Bulage
上海具身 AI 实验室,创始人履历和资本都强,但 $2B 种子估值下公开证明很薄
Bulage 有机会长成中国具身 AI 的关键平台,但公开证据还不足以支撑今天按据称 $2B 估值买入。
封面要素
公司概况
Bulage 是一家上海具身智能初创公司,由 Alibaba Qwen / 通义千问模型项目此前的技术负责人和对外代表 Lin Junyang 于 2026 年 5 月创立。公开报道称,公司瞄准世界模型,以及面向机器人或其他物理终端的具身大脑;但官网几乎没有可读的产品、治理、客户或财务细节。到 2026 年 6 月中旬,中英文报道均称 Bulage 已完成约 $220M 种子轮或首轮融资,估值约 $2 billion。尽管公开运营证明很有限,它已成为中国定价最激进的新一批人形机器人 / 机器人 AI 实验室之一。
- 成立时间
- 2026-05-27
- 创始人
- Lin Junyang
- 创立地点
- Shanghai, China
- 总部
- Shanghai, China
- 产品
- 公开层面,Bulage 主要被定义为世界模型 / 具身大脑公司,而不是一家已披露机器人 SKU、软件模块或部署平台的公司。公司网站未发现可读的官方产品文档、路线图、开发者入口、安全栈或客户案例。
- 客户
- 未公开披露;现有叙事暗示未来用例更可能是企业机器人、机器人 OEM 或工业部署,而不是消费软件。
- 商业模式
- 未公开披露。公开证据尚未说明 Bulage 计划靠机器人硬件、模型 / 软件授权、部署服务,还是打包式全栈方案变现。
- 阶段
- Seed-stage private (formed May 2026; first round reported closed June 2026)
- 融资情况
- 据报道,2026 年 6 月完成约 $220M 种子轮 / 首轮融资,估值约 $2B,由 Gaorong 和 HSG / HongShan 领投,中文报道中还点名 Tencent。
执行摘要
主要优势
- 创始人与市场匹配度很强:Lin Junyang 直接来自 Alibaba Qwen 体系,在中国新一代具身 AI 实验室里技术履历最清晰。
- 据报道约 $220M 的种子 / 首轮融资,让 Bulage 有资源做产品、招人、推进早期试点,不必马上承受再融资压力。
- 品类时点顺风:2026 年,具身 AI、人形机器人和工业机器人智能在中国及全球私募市场都进入高关注窗口。
- 如果 Bulage 真能做成 OEM 中立的具身大脑层,而不是又一家机器人整机 OEM,上行空间可能有战略级别。
主要风险
- 公开验证仍极薄:未找到具名客户、收入、产品架构或已披露部署指标。
- 据报道 $2B 估值相比 Figure、Agility、Apptronik、UBTECH 等证据更充分的可比公司,已经跑在 Bulage 披露证据前面。
- 公司仍很年轻,治理和披露不透明:董事会、联合创始人梯队、投资人权利、股权结构表、烧钱速度、现金跑道都缺少公开视图。
- 这个赛道仍重资本、也暴露于地缘风险,出口管制、算力、安全和商业化风险都还没关闭。
未决问题
- 具名客户、试点或部署证据,以及任务级运营结果。
- 清晰产品边界:模型层、机器人栈、服务业务,还是混合模式。
- 烧钱速度、现金跑道、招聘计划,以及后续融资是可选还是不可避免。
- 治理材料,包括董事会构成、投资人权利和高管梯队。
- 真实世界具身 AI 部署的安全、隐私和合规控制。
目录
01公司概览
1.1 身份、命名与公开足迹
2026 年的公开报道一致把 Bulage 视为一家很新的上海主体,而不是已有长期运营史的隐身项目。本次审阅的中文资料将其登记式公司名识别为 Shanghai Bulage Technology Co., Ltd.,写作上海卜拉格科技有限公司;任务简报则给出了另一个写法:布拉格。由于 2026-07-22 可访问的公开资料压倒性使用卜拉格,本章将 Bulage 作为罗马化标签,并把字面命名差异列为尽调点,而不是直接抹平。公司被描述为成立于 2026-05-27,因此在报告日仍不足一岁。即便按刚成立的前沿 AI 实验室标准,它的公开沟通界面也异常单薄:bulage.cn 可解析但没有返回可读的公司内容,bulage.com 则托管无关游戏内容,两个域名都没有给投资人提供正常的官方产品、领导层或客户参考页面。这一缺口很重要,因为公司在发布可用官方网页叙事前,已经被定价到独角兽量级。[CO001, CO002, CO003, CO004, CO005, CO006]
| 指标 | 值 | 日期 | 置信度 | 缺口 |
|---|---|---|---|---|
| 来源使用的公开公司名称 | Shanghai Bulage Technology Co., Ltd. / 上海卜拉格科技有限公司(公开名称) | 2026-06 | 高 | |
| 中文用字差异 | 已审阅公开来源使用「卜拉格」;任务简报提供「布拉格」 | 2026-07 | 中 | 需要官方公司发布确认首选中文写法 |
| 成立时间 | 2026-05-27 | 2026-05-27 | 中 | |
| 运营基地 | 中国上海 | 2026-06 | 高 | 已抓取来源未确认确切办公地址 |
| 阶段 | 种子阶段的私营前沿 AI 实验室 | 2026-06 | 中 | 各来源对轮次叫法不同:种子轮、天使轮或首轮 |
| 首轮融资完成 | 已披露出资方金额合计推算为 $220M;更广泛报道称数亿美元 | 2026-06 | 中 | 已具名出资合计为 $220M,但并非所有参与方都公开可识别 |
| 投后估值 | $2B / 约 RMB 13.5B | 2026-06 | 高 | |
| 收入 | 未公开收入或 ARR | |||
| 客户 | 已审阅来源未披露公开客户 logo、试点或已签约部署 | |||
| 员工数 | 未公开披露员工数量 | |||
| 官方 .cn 域名 | 可以解析,但 2026-07-22 未返回可读公司内容 | 2026-07-22 | 高 | 需要管理层确认网站是否仍待上线 |
| bulage.com 状态 | 与公司无关的网络游戏内容 | 2026-07-22 | 高 | 公司似乎未控制 .com 域名 |
表格结合了直接来源事实;未披露公开封面指标处明确标为空值。
[CO001, CO002, CO003, CO004, CO005, CO006]创始人履历、主体设立、技术命题、公开足迹缺口与资本如何交织,构成 Bulage 当前公司快照。
[CO004, CO005, CO008, CO017, CO020, CO024]1.2 创始人背景与治理不透明
在概览阶段,Bulage 的可投性几乎全靠创始人支撑。已审阅资料一致认为,Lin Junyang 是新实验室的创始人和对外代表;英文报道也称其为 Justin Lin。他此前担任 Alibaba Qwen / 通义千问模型项目的技术负责人或核心负责人。对如此年轻的公司来说,创始人画像异常强:公开传记和第三方报道将其出生年份写为 1993 年,提到他毕业于 University of International Relations 和 Peking University’s School of Foreign Languages,并梳理了他从 2019 年加入 DAMO Academy 到 2022 年底领导 Qwen 的 Alibaba 路径。多篇报道还称,他 2025 年在 Qwen 内部搭建机器人与具身智能团队,因此 Bulage 更像是已孵化研究方向的延续,而不是冷启动。主要反向因素是治理不透明。已审阅材料没有识别出联合创始人名单、已披露董事会构成、独立董事,也没有 Lin 之外的高管梯队。对一家成立数周内以 $2 billion 估值融资的公司而言,治理披露缺口是实质尽调问题,不是表面瑕疵。[CO008, CO009, CO010, CO011, CO012, CO013]
