12.3) (601) ... Cash and cash equivalents at end of year 8,991. |
| SI024 | Lifeward | Medicare Reimbursement Established for Medically Eligible Beneficiaries | Nearly 40% of individuals with spinal cord injury are insured by payers that have reimbursed claims for medically eligible individuals. |
| SI025 | VPM News | Medicare begins covering cost of robotic exoskeletons | Medicare covers about 80% of the price tag, which is just over $91,000. |
| SI026 | Sheltering Arms Institute | Medicare Covers Personal Robotic Exoskeletons: First person on the East Coast to apply for Medicare personal device coverage completing training | FDA-approved robotic exoskeletons ... cost approximately $100,000 each. |
| SI027 | Interact Analysis | Despite hype and large addressable market, humanoid robot adoption will remain low | Key barriers include regulatory and safety concerns, dexterity limitations, cost, and questions over whether humanoid is the optimal form-factor. |
| SE001 | Fourier | FOURIER-Robotics | Trusted by world-leading research institutes. |
| SE002 | Fourier Rehab | The_Most_Advanced_Rehabilitation_Robots | Today, Fourier Rehab serves more than 2, 000 institutions across over 40 countries, delivering one-stop solutions that empower therapists and patients alike. |
| SE003 | PR Newswire / Fourier | Fourier Makes CES Debut With GR-3, a Next-Generation Care-Focused Humanoid Robot | Standing about 165 cm tall with 55 degrees of freedom, the robot combines compact engineering with a soft, friendly exterior that feels approachable in homes, public spaces, commercial environments, and assisted settings. |
| SE004 | Gasgoo | CES 2026: Fourier Intelligence unveils full-size humanoid robot GR-3 | GR-3 also carries a full-perception interaction system that integrates visual recognition, sound-source localization, and haptic feedback to enable natural human-robot interaction. |
| SE005 | Fourier Documentation Center | About GR-2 | Fourier 文档中心 | Fourier's GR-2 General-Purpose Humanoid Robot stands 1.75 meters tall, weighs 63 kilograms, and has up to 53 Degrees of Freedom covering its entire body. |
| SE006 | Fourier Documentation Center | About Fourier N1 | Fourier 文档中心 | The Fourier Nexus-01 General-purpose Humanoid Robot stands 1.3 meters tall and weighs 38 kg, combining flexibility and dynamic stability. |
| SE007 | Fourier Documentation Center | Teleoperation Resources Overview | Fourier 文档中心 | We provide a complete pipeline for upper-limb imitation learning based on the Fourier Robotics Teleoperation System, including data collection, a training framework, and an open-source dataset. |
| SE008 | Fourier Documentation Center | About Dexterous Hand | Fourier 文档中心 | The Fourier Bionic Five-Finger Dexterous Hand features a humanoid five-finger design with built-in 6 linear actuators and control circuits. |
| SE009 | GitHub / FFTAI | Fourier | We actively contribute to the open-source community and share projects that can help advance the field of robotics. |
| SE010 | GitHub / FFTAI | GitHub - FFTAI/fourier-grx-GR2: Documents for fourierGR2 | 198 Commits. No releases published. |
| SE011 | GitHub / FFTAI | GitHub - FFTAI/fourier-grx-N1: Documents for fourierN1 | This repository mainly maintains GitHub Pages / Jekyll documentation for Fourier-GRX-N1 SDK installation, run modes, interfaces, task instructions, and FAQs; it is not the SDK source-code repository. |
| SE012 | GitHub / FFTAI | GitHub - FFTAI/teleoperation: A.K.A. Fourier Advanced Robot Teleoperation System (F.A.R.T.S.) | This system enables real-time humanoid robot control using a VR headset and hand tracking. |