| 人物或角色 | 状态 | 背景 | 创始人或职能匹配度 | 关键人物依赖 |
|---|---|---|---|---|
| Lin Junyang / Justin Lin | 创始人,也是唯一持续被公开点名的领导者 | 1993 年出生,曾任 Alibaba DAMO 和 Qwen 技术负责人;毕业于 University of International Relations 和 Peking University School of Foreign Languages;2025 年在 Qwen 内部组建机器人与具身智能团队。 | 创始人-市场匹配度强:集合了前沿模型领导经验、多模态背景,以及创业前明确的具身 AI 准备。 | 极高;已审阅公开公司身份几乎都绑定在 Lin 个人身上。 |
| 联合创始人名单 | 未公开披露 | 已抓取来源未识别出 Lin 之外的联合创始人、CTO、总裁或运营联合负责人。 | 未知,直到尽调拿到创始团队名单。 | 高;缺少具名互补型运营者,会抬高执行集中风险。 |
| 董事会 / 独立治理 | 未公开披露 | 已审阅来源未发布融资后公司的董事、观察员席位或治理文件。 | 未知;公开证据还无法评估治理质量。 | 高;投资者应核实董事会控制权、保留事项和继任计划。 |
列举有意保持不完整,因为已抓取来源集合中只有一位具名个人被公开披露。
[CO008, CO009, CO010, CO011, CO012, CO013]1.3 融资、估值与利益相关方地图
资本故事是 Bulage 登上 2026 年前沿实验室地图的核心事实。Crunchbase News 明确称,公司是一家总部位于上海的机器人智能初创公司,完成由 Gaorong Capital 和 HSG 领投的 $220 million 种子轮,估值 $2 billion。中文报道对轮次标签说法不一——首轮、天使轮,或泛称数亿美元融资——但反复指向同一组已披露出资:Gaorong 和 HongShan / Sequoia China 各 $100 million,Tencent $20 million。这些具名出资合计 $220 million,与 Crunchbase 的数字一致;MarketScreener 称公司获得 HongShan、Gaorong、Tencent 和“其他投资者”融资,则留下了仍有未披露参与方的可能。公开报道还称,Lin 在 2026 年 5 月至 6 月控制一组新注册主体,包括 100% 持股的 Yuyong、99% 持股的上海 Bulage Technology,以及一家通过 Bulage 关联的管理合伙企业。这一结构显示,围绕融资快速搭建了实体,但公开资料仍未披露股权比例、投资人治理权、清算条款或董事席位分配。[CO017, CO018, CO019, CO020, CO021, CO022]
| 利益相关方 | 轮次角色 | 公开披露的重要性 | 已知金额或状态 | 尽调问题 |
|---|---|---|---|---|
| Lin Junyang | 创始人 / 控制人 | 公开报道把所有实体设立和战略叙事都系在 Lin 身上。 | 2026 年 6 月媒体摘要称其控制 Yuyong(100%)和 Bulage(99%)。 | 确认融资后直接和间接持股、归属安排,以及任何超级投票权。 |
| Gaorong Capital | 领投方 | Crunchbase 和中文报道均将其列为两家领投机构之一。 | 多篇报道公开归因金额为 $100M。 | 确认 Gaorong 是否拥有董事会或观察员权利,以及是否仍有分期资本附带条件。 |
| HongShan / Sequoia China / HSG | 领投方 | 同一投资者家族在不同来源中以新旧品牌出现;被识别为共同领投方。 | 多篇报道公开归因金额为 $100M。 | 确认法律投资实体、治理权利,以及英文报道中的「HSG」是否对应同一个领投载体。 |
| Tencent | 战略参与方 | 反复被描述为 $20M 参与方,而不是领投方。 | 多篇报道公开归因金额为 $20M。 | 确认 Tencent 是否获得商业合作、数据或云服务优先权。 |
| 其他投资者(未具名) | 可能的其他参与方 | MarketScreener 表述称公司获得已具名投资者和「其他投资者」融资。 | 未公开识别。 | 获取完整的融资后股本表和全部附函条款。 |
公开图谱不完整,因为没有备案文件或官方融资公告披露完整财团或持股比例。
[CO017, CO020, CO021, CO022, CO023, CO024]公开可见的概览指标主要是资本和公司年龄;商业 KPI 仍未披露。
[CO006, CO020, CO021, CO023, CO024, CO036]1.4 战略论点、里程碑与眼前尽调旗标
资料集中对 Bulage 的战略叙事一致:公司面向世界模型,以及服务机器人或其他物理终端的具身大脑。资料描述的方向,是把 AI 从屏幕内推理推向感知、物理世界建模和行动规划。这个论点解释了为什么公司即便没有公开产品目录、客户 logo 或已披露收入,仍能与前沿实验室放在同一语境中讨论。它也解释了为什么投资人看重创始人质量和赛道时点,超过当下商业证据。里程碑链条被压得很短:Lin 在 2026 年 3 月初离开 Qwen;5 月媒体报道他正为新实验室融资,目标估值约 $2 billion;运营主体被报道在 5 月至 6 月注册;到 6 月中旬,多家媒体称首轮已关闭,管理层正在试探下一轮融资。这种速度有两面性。它制造了即时独角兽身份,但也意味着今天的公开尽调仍更多依赖团队履历和市场叙事,而不是已验证的 PMF。[CO027, CO028, CO029, CO030, CO031, CO032]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2019 | Lin Junyang 加入 Alibaba DAMO Academy | 治理 | Lin Junyang;Alibaba DAMO Academy 相关方 | 开始积累后来支撑创始人可信度的技术履历。 | |
| 2022-12 | 据报道,Lin 接手 / 牵头 Qwen 技术工作 | 治理 | Lin Junyang;Alibaba Qwen/Tongyi 相关方 | 将 Lin 推到全国可见的前沿模型角色。 | |
| 2025-10 | 据报道,Lin 在 Qwen 内部组建机器人与具身智能团队 | 产品 | Lin Junyang;Qwen team 相关团队 | 显示 Bulage 论点在创业启动前已经孵化。 | |
| 2026-03-03/04 | Lin 离开 Alibaba Qwen,并公开发布辞职信息 | 不利 | Lin Junyang;Alibaba Qwen 相关方 | 形成创业前的创始人断点。 | |