| SE013 | GitHub / FFTAI | GitHub - FFTAI/Wiki-GRx-Pipeline: The development pipeline of Fourier Intelligence GRx Series Robot. | The complete process for training Fourier GRx robots for flat-ground walking can be divided into model preparation, Isaac Gym training, MuJoCo verification, and deployment to the real robot. |
| SE014 | GitHub / FFTAI | GitHub - FFTAI/Wiki-GRx-Gym | This repository provides an NVIDIA Isaac Gym training environment ... to train the Fourier N1 robot on walking over complex terrain. |
| SE015 | GitHub / FFTAI | GitHub - FFTAI/Wiki-GRx-Mujoco | This repository provides a MuJoCo-based implementation for RL policy verification and visualization for the Fourier N1 robot. |
| SE016 | GitHub / FFTAI | GitHub - FFTAI/Wiki-GRx-Models | This repository provides the Unified Robot Description Format (URDF) files for the GRx robot, enabling enthusiasts and developers to explore, modify, and extend the capabilities of this robotic platform. |
| SE017 | GitHub / FFTAI | GitHub - FFTAI/Wiki-GRx-MJCF: This repository provide a tool to convert robot urdf file to mjcf. | Parse a URDF model into MJCF format. |
| SE018 | GitHub / FFTAI | GitHub - FFTAI/fourier_dexhand_sdk: The DexHand SDK provides a set of interfaces for communication with the underlying dexterous hand device. | The SDK supports both Python3 and C++ interfaces, and currently can only be used in Linux systems. |
| SE019 | GitHub / FFTAI | GitHub - FFTAI/fourier_actuator_sdk: This is the SDK repository for Fourier Smart Actuators(FSA) on Fourier Humanoid Robots. | This is the SDK repository for Fourier Smart Actuators(FSA) on Fourier Humanoid Robots. |
| SE020 | GitHub / FFTAI | GitHub - FFTAI/fourier-grx-client: Client library for the Fourier GRx series robot | This software is highly unstable and subject to change at any time. |
| SE021 | NVIDIA | Spotlight: Fourier Trains Humanoid Robots for Real-World Roles Using NVIDIA Isaac Gym | NVIDIA Technical Blog | When transferred directly to GR-2’s physical controls, the model’s action tensors achieved an 89% success rate. |
| SE022 | Thor Assistive Technologies / Fourier Intelligence | RehabHub Brochure A5 size 2023 compressed | One therapist can now monitor multiple patients at the same time as all the devices are linked, replacing the traditional one-to-one training. |
| SE023 | Ascot Rehabilitation Therapy | RehabHub by Fourier Intelligence - Ascot Rehabilitation Therapy | RehabHub by Fourier Intelligence brings robotics, sensor-based feedback and interactive therapy into the rehabilitation journey. |
| SE024 | STEPS Rehabilitation | STEPS RehabHub - STEPS Rehabilitation | STEPS is home to the only ArmMotus™ EMU in the UK. |
| SE025 | Hobbs Rehabilitation | Hobbs Rehabilitation Partners With Thor Assistive Technologies And Becomes A Fourier Intelligence RehabHub™ | Hobbs Rehabilitation | Hobbs Rehabilitation ... has partnered with Thor Assistive Technologies to become a Fourier Intelligence RehabHub™. |
| SE026 | 36Kr / The Frontline | Fourier Aims to Introduce Humanoid Robots into Health and Rehabilitation Care Scenario on Its 10th Anniversary | The Frontline | As of now, Fourier Intelligence has released two self-developed general humanoid robots in the GRx series, with a cumulative shipment of hundreds of units. |
| SE027 | Fourier | Fourier_General Purpose_Humanoid Robot_Assistant | The First Mass-Produced Humanoid Robot. |
| SU001 | Fourier Rehab | The Most Advanced Rehabilitation Robots | Today, Fourier Rehab serves more than 2,000 institutions across over 40 countries. |