| 2026-05 | 媒体报道称 Lin 正在为一家新 AI 实验室融资,估值约 $2B | 融资 | 目标估值约 $2B | Lin Junyang;潜在投资者包括 Sequoia/HongShan 和 Gaorong | 轮前估值叙事早于公司广泛公开出现。 |
| 2026-05-27 | 据报道,Shanghai Bulage Technology 成立 | 创立 | Lin Junyang 控制的实体 | 正式运营实体只在融资确认前数周出现。 | |
| 2026-06-15 至 2026-06-17 | 多家媒体报道首轮融资已完成 | 融资 | 投后估值 $2B;已具名出资合计 $220M | Gaorong、HongShan/Sequoia China/HSG、Tencent,可能还有其他投资者 | Bulage 几乎在成立后立刻成为独角兽。 |
| 2026-06 | 已有报道显示融资后开始接触新一轮融资 | 融资 | 探索下一轮 | Lin Junyang 创业团队 | 暗示融资节奏异常压缩,计划烧钱速度可能较高。 |
时间线以创始人为中心,因为官方公司渠道尚未披露公开产品发布或客户部署时间表。
[CO011, CO012, CO013, CO014, CO019, CO020]从创始人离开 Alibaba,到 Bulage 首次被报道融资和随即跟进的下一轮募资,时间线被压缩得很紧。
[CO011, CO012, CO013, CO014, CO019, CO020]1.5 图表
02市场分析
2.1 市场边界、纳入支出与替代品
分析 Bulage,应把它放在具身智能与人形机器人栈里,而不是拿它对标所有机器人或所有生成式 AI。公开报道反复把公司与世界模型、具身大脑、以及物理世界中的模型驱动行动绑定。因此,相关支出比单一机器人 SKU 更宽,但比中国整个自动化市场更窄。应纳入的支出包括具身基础模型研发、机器人训练数据与仿真基础设施、用于训练或部署的通用人形和轮式具身机器人,以及 OEM 或企业付费在工厂、物流、零售或服务场景测试具身系统的试点项目。应排除的支出包括传统固定式工业机器人、仅限仓储的 AMR、手术机器人、自动驾驶,以及不依赖 Bulage 所称具身 AI 论点的扫地机器人等消费设备。现状替代品仍是人力、传统任务专用自动化,以及大型在位企业的内部 AI 或机器人团队。这个边界很重要,因为 Bulage 尚未披露产品目录;投资人实际上是在承销具身 AI 栈中的一层,而不是一个已验证的终端市场细分。[CM001, CM002, CM003, CM004, CM005, CM006]
| 细分领域或支出类别 | 纳入支出 | 剔除支出 | 买方或付款方 | 与 Bulage 的相关性 |
|---|---|---|---|---|
| 具身模型层 | 世界模型、具身大脑、控制栈、训练系统、语料库、仿真 | 不带物理动作闭环的通用纯文本 LLM 应用 | 机器人 OEM、实验室、企业试点项目 | 最贴近 Bulage 论点 |
| 通用人形机器人 | 面向试点或部署销售的人形机器人硬件和系统集成 | 固定式工业机械臂和非具身自动化 | OEM、工厂、公共部门试点 | Bulage 式软件的重要下游载体 |
| 具身数据和训练基础设施 | 真实世界数据采集、数字孪生、试点测试、算力租赁、语料服务 | 未绑定具身系统的通用云支出 | 训练基地、OEM、市级平台 | 模型改进的关键使能层 |
| 工业部署项目 | 工厂巡检、装配、物流、物料搬运、仓库试验 | 没有运营工作流的纯展示机器人 | 制造运营和自动化预算 | 产品跑通后,可能成为最早规模化收入池 |
| 服务和公共交互项目 | 接待、零售、巡逻、教育、娱乐、养老和康复试点 | 消费电子小工具和简单家电 | 商场、场馆、SOE、医疗机构、地方政府 | 真实存在,但常由信号展示驱动 |
| 传统机器人和自动化 | 核心范围不纳入;仅作替代基准 | 传统工业机器人、AMR、手术机器人、自动驾驶 | 现有工厂资本开支决策方 | 替代和对比集合,不是主要市场 |
表格把可触达市场窄化到具身智能系统和使能层,而不是全部机器人。
[CM001, CM002, CM003, CM004, CM005]从最窄的中国人形机器人销售口径,到更宽的中国具身智能集群和相邻工业自动化基础,呈现四个嵌套视角。各层不可相加。
[CM006, CM010, CM011, CM012, CM013, CM014]2.2 用多重视角做市场规模,而非只看一个 TAM
公开资料支持几组真实但互不兼容的规模口径。CNBC 报道的 Morgan Stanley 2026 年 6 月报告称,中国人形机器人市场 2026 年约 $2 billion,当年外部销售出货 50,000 台,到 2030 年升至 $15 billion 和 446,000 台。TrendForce 另行预测,中国人形机器人产量将在 2026 年增长 94%,并称 2026 年下半年是行业商业化拐点。China Economic Net 引用了更宽口径估算:Development Research Center 预测中国具身智能产业到 2030 年达 RMB 400 billion,2035 年超过 RMB 1 trillion;另有 IDC 估算,中国用户在具身智能机器人上的支出 2025 年已超过 $1.4 billion,到 2030 年可能达 $77 billion。来自 Omdia 的出货报道又给出一个视角:2025 年全球通用具身机器人出货约 13,318 台,2035 年预测为 2.6 million 台。这些数字不应揉成一个头条 TAM,因为它们测量的是不同范围:外部人形机器人销售、用户支出、更宽的具身智能产业产值,以及全球出货。正确结论不是市场精确等于某一个数,而是中国已经是该品类的量级中心;但像 Bulage 这样的模型层初创公司,其经济上真正相关的 SAM 仍无法从免费公开资料中单独剥离。[CM008, CM009, CM010, CM011, CM012, CM013]
| 发布方 | 年份或日期 | 地区 | 数值或单位 | 方法或范围 | 置信度 | 局限 |
|---|---|---|---|---|---|---|
| Morgan Stanley 经 CNBC | 2026-06 | 中国 | 2026 年 50,000 台出货、$2B 市场;到 2030 年 446,000 台出货、$15B | 基于供应链实地研究的人形机器人对外销售预测 | 中 | 仅限人形机器人,排除原型机 / 内部使用 |
| TrendForce | 2026-04 | 中国 | 2026 年人形机器人产量 +94% YoY | 中国产量增长预测,附商业化评论 | 中 | 是产量增长率,不是市场收入 |
| IDC 经 China Economic Net | 2025-2026 | 中国 | 2025 年用户支出 $1.4B+;到 2030 年 $77B | 具身智能机器人用户支出预测 | 中 | 范围宽于人形机器人对外销售,方法未完全公开 |
| Development Research Center 经 China Economic Net | 2025-2026 | 中国 | 到 2030 年 RMB 400B;到 2035 年 >RMB 1T | 更广义的具身智能产业规模估算 | 中 | 宽于 Bulage 可能的变现层 |