| SU002 | NHG Health | NHG Health and Fourier Rehab Advance Rehabilitation Robotics and Healthcare Innovation Further | Since then, both organisations have embarked on clinical rehabilitation initiatives, including the deployment and evaluation of selected Fourier rehabilitation technologies for rehabilitation services at Tan Tock Seng Hospital. |
| SU003 | Exoskeleton Report | Fourier Rehab’s GReAT Summit 2026 Highlights Rehabilitation Robotics, ExoMotus M4 Updates, and New Global Partnerships | Delegates also visited Xinchang Medical Center and Yong Ci Hospital, where Fourier technologies were presented in real-world clinical settings. |
| SU004 | Brooks Rehabilitation | Fourier Rehab and Brooks Rehabilitation Sign Strategic MOU to Advance Research in Robotic Neurorehabilitation | Fourier Rehab will deploy its flagship upper- and lower-limb robotic rehabilitation systems at Brooks Rehabilitation as part of the collaboration. |
| SU005 | Fourier Rehab | Fourier Rehab and Brooks Rehabilitation Sign Strategic MOU to Advance Research in Robotic Neurorehabilitation | With ArmMotus EMU and ExoMotus M4 now in place at Brooks, the partnership will pave the way for robust joint studies. |
| SU006 | STEPS Rehabilitation | STEPS RehabHub | STEPS is home to the only ArmMotus EMU in the UK. |
| SU007 | Ascot Rehabilitation Therapy | RehabHub by Fourier Intelligence | London’s only Fourier Centre of Excellence. |
| SU008 | Hobbs Rehabilitation | Hobbs Rehabilitation Partners With Thor Assistive Technologies And Becomes A Fourier Intelligence RehabHub | Hobbs Rehabilitation at Laverstock Care Centre, is now home to three new neurotech devices, provided by Thor Assistive Technologies. |
| SU009 | Independent Neurorehabilitation Providers Alliance | The EMU has landed | STEPS Rehabilitation is home to one of the only two Exomotus machines in the UK, and to the only ArmMotus EMU. |
| SU010 | Thor Assistive Technologies / Fourier Intelligence | RehabHub Brochure | One therapist can now monitor multiple patients at the same time as all the devices are linked, replacing the traditional one-to-one training. |
| SU011 | Neuro Rehab Times | RehabHub – next-gen neuro-rehab | Globally, there are more than 200 RehabHubs in operation, including the Shirley Ryan AbilityLab in Chicago, STEPS Rehabilitation in the UK and the Yongchi Rehabilitation Hospital in China. |
| SU012 | PRLog / Fourier Intelligence | National Healthcare Group and Fourier Intelligence ink MOU to Advance MedTech Innovations | Both parties will co-develop home-based rehabilitation technologies and devices that will help improve patients’ accessibility to care and therapy outside the hospital, as well as reduce the manpower required in the community settings. |
| SU013 | Gasgoo Auto News | Fourier Intelligence Partners with Singapore's National Healthcare Group and Nagoya University | The two will conduct muscle synergy analysis using Fourier’s ArmMotus EMU 3D upper-limb rehabilitation robot. |
| SU014 | Fourier Rehab | Fourier Rehab Leads the Future of Neurorehabilitation at IFNRCON 2025 | Held from the 27th to the 29th of March 2025 at The Christian Medical College and Hospital (CMC Ludhiana) campus. |
| SU015 | Indian Federation of Neurorehabilitation | IFNRCON 2025 | Global implementation of evidence based practice in neuro rehabilitation. |
| SU016 | Ren Ci Hospital | Effects of robotic assisted gait rehabilitation: A pilot study | A single arm pre-post interventional study design was conducted at Ren Ci Community Hospital and Day Rehab Centre. 47 patients were recruited. |
| SU017 | IndiaMedToday | Saifee Hospital Introduces ExoMotus M4 Robotic Gait Trainer to Advance Neuro and Orthopedic Rehabilitation | The ExoMotus M4 is well-suited for high-volume tertiary care hospitals like Saifee Hospital. |