| Omdia 经 Telecoms / Agibot | 2025-2026 | 全球 | 2025 年出货 13,318 台;到 2035 年年出货 2.6M 台 | 通用具身智能机器人出货预测 | 中 | 全球范围且以出货台数计,不是中国收入 |
| IFR | 2024-2025 | 中国 | 工业机器人保有量 2.027M;2024 年年安装量 295,000 台 | 工业机器人基数和采用度代理指标 | 高 | 工业机器人是相邻市场,不是 Bulage 直接收入 |
| 上海政府计划 | 2025-2028 | 上海 | 到 2027 年核心产业产出目标 50B yuan | 市级产业集群建设目标 | 高 | 政策目标,不是已测算的市场结果 |
不同视角衡量不同层:对外销售、用户支出、更广义产业产出、出货量,以及相邻装机基础背景。
[CM006, CM008, CM010, CM011, CM012, CM013]以十亿美元计的低、中、高三档已发布 2030 年中国市场价值口径。差异来自范围不同,不是统计置信区间。
低 / 中 / 高是对已发布 2030 年口径的编辑整理,单位为十亿美元。Morgan Stanley 衡量人形机器人外部销售,发展研究中心估算的是更宽的具身智能产业规模,IDC 口径是用户在具身智能机器人上的支出。
[CM011, CM012, CM013, CM014, CM017, CM042]2.3 今天谁买、谁用、谁真正付钱
具身 AI 的买方分层仍在过渡期。Interact Analysis 把应用分成八个桶,从学术 R&D、机器人训练,到娱乐、制造、仓储、公共服务和家庭用途。反向证据显示,最早的真实买方往往是研究实验室;较新的客户画像虽然出现,但仍偏弱,包括把机器人放进大堂或展陈场景、主要追求示范价值而非替代劳动力的国企。即便如此,已审阅官方资料也显示支出下一步想往哪里走:上海政策框架优先支持物流、工业制造、商业零售、医疗康复和家政服务;杭州国家级试点基地已在 30 多个职业场景运行 130 多台机器人;WAIC 公开报道指向工厂、门店和家庭中的快速试验。实践中,付款方通常是 OEM、工厂运营方、国家支持的试点项目、地方政府平台或大型企业预算负责人,而不是终端消费者。运营买方可能是制造 VP、自动化负责人、实验室负责人或公共部门创新办公室。用户随场景变化:工厂里的操作员和流程工程师、训练场景的数据采集技术员,以及零售和接待场景的服务人员或访客。对 Bulage 来说,最可信的近期客户因此不是家庭消费者,而是寻求具身模型、训练系统或早期部署能力的机器人厂商或标杆企业。[CM018, CM019, CM020, CM021, CM022, CM023]
| 细分市场 | 买方 | 用户 | 付款方 | 工作流或用例 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 学术 R&D 和机器人训练 | 实验室负责人或训练基地运营方 | 研究人员、数据采集团队、机器人操作员 | 大学、市级平台或国资支持基地 | 采集数据、训练策略、基准测试模型 | 科研经费或平台运营方 | 数据和模型迭代需求 |
| 制造业 | 工厂自动化负责人或运营 VP | 工艺工程师、产线工人、维护团队 | 工厂业主、OEM 或 SOE 试点资助方 | 巡检、装配、搬运、物料移动 | 工业 capex / 智能工厂预算 | 劳动力短缺和质量控制 |
| 仓储和物流 | 仓库自动化负责人或物流运营方 | 仓库员工和调度团队 | 3PL、物流企业或地方试点基金 | 分拣、搬运、最后一公里或内部物流 | 自动化或数字化转型预算 | 吞吐提升和劳动力替代 |
| 公共服务 / 零售 / 接待 | 场地运营方或公共部门创新办公室 | 员工、访客、安保或接待团队 | 商场运营方、国企、地方政府或赞助方 | 导览、巡逻、接待、展示、客户互动 | 运营或创新预算 | 品牌展示叠加早期工作流自动化 |
| 医疗 / 康复 / 养老 | 医院管理层或照护机构运营方 | 护士、护理员、患者、住民 | 医院、照护服务商或政府项目 | 辅助照护、转运、康复支持 | 机构采购预算 | 老龄化和人手压力 |
| 家庭服务 | OEM 的消费产品或生态团队;成熟零售买家尚未出现 | 家庭用户 | 未来家庭买家或打包服务平台 | 清洁、助餐、安全监测、陪伴 | 消费预算或打包订阅 | 安全性、成本和可靠性还要继续改善 |
表中各行反映了从实验室和演示走向企业及公共部门部署的迁移;家庭场景仍偏未来。
[CM017, CM018, CM019, CM020, CM021, CM024]映射 2025-2026 年证据中最可见的市场分段里,谁付款、谁决策、谁使用系统。
[CM017, CM018, CM019, CM020, CM021, CM024]呈现从政策支持试点到真正规模化部署的观察路径,并标出当前商业化卡住的位置。
[CM009, CM020, CM028, CM032, CM035, CM038]2.4 增长驱动、采用约束与仍未证明的事
从生态层面看,增长逻辑很强。中国已经拥有全球最大的工厂机器人装机基数、上升中的国产供应商份额、强力国家支持,以及越来越正式的标准、测试基础设施和补贴体系。官方与研究资料把具身 AI 需求同劳动力短缺、劳动力成本上升、制造业升级,以及把 AI 和机器人价值链留在中国的战略意愿相连。但市场仍足够早,商业化风险压过估值风险。MERICS 发现,当前人形机器人仍缺乏精度和灵巧性,大多停留在小规模、场地定制化试验,且对大规模部署来说仍太贵;商业可行阈值约 CNY 160,000,而报道中的当前均价约 CNY 300,000 至 CNY 500,000。多份反向资料还指出,PMF 仍不清晰,一些工厂部署更像 demo 而不是成交,最难的瓶颈仍是真实世界数据和跨环境泛化,而非单纯硬件可得性。北京方面也已警示泡沫风险,因为已有 150 多家公司挤入该领域。因此,Bulage 所在市场战略动能极强,但在客户 ROI、复购和模型层变现上公开证明仍弱。关键尽调问题不是具身 AI 在中国是否有战略重要性,而是 Bulage 能否在多份资料暗示的 2027–2028 洗牌前,把国家级推动转化成商业上可防守的位置。[CM028, CM029, CM030, CM031, CM032, CM033]
| 驱动因素或约束 | 方向 | 时点 | 影响 | 尽调问题 |
|---|---|---|---|---|
| 中国工业自动化基础与本土供应商进步 | 正向 | 当前 | 相邻领域的大量装机基础降低采用摩擦,也利于本地采购 | 其中有多少能现实升级为具身系统? |
| 市级与国家级政策支持 | 正向 | 当前至中期 | 补贴、测试平台、标准和试点项目加速试用 | 需求有多依赖补贴? |
| 用工短缺与制造升级压力 | 正向 | 中期 | 支撑自动化 ROI 叙事,也提高部署紧迫性 | 当前付费意愿最强在哪里? |
| 具身数据与训练基础设施建设 | 正向 | 当前 | 训练基地和语料平台改善模型迭代闭环 | Bulage 是否有专属数据入口? |
| 机器人成本高、灵巧性弱 | 负向 | 当前 | 许多工作流仍无法按经济账大规模部署 | Bulage 能否把任务成功率拉高到足以支撑 ROI? |
| 数据稀缺、跨环境泛化弱 | 负向 | 当前 | 模型离开已训练工厂或演示场地后可能失效 | 有什么证据证明可跨客户迁移? |
| 泡沫风险与赛道拥挤 | 负向 | 当前至中期 | 可能压缩融资窗口,并在 2027-2028 年前迫使整合 | 如果融资回归常态,Bulage 护城河有多耐久? |