| SU018 | Express Healthcare | Saifee Hospital adds ExoMotus M4 exoskeleton for gait rehabilitation | Robotic gait trainers allow early mobilisation, repeated task-specific training, monitoring and consistent therapy delivery. |
| SU019 | Healthcare Executive | Saifee Hospital Introduces ExoMotus M4 Robotic Gait Trainer to Advance Neuro and Orthopaedic Rehabilitation | The ExoMotus M4 enables us to provide safe, personalized, and data-driven gait training while minimizing physical effort for therapists. |
| SU020 | Robocare Solutions | Physical Therapy and Rehabilitation Medicine Products | Physical Therapy and Rehabilitation Medicine ArmMotus M2 Pro ... Fourier Intelligence ExoMotus M4. |
| SU021 | Healthlink Holdings | ArmMotus EMU Neurorehabilitation | It can precisely measure every movement trajectory, strength, cognitive ability, ROM, motor control ability and generate a report after the training. |
| SU022 | Fourier Rehab | Fourier Rehab and Kurage Announce Strategic Partnership | A key milestone in this partnership will be the global launch of the bundled solution, starting with its planned presentation at Medica 2025 in Germany. |
| SU023 | PR Newswire | HealthTech Partners Global Establishes U.S. Headquarters in Charlotte, NC; Announces Exclusive Strategic Partnership With Fourier Rehab | Through this partnership, HTPG will bring Fourier Rehab’s advanced solutions—including the RehabHub Robotics Suite and Galileo Systems—to rehabilitation providers nationwide. |
| SU024 | Centers for Medicare & Medicaid Services | Complying with Outpatient Rehabilitation Therapy Documentation Requirements | Outpatient rehabilitation therapy includes physical therapy, occupational therapy, and speech-language pathology services, which we cover when the physician or NPP certifies the treatment plan. |
| SU025 | NIH SEED Innovator Support Team | Reimbursement Knowledge Guide for Medical Devices | Developing a strategy for obtaining coverage, coding, and payment should begin early, and well in advance of FDA clearance or approval. |
| SU026 | robotics.press | Fourier Intelligence | Its humanoid pivot remains unproven with no verified paid industrial deployments, undisclosed shipping volumes, and a declining CB Insights Mosaic Score. |
| SU027 | robotics.press | Fourier Intelligence: Competitive Response | Our DRES event tracking shows zero publicly confirmed paid deployments for either the GR-1 or GR-3 platforms. |
| SR001 | Fourier | FOURIER-Robotics | |
| SR002 | Fourier Rehab | The_Most_Advanced_Rehabilitation_Robots | Today, Fourier Rehab serves more than 2,000 institutions across over 40 countries. |
| SR003 | Fourier Rehab | Fourier Rehab and Brooks Rehabilitation Sign Strategic MOU to Advance Research in Robotic Neurorehabilitation | |
| SR004 | Brooks Rehabilitation | Fourier Rehab and Brooks Rehabilitation Sign Strategic MOU to Advance Research in Robotic Neurorehabilitation | With ArmMotus™ EMU and ExoMotus™ M4 now in place at Brooks, the partnership will pave the way for robust joint studies and the co-creation of scalable therapy protocols informed by clinical data. |
| SR005 | PR Newswire / Fourier | Fourier Makes CES Debut With GR-3, a Next-Generation Care-Focused Humanoid Robot | |
| SR006 | EUR-Lex | Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence | |
| SR007 | EUR-Lex | Regulation (EU) 2017/745 on medical devices | |
| SR008 | European Commission | New Regulations | |
| SR009 | Artificial Intelligence Board and Medical Device Coordination Group | Interplay between the Medical Devices Regulation (MDR) & IVDR and the Artificial Intelligence Act (AIA) | |