市场享受战略顺风,但商业化仍不成熟;因此分析驱动因素时,必须同时把采用约束讲清楚。
[CM028, CM029, CM030, CM033, CM034, CM035]2.5 图表
03竞争格局
3.1 格局类型与 Bulage 的起点
Bulage 不是进入一场只同另一家初创公司正面竞争的整齐赛道。相关的 2026 年购买格局至少横跨三类玩家加上现状替代品。第一类是全栈具身 AI 实验室,试图同时掌握机器人和控制栈;公开代表包括 Figure、1X、AGIBOT、Unitree、UBTech、Agility 和 Boston Dynamics。第二类是工业在位企业或准在位企业,它们带来服务、安全、制造和部署经验,这些不是新实验室一夜之间能买来的。第三类是中国规模优先的竞争者,它们未必已拥有最强公开模型叙事,但已经有价格锚、供应链杠杆和出货动能。现状替代品仍是任务专用自动化、人力和内部系统集成。 Bulage 起跑时有一个不可否认的优势,也有一个同样实质的短板。优势是资本加创始人履历:据报道的 $220 million 种子轮和 $2 billion 估值,给了它快速招聘、训练模型和购买硬件的资源。短板是披露与证明。与本章几乎所有具名同行相比,Bulage 在官方界面上没有公开产品页、没有发布包装、没有客户故事,也没有经验证的部署记录。这意味着公司目前更多是在用论点竞争——世界模型和具身大脑——而不是用可观察的商业执行竞争。市场正从 demo 走向狭窄但真实的工厂、物流和开发者部署,这个缺口很重要。[CP001, CP023, CP025, CP030, CP034, CP035]
| 竞争对手 | 类别 | 规模 / 融资信号 | 目标细分市场 | 差异化 | 局限 |
|---|---|---|---|---|---|
| Bulage | 具身智能新进入者 | 种子轮 $220M,估值 $2B,但没有公开产品或客户证明 | 未知;可能面向 OEM、企业或一体化机器人部署 | 资金、创始人履历、世界模型 / 具身大脑论点 | 没有公开产品页、定价、部署或分销证明 |
| Figure | 模型优先的全栈人形机器人实验室 | Series B 轮 $675M,估值 $2.6B;与 BMW 签署商业协议 | 短期面向制造业,长期面向家庭 | Helix VLA、强合作伙伴品牌组合、高端前沿模型叙事 | 没有公开标价;商业规模仍在早期 |
| 1X | 家庭优先的人形机器人实验室 | Series B 轮 $100M;12 个月内融资 >$125M | 家庭,以及物流和安保中的企业客户 | 柔软安全设计、远程操作辅助学习闭环、消费者定位 | 家庭安全和规模化商业证明仍有限 |
| Agility Robotics | 工业部署领先者 | 计划公开交易,总募资 >$620M;有具名企业部署 | 仓储、制造、配送 | Digit 加 Arc 工作流软件、服务和具名客户 | 定价不透明;仍聚焦狭窄工业任务 |
| Boston Dynamics | 工业老牌标杆 | Hyundai 支持的产品化 Atlas;2026 年交付设备已锁定 | 汽车与工业自动化 | 可靠性、服务经验和清晰的工业优先打法 | 公开价格透明度低,通用用途大范围落地证据有限 |
| Unitree | 中国规模优先的成本领先者 | 公开零售人形机器人定价;TrendForce 招股书摘要给出 IPO 和利润率信号 | 开发者、实验室、工业及近消费买家 | 低公开价格锚、垂直整合、广泛硬件线 | 应用层企业打包不如工业集成商清晰 |
| UBTech | 工业集成商 | Walker 大规模交付说法;报道中的订单和工厂伙伴 | 汽车和工厂运营方 | 工厂工作流契合、LLM 辅助规划、换电保障稼动率 | 公开定价和实际经济性仍未披露 |
| AGIBOT | 中国出货和数据集领先者 | Omdia 评为 2025 年出货第一,份额 39% | 制造、物流、酒店接待、培训、研究 | 大型具身数据集叙事、多元机器人组合、规模叙事 | 公开 ASP、毛利率和复购细节仍稀缺 |
画像聚焦 2026 年与 Bulage 最相关的买家路径:具身模型领导者、工业部署领导者和中国规模领导者。公开证据缺位处保留未知,不做补全。
[CP001, CP004, CP007, CP010, CP012, CP015]按商业部署成熟度和具身模型 / 数据雄心,对关键竞争对手做序数定位。
坐标轴是基于证据的序数判断,综合采信的产品、融资、部署及负面来源;不是单一第三方排名。
[CP002, CP005, CP010, CP012, CP015, CP019]3.2 西方参照玩家:模型、安全与服务
西方参照组分成两种不同的竞争肌肉。Figure 和 1X 主打具身智能与家庭泛化。Figure 的公开叙事是,搭载 Helix 的 Figure 03 能在不可预测的家庭环境中移动,拿起数千种新的家用物体,并最终把语言条件行为扩展到狭窄脚本之外。1X 讲的是一条相近但更柔和的故事:NEO 面向家庭,设计上轻且安静,但对于它不知道的任务,仍依赖专家引导路径。这些公司对 Bulage 重要,因为它们是最清晰的公开基准,支撑“模型质量和数据闭环可以成为人形机器人主要护城河”这一主张。 Agility 和 Boston Dynamics 则锚定光谱另一端。它们的优势不是宏大的 AGI 叙事,而是部署纪律。Agility 把 Digit、工作流软件和现场支持组合起来,并在测试或部署主张中公开点名 Amazon、GXO、Schaeffler、Toyota Motor Manufacturing Canada 和 Mercado Libre。Boston Dynamics 更明确地说,家庭不是正确的首发市场,因为安全标准、成本和能力仍不成熟。这给 Bulage 设下了不舒服的基准:即便模型能力正在成为头条故事,今天看起来最真实的买方,仍在奖励服务、安全、工作流适配和现场可靠性,而不只是前沿模型野心。[CP002, CP003, CP004, CP005, CP006, CP007]
| 公司 | 具身模型雄心 | 工业部署证明 | 家庭导向 | 服务 / 工作流层 | 价格透明度 |
|---|---|---|---|---|---|
| Bulage | 宣称聚焦世界模型 / 具身大脑 | 未披露 | 未披露 | 未披露 | 未披露 |
| Figure | 高:Helix VLA 支持语言条件控制 | 中:BMW 阶段性部署协议 | 高:Figure 03 定位家庭 | 低-中:公开工作流栈不如 Agility 细 | 低 |
| 1X | 中-高:视觉操作加专家引导学习 | 低-中:有企业引用,但规模化部署细节稀疏 | 高:NEO 明确是家庭机器人 | 中:专家模式和早期访问支持路径 | 中:订金公开,完整 ASP 未公开 |
| Agility Robotics | 中:支持 AI,但信息表达不那么围绕前沿模型 | 高:具名部署和客户故事 | 低 | 高:Arc 加服务和支持 | 低 |
| Boston Dynamics | 中:AI 伙伴强,但工业框架占主导 | 高:Hyundai 证明、2026 年交付设备已锁定、既有机器人收入基础 | 低 | 高:强调可服务性和现场维护 | 低 |
| Unitree | 中:硬件和运动能力比应用模型栈更可见 | 中:规模信号强,但这里缺少更多具名企业工作流细节 | 低-中 | 低-中 | 高 |