| SR010 | U.S. Food and Drug Administration | Cybersecurity | These same features also increase potential cybersecurity risks. Medical devices, like other computer systems, can be vulnerable to security breaches, potentially impacting the safety and effectiveness of the device. |
| SR011 | U.S. Food and Drug Administration | Device Approvals, Denials and Clearances | |
| SR012 | U.S. Food and Drug Administration | Medical Device Databases | |
| SR013 | Centers for Medicare & Medicaid Services | Therapy Services | |
| SR014 | Centers for Medicare & Medicaid Services | Medicare Physician Fee Schedule Final Rule Summary: CY 2026 | |
| SR015 | U.S. Food and Drug Administration | Recalls, Corrections and Removals (Devices) | |
| SR016 | European Commission | Market surveillance and vigilance | |
| SR017 | Bureau of Industry and Security | Department of Commerce Revises License Review Policy for Semiconductors Exported to China | BIS will now review export license applications for the Nvidia H200, AMD MI325X, and similar chips on a case-by-case basis provided certain security requirements are met. |
| SR018 | International Trade Administration | China - U.S. Export Controls | |
| SR019 | Xinhua / State Council Information Office | China releases national standard system for humanoid robotics and embodied AI | |
| SR020 | TrendForce | China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce | In China, vendors are rapidly clarifying commercial use cases and scaling up production, which is expected to drive annual output growth up to 94% in 2026. |
| SR021 | TechCrunch | Why China’s humanoid robot industry is winning the early market | |
| SR022 | CNBC | Elon Musk envisions humanoid robots everywhere. China may be the first to make it a reality | |
| SR023 | Exoskeleton Report | Fourier Rehab’s GReAT Summit 2026 Highlights Rehabilitation Robotics, ExoMotus M4 Updates, and New Global Partnerships | |
| SR024 | Interesting Engineering | Fourier's humanoid robot brings 'warm tech companionship' to CES 2026 | |
| SR025 | BotInfo.ai | Fourier Humanoid Robots: GR-2, GR-3 Specs & Pricing (2026) | |
| SR026 | Humanoid.guide | GR-2 | |
| SR027 | robotics.press | Fourier Intelligence: Company Profile | |
| SR028 | robotics.press | Fourier Intelligence: Competitive Response | Our DRES event tracking shows zero publicly confirmed paid deployments for either the GR-1 or GR-3 platforms — no named customers, no disclosed KPIs, no measurable ROI. |
| SR029 | Hill Dickinson | Humanoid robots and the law - preparing for a new era of risk | A clear contractual allocation of responsibility between developers, integrators and end users will become increasingly important. |
| SR030 | DQS | AI Act & AI-Enabled Medical Devices: Regulatory Status 2026 | MDR certification remains the sole regulatory pathway for placing AI-enabled medical devices on the EU market. |
| SV001 | CNINFO / Runyang Technology | 关于对外投资暨增资上海傅利叶智能科技有限公司的进展公告 | 协议约定公司按照投前人民币80亿元的估值,以人民币3亿元认购傅利叶新增注册资本。 |
| SV002 | Grandall Law Firm | Grandall Shanghai Assisted State-owned Companies Completing Equity Investment in Frontier | The total amount of Series E financing has reached nearly RMB 800 million. |
| SV003 | Fourier Intelligence | 傅利叶宣布已完成新⼀轮融资,E系列融资总额近8亿 | 傅利叶宣布已完成新一轮融资,E系列融资金额合计达到近8亿元。 |
| SV004 | Fineline Cube | Fourier Intelligence Secures USD 109 Million in Series E Funding for Robotics Innovation | Fourier Intelligence ... has reportedly raised close to RMB 800 million (USD 109 million) in a Series E financing round. |