| UBTech | 中:用于工厂任务的多模态和 LLM 规划 | 高:工厂集成和交付信息 | 低 | 中-高:制造系统集成 | 低 |
| AGIBOT | 高:具身数据集和多形态产品策略 | 中-高:出货和部署广度说法 | 低-中 | 中 | 低 |
取值是有证据支撑的定性判断。没有支撑的单元格保留未披露或低透明度,不猜测。
[CP002, CP003, CP005, CP006, CP008, CP010]矩阵显示,行业分化不看功能数量多寡,而看验证场、当前收入路径、运营负担和可能的护城河候选。
[CP009, CP011, CP014, CP023, CP033, CP036]3.3 中国规模领导者:价格、吞吐与组件杠杆
如果西方玩家定义了公开前沿叙事,中国赛场则定义了近期商业化压力。Unitree 是最清晰的公开价格锚,因为它明示 G1 起价 $13,500,同时推进能力更强的全尺寸 H1 系列,并声称对电机、减速器、控制器、LiDAR 和算法有强内部控制。TrendForce 补上了关键信号:据称 Unitree 招股书显示,2025 年人形机器人收入超过四足机器人,综合毛利率 60%,并规划 75,000 台人形机器人产能。AGIBOT 给出另一套同样重要的模板:多样化机器人形态、大型具身数据集计划,以及 Omdia 支持的 2025 年 5,168 台人形机器人出货、全球份额 39% 的说法。 UBTech 位于这两种模式之间。它更显眼地聚焦工业集成、大模型辅助决策、工厂系统连接和换电式续航。更广泛地看,TrendForce、CNBC、SCIO、MERICS 和反向投资人报道都指向同一个方向:中国如今的严肃竞争者,远多于围绕单一独角兽叙事所能解释的数量。国内制造商超过 140 家,型号超过 330 个,政府也明确推动接口标准化并加速工业部署。因此,Bulage 面对的不只是几个知名 logo,而是一个正在压缩实验室原型、供应链规模化与商业可读产品包装之间距离的本土市场。[CP015, CP016, CP017, CP018, CP019, CP020]
| 公司 | 公开价格 / 合同信号 | 打包 / 合同模式 | 公开包含内容 | 未披露内容 | 影响 |
|---|---|---|---|---|---|
| Bulage | 未披露 | Unknown | 只有高层融资和战略报道 | 产品形态、ASP、许可模式、服务义务、客户 | Bulage 还不能用真实打包经济性做基准比较 |
| Figure | 无公开 ASP | 以里程碑为基础的 BMW 商业协议;Figure 03 定位家庭 | Helix 能力叙事和具名企业关系 | 实际价格、支持模式、部署节奏 | 叙事强,但公开变现细节有限 |
| 1X | NEO 早期访问订金 $200 | 消费者早期访问加专家引导任务支持 | 家庭用例、安全叙事、订金信号 | 完整设备价格、支持成本、转化率、企业定价 | 组内最面向消费者的公开打包,但经济性仍不透明 |
| Agility Robotics | 无公开标价 | 商业部署和 RaaS 式工业机器人模式 | Digit、Arc 和支持栈 | 实际机器人价格、服务利润率、多年期单元经济 | 工业买家买的可能是结果,而不是标价 |
| Boston Dynamics | 无公开标价 | 企业部署和服务牵引的工业打包 | Atlas 可维护性和维护叙事 | 定价、融资结构、合同范围 | 老牌厂商可信度高,但价格发现仍在私下完成 |
| Unitree | G1 起价 $13,500 | 直接硬件销售 / 开发者式采购路径 | 标价和核心规格 | 企业支持、完整解决方案服务、实际折扣 | 强低端锚,压低市场预期 |
| UBTech | 无公开标价 | 定制工业部署 | 换电、载荷、工厂集成功能 | ASP、维护负担、合同期限 | 竞争点是工业工作流适配,而不是公开标价 |
| AGIBOT | 无公开标价 | 组合销售和部署模式公开披露不完整 | 出货、份额、数据集和场景广度叙事 | ASP、附加软件收入、复购结构 | 规模叙事清晰,变现架构不清晰 |
企业人形机器人供应商的公开价格发现异常稀薄。Unitree 和 1X 是主要公开锚;多数公司仍通过定制协议或不透明企业打包销售。
[CP006, CP008, CP012, CP015, CP019, CP021]3.4 护城河耐久性与被替代风险
公开证据还不足以支持“人形机器人护城河已经在任何地方锁定”的强主张,更不用说 Bulage。家庭优先玩家仍需证明安全、低接触支持和可复用效用。工业优先玩家仍面对任务范围窄、定制集成重的负担。中国规模优先玩家受益于更低硬件成本和政策支持,但仍要面对泛化、数据稀缺和进口高端组件瓶颈。因此,来自 KrASIA、Yicai 和 Unite.ai 的反向报道很重要,因为它们认为估值膨胀跑在已验证 PMF 前面。换句话说,品类是真的,但护城河故事往往讲得太早。 对 Bulage 来说,竞争含义很直接。当 Unitree 能公布价格锚、AGIBOT 能引用出货领先、UBTech 能引用订单和工厂部署、Agility 或 Boston Dynamics 能指向服务能力和具名企业工作流时,融资本身不是护城河。可持续的 Bulage 护城河需要在四个地方之一显现:在高价值任务上有独特有用的具身模型表现;通过 OEM 中立软件层更快部署;经济性强于全栈对手;或形成别人难以低成本复制的专有伙伴 / 数据飞轮。上述几点目前都没有公开证据。竞争结论因此不是 Bulage 不可能赢,而是举证责任几乎全部还在前方。[CP026, CP028, CP029, CP030, CP032, CP033]
| 护城河主张 | 威胁 / 竞争对手回应 | 严重程度 | 当前证据 | 缓释措施 / 尽调问题 |
|---|---|---|---|---|
| 具身模型优势 | Figure 和 1X 已经公开强 VLA 或家庭学习叙事 | 高 | 模型差异化在营销中可见,但尚未出现在经验证的 ROI 排名中 | 要求 Bulage 提供任务级基准、推理架构和伙伴试点 |
| 中国制造优势 | Unitree、AGIBOT 和 UBTech 已经处在更强的本地供应链和规模计划中 | 高 | TrendForce、SCIO 和 MERICS 都指向中国规模化和标准化动能 | 测试 Bulage 是模型层中立,还是也必须打硬件成本战 |
| 资本作为护城河 | 种子资金本身可以被已有部署或 IPO 路径的同行匹配甚至绕开 | 高 | Bulage 有资本,但同行拿出了更丰富的公开执行证据 | 在把融资视为可防守优势之前,要求首个试点证明 |
| 家庭优先的通用性 | Boston Dynamics 和反向来源认为,家庭场景在成本和安全上仍太早 | 中-高 | 家庭领域领先者仍缺少规模化变现证明 | 先看狭窄付费工作流,再看宽泛家庭叙事 |
| 工业集成 | Agility 和 UBTech 已将工作流集成和支持包装为核心产品 | 高 | 具名客户和工厂系统集成先于 Bulage 的公开引用出现 | 澄清 Bulage 卖的是软件、机器人还是联合方案 |
| 低价颠覆 | Unitree 的公开定价可能把买家预期向下重置 | 高 | G1 价格公开;多数竞争对手不透明 | 用任务 ROI 给 Bulage 做基准,而不是只对照高端叙事 |
| 标准和接口带来的锁定 | 政府主导的标准化可能随时间降低切换成本 | 中 | SCIO 突出互操作性、模块化和更低协调成本 | 测试 Bulage 是否拥有专有数据,还是只有可替换组件 |
| 品类热度支撑 | 泡沫风险可能收紧资本,也可能迅速惩罚 PMF 弱的公司 | 高 | KrASIA、Yicai 和 Unite.ai 都警告商业化落后于估值 | 按证明里程碑评估 Bulage,不按可比公司头条下结论 |
登记表检验 Bulage 的潜在护城河能否扛住真实竞争对手回应,而不是只列吸引人的属性。
[CP014, CP023, CP026, CP028, CP029, CP030]简要汇总最影响 Bulage 当前防御性的竞争特征。
数值是基于采信来源证据得出的定性判断,不是已披露基准分数。
[CP023, CP026, CP028, CP029, CP032, CP034]3.5 图表
04财务
4.1 公开收入模型:已知与未知
Bulage 的公开财务界面几乎是空白。官方域名仍没有可读的产品、定价或客户细节,而推动公司独角兽叙事的融资报道,重点放在创始人履历、估值和世界模型论点上,不在订单或收入上。这意味着第一步分析是负面的,而不是肯定的:公开资料没有基础断言 Bulage 是机器人销售商、模型授权方、企业软件厂商、集成承包商,或四者的某种组合。公司未来可能通过硬件、软件、部署服务、维护和数据工具的任意组合变现,但先后顺序没有披露。 同行证据很重要,因为它显示这个品类里财务上可读的收入层长什么样。Agility 强调的部署栈包括机器人硬件、工作流软件和支持。UBTECH 的报告把人形机器人产品与服务同其他机器人产品和硬件设备分开。Research and Markets 的 2026 年行业报告明确把软件、系统集成和维护列为更宽市场的一部分,而不只是机器人 ASP。换句话说,即便 Bulage 自身模式尚不清楚,公开证据也表明人形机器人变现通常是打包且运营很重。任何简单的 SaaS 式解读都缺乏支持。[CI001, CI002, CI003, CI018, CI021, CI034]
| 潜在收入流 | Bulage 公开证据 | 可比公开信号 | 当前状态 | 收入质量判断 | 尽调问题 |
|---|---|---|---|---|---|
| 一体化机器人销售 | 未披露 | Unitree、UBTECH 和 Agility 都围绕机器人部署变现 | 可能存在,但未验证 | 如果没有服务或复购,单纯硬件收入质量可能偏低 | 要求提供首个产品包、BOM 和目标 ASP |
| 模型 / 软件许可 | 未披露 | Agility 和 UBTECH 都在硬件之外强调软件 / 平台价值 | 可能存在,但未验证 | 如果 OEM 中立且毛利率高,可能有吸引力,但目前没有证据 | 要求提供软件架构、定价指标和附加率假设 |
| 集成 / 部署服务 | 未披露 | Agility Arc 和 UBTECH 工厂集成暗示上线初期服务占比高 | 可能是品类常态 | 服务收入能验证需求,但可能压缩利润率 | 要求提供试点 SOW、实施人员配置和验收标准 |
| 维护 / 支持 | 未披露 | Research and Markets 在更宽口径的市场规模测算中纳入维护 | 可能是品类常规项 | 关系到在线率和留存,但可能高度依赖人力 | 要求提供支持义务、SLA 设计和现场服务计划 |
| 数据 / 工具 / 开发者工作流 | 未披露 | 同业靠数据飞轮或面向开发者的硬件间接变现能力 | Unknown | 长期可能抬高利润率,但缺少证据 | 要求说明 Bulage 是销售工具、API,还是只交付集成结果 |
本表区分品类可以变现什么,以及 Bulage 实际披露了什么。目前后者基本仍是空白。
[CI001, CI002, CI003, CI010, CI021, CI034]定性映射 Bulage 这类人形机器人创业公司必须如何把研发转成可产生收入的部署层。
[CI002, CI003, CI019, CI021, CI024, CI034]4.2 定价、变现与销售周期代理
人形机器人公开定价记录很薄,但已有片段很有提示性。Unitree G1 起价 $13,500,提供了少见的低端公开锚。Research and Markets 的 2026 年中国报告随后给出更宽的工业价格阶梯:低端系统 RMB100,000–200,000,中端系统 RMB250,000–500,000;一旦加入软件、集成和维护,更宽市场继续扩大。1X 对 NEO 给出 $200 订金信号,但没有完整消费端或企业端 ASP。Agility、Figure 和 UBTECH 披露能力、客户、订单或融资,却仍把真实定价保密。 这种不透明对 Bulage 有两层含义。第一,可能的销售动作更像咨询式、按里程碑推进,而不是自助式。Agility 的文件和投资人材料谈的是多年订单、部署承诺和客户管线,不是无摩擦的经常性订阅。第二,即便公开价格锚不完整,心理上仍有影响。一家未披露包装的种子阶段公司,最终会被放进一个已有参照的赛场:Unitree 设定低标价基准,工业同行则用工作流集成、安全、正常运行时间和服务来证明更高总合同价值。Bulage 还没有显示自己会落在这条连续谱的哪里。[CI006, CI007, CI008, CI009, CI010, CI018]
| 公司 | 公开价格 / 合同信号 | 产品包装信号 | 已披露内容 | 未知项 | 含义 |
|---|---|---|---|---|---|
| Bulage | None | Unknown | 仅披露融资和战略 | 价格、收入模式、合同结构、客户 | 尚无法对标已兑现变现 |
| Figure | 无公开 ASP | 与 BMW 达成里程碑式商业协议 | 资金用途和企业关系 | 已兑现价格、毛利率、支持成本 | 叙事强度高于公开财务细节 |
| 1X | NEO 订金 $200 | 家庭早期体验加专家指导 | 订金和融资用途 | 完整 ASP、转化率、支持经济性 | 消费者端最可见的包装,但仍不是完整单位经济性 |
| Agility | > $300M 多年期 Digit v5 订单;预计总募资 >$620M | 机器人 + 软件 + 部署 | 订单、管线、资本计划 | 已兑现机器人价格和毛利率 | 同业中最接近公开工业收入验证 |
| Unitree | G1 起价 $13,500 | 开发者 / 硬件直销信号 | 公开标价和规格 | 服务和企业合同经济性 | 给出最清晰的低端价格锚 |
| UBTECH | 无公开标价 | 工业部署和产品 / 服务收入 | 分部收入、出货量、部署行业 | 单机器人 ASP、维护负担、客户集中度 | 最佳公开财务可比对象,但仍不完整 |
| AGIBOT | 无公开标价 | 出货和场景宽度叙事 | 份额和出货声明 | ASP、毛利率、服务结构 | 有规模叙事,缺少公开定价深度 |
人形机器人领域公开变现证据仍主要是订金、融资计划、订单管线,以及少数标价锚点,而不是完整的已兑现 ASP 披露。
[CI006, CI007, CI008, CI009, CI018, CI019]以 CNY 计的已发布价格和可行性口径,显示当前市场价格带仍可能高出商业 ROI 门槛多远。
低 / 中 / 高数值在有已发布区间时使用该区间。可行性门槛画成一条平线,因为引用来源给出的是单一门槛,而非区间。
[CI008, CI009, CI015, CI017]4.3 成本结构、单位经济性与可比公开结果