| SV005 | 36Kr | "Fourier" has completed a new round of financing, with the total amount reaching nearly 800 million yuan. | The total amount of the Series E financing has reached nearly 800 million yuan. |
| SV006 | TechCrunch | Famed roboticist says humanoid robot bubble is doomed to burst | today’s billions are funding expensive training experiments that will never scale to mass production. |
| SV007 | CNBC | China ships more humanoid robots than the U.S. as investors diverge on AI bets | U.S. humanoid startups are being priced as wide-reaching artificial intelligence platforms, while Chinese ones are seen more as industrial hardware plays. |
| SV008 | Apptronik | Apptronik Closes Over $935 Million Series A with New $520 Million Extension Round | bringing Apptronik’s total Series A to more than $935 million and total capital raised to nearly $1 billion. |
| SV009 | Forbes | Apptronik Scores $935 Million, Hits Top 3 For Humanoid Robotics Funding | |
| SV010 | Sacra | Apptronik valuation, funding & news | Valuation $5.50B ... On February 11, 2026, Apptronik closed a $520M Series A extension. |
| SV011 | Tracxn | Figure | Figure has raised $1.75B in funding ... with a current valuation of $39B. |
| SV012 | Ekso Bionics | Annual report on Form 10-K for the fiscal year ended December 31, 2025 | Revenue $12,799 ... Gross profit % 53 % ... Net loss $ (11,695 ). |
| SV013 | Yahoo Finance | Ekso Bionics Holdings Inc (EKSO) Valuation Measures & Financial Statistics | Market Cap 42.01M ... Price/Sales 2.91 ... Enterprise Value/Revenue 3.93. |
| SV014 | CompaniesMarketCap | Ekso Bionics (EKSO) - Revenue | Revenue in 2025 (TTM): $14.74 Million USD. |
| SV015 | Stock Analysis | Lifeward (LFWD) Financials & Income Statement | FY 2025 Revenue 22.03 ... Gross Margin 38.25%. |
| SV016 | Stock Analysis | Lifeward (LFWD) Market Cap & Net Worth | Lifeward has a market cap or net worth of $18.03 million as of May 20, 2026. |
| SV017 | Yahoo Finance | Lifeward Ltd. (LFWD) Valuation Measures & Financial Statistics | Market Cap 17.71M ... Price/Sales 0.34 ... Enterprise Value/Revenue 0.90. |
| SV018 | CompaniesMarketCap | Lifeward (LFWD) - Revenue | Revenue in 2025 (TTM): $24.49 Million USD. |
| SV019 | CYBERDYNE | Consolidated Financial Results (Tanshin) for the Fiscal Year Ended March 31, 2026 | Fiscal year ended March 31, 2026 Revenue 3,846 ... Gross profit 2,265. |
| SV020 | Yahoo Finance | CYBERDYNE Inc. (7779.T) Stock Price, News, Quote & History | Market Cap 54.89B ... Price/Sales (ttm) 14.17 ... Enterprise Value/Revenue 10.54. |
| SV021 | CYBERDYNE | INVESTOR RELATIONS - CYBERDYNE | |
| SV022 | Stock Analysis | Ubtech Robotics Corp (HKG:9880) Market Cap & Net Worth | Ubtech Robotics Corp has a market cap or net worth of 56.54 billion as of May 20, 2026. |
| SV023 | Yahoo Finance | Ubtech Robotics Corp Ltd (9880.HK) Stock Price, News, Quote & History | Market Cap 55.42B ... Price/Sales (ttm) 21.73 ... Enterprise Value/Revenue 22.12. |
| SV024 | CompaniesMarketCap | UBTECH Robotics (9880.HK) - Revenue | In 2024 the company made a revenue of $0.17 Billion USD. |
| SV025 | UBTECH Robotics | Financial Reports | UBTECH Robotics | |
| SV026 | UBTECH Robotics | Investor Relations | UBTECH Robotics | |
| SV027 | Lifeward | Investors | Lifeward Ltd. | |
| SV028 | Ekso Bionics | Ekso Bionics SEC filings | |
| SV029 | HKEX News | Listed Company Information Title Search — UBTECH ROBOTICS (09880) | |
| SV030 | CompaniesMarketCap | Ekso Bionics (EKSO) - Market capitalization | As of May 2026 Ekso Bionics has a market cap of $42.04 Million USD. |
| SV031 | CompaniesMarketCap | Lifeward (LFWD) - Market capitalization | On May 20th, 2026 the market cap of Lifeward was reported to be: $18.03 Million USD. |