这个市场中最好的公开成本与利润窗口来自 UBTECH,因为它已经上市。其 2025 年年报说明,哪怕收入已经出现,人形机器人承销仍是资本密集型。收入增长 53.3% 至 RMB2.0 billion,毛利翻倍以上,人形机器人产品与服务成为最大分部,收入 RMB820.6 million。但净亏损仍有 RMB789.8 million,R&D 消耗 RMB507.5 million;公司看起来从容,只是因为多轮融资后现金升至接近 RMB4.9 billion。这是 Bulage 应吸收的可比教训:商业进展不会自动意味着轻资产或自我造血模式。 更宽行业证据也强化了这一点。MERICS 认为,许多中国人形机器人当前平均成本仍为 RMB300,000–500,000,并引用约 RMB160,000 的商业可行阈值,对应两年 ROI 场景。SCIO 和标准化推进意味着成本下降应来自模块化与互操作性,但这些是未来助推器,不是当下事实。换句话说,在有吸引力的单位经济性得到证明前,该品类的公开数字仍显示沉重的工程、数据、制造和服务负担。Bulage 今天没有披露这些项目中任何一项的数据。[CI011, CI012, CI013, CI014, CI015, CI016]
| 指标 | 公开数值 / 状态 | 置信度 | 重要性 | 可比证据 | 尽调请求 |
|---|---|---|---|---|---|
| Bulage 收入 | 未披露 | 公开不可得,置信度高 | 没有收入基数,就无法判断收入质量或增长 | 官网和融资报道均未披露 | 要求提供截至目前收入、订单额和收入确认政策 |
| Bulage 毛利率 | 未披露 | 公开不可得,置信度高 | 毛利率路径是机器人公司的核心 | 公开同业很少披露已兑现合同级毛利率 | 要求按收入流提供毛利率和支持服务附着率 |
| Bulage 烧钱 / 跑道 | 未披露 | 公开不可得,置信度高 | 种子轮规模只有放在烧钱速度旁边才有意义 | 同业融资显示回收周期很长 | 要求提供月度烧钱、现金和 18 个月经营计划 |
| 中国可行性门槛 | 两年 ROI 约需每台 RMB160,000 | 中 | 说明要大规模采用,成本还需压缩多少 | MERICS 引用 Guotai Securities 门槛 | 用该门槛检验 Bulage 目标成本结构 |
| 中国机器人平均成本区间 | 平均 RMB300,000-500,000;低 / 中端价格带 RMB100,000-500,000 | 中 | 公开价格区间显示距离大众可负担仍很远 | MERICS 和 Research and Markets | 要求 Bulage 目标 ASP 和计划降本曲线 |
| UBTECH 毛利率 | FY2025 为 37.7% | 高 | 说明早期规模仍可伴随持续亏损 | HKEX 文件和 Yicai 摘要 | 对标 Bulage 预期的组合经济性更好还是更差 |
| UBTECH 研发强度 | FY2025 研发支出 RMB507.5M,占收入 25.4% | 高 | 具身 AI 即便有收入规模,仍然吃研发 | HKEX 文件和 RobotToday 摘要 | 要求 Bulage 计划研发强度和招聘画像 |
| Agility 订单管线 | > $300M 多年期 Digit v5 订单,30+ 客户 | 高 | 管线比宽泛 TAM 口号更能作为早期信号 | Nasdaq 新闻稿和 SEC 证物 | 要求 Bulage 管线质量、里程碑结构和转化假设 |
本章用公开可比公司界定 Bulage 最终应被衡量的指标。多数 Bulage 专属指标仍不可得,因此表格刻意保留大量缺口。
[CI011, CI012, CI014, CI015, CI017, CI024]从组件与运营负担,到人形机器人的毛利和回本周期,建立定性桥梁。
[CI010, CI015, CI017, CI025, CI033, CI036]4.4 资本充足性与承销结论
Bulage 据报道的 $220 million 种子轮按绝对值很大,但周边证据表明,这只是起点,不是财务充足性的证明。Figure 单轮融资 $675 million,用于加速 AI 训练、制造和部署。Agility 正推进一笔总收益超过 $620 million 的交易,同时称已有超过 $300 million 的多年订单和 30 多个客户管线。UBTECH 做到 RMB2.0 billion 收入后,仍报告大额亏损和大量 R&D 需求。1X 融资 $100 million,仍把家庭部署描述为持续建设。共同模式是:该品类在稳定盈利可见前,会长时间消耗资本。 因此,承销结论偏谨慎。Bulage 可能有足够现金支持核心研究、早期硬件迭代,以及一个或多个试点项目,但公开证据不足以承销收入质量、毛利率、烧钱速度、营运资本需求或现金跑道。如果 Bulage 打算打造全栈人形机器人平台,资本负担很可能包括算力、数据采集、原型迭代、安全工程、制造搭建、部署支持和伙伴集成。如果它打算做 OEM 中立模型层,成本基数可能更轻,但仍取决于能否证明真实任务表现。无论哪条路径,当前估值更多压在期权价值上,而不是已披露财务牵引力上。[CI022, CI023, CI024, CI027, CI028, CI029]
| 项目 | 公开数值 | 来源 / 可比对象 | 资金用途或含义 | 投研解读 |
|---|---|---|---|---|
| Bulage 种子轮 | $220M,估值 $2B | Crunchbase 和中文融资报道 | 足以支撑研究、招聘和试点 | 有助于起步,但不能证明资金足够 |
| Figure Series B 轮 | $675M,估值 $2.6B | Figure 新闻稿 | AI 训练、制造、人员扩张、部署 | 前沿人形机器人项目吸收的资本可能远超种子轮 |
| Agility 交易募资 | 总募资 > $620M | SEC 证物 / Nasdaq 新闻稿 | 履行订单、扩大部署、扩展 Digit v5、投资软件和安全 | 即便有客户,工业规模化仍然烧钱 |
| Agility 已签约订单 | 多年期订单 > $300M | Nasdaq 新闻稿 | 提供部分需求支撑的融资逻辑 | 订单比抽象 TAM 更重要 |
| UBTECH 现金 | 2025 年底 RMB4,887.9M | RobotToday 年度业绩摘要 | 支撑继续规模化和研发 | 现金强劲仍可与亏损并存 |
| UBTECH 净亏损 | FY2025 净亏损 RMB789.8M | HKEX 文件和 Yicai | 说明规模还不等于盈利 | 提醒不要假设早期毛利率已成熟 |
| UBTECH 2025 年配股 | 2025 年总募资约 HK$7.3B | RobotToday 摘要 | 资产负债表大幅增强 | 外部融资仍是品类规模化核心 |
| Bulage 下一轮可能触发条件 | 首次真实部署和包装证明 | 从同业模式推断 | 需要把期权价值转化为可融资的经营计划 | 未来资本应取决于证据,而不只是叙事 |
历史融资时间线见公司概览;本表聚焦当前可见资本是否足以支撑可能的建设路径。
[CI004, CI006, CI007, CI022, CI023, CI028]| 缺失指标 | 公开状态 | 对投资判断的影响 | 现在为何重要 | 具体尽调路径 |
|---|---|---|---|---|
| 截至目前收入 | 未披露 | 重大 | 无法评估增长或 PMF | 要求月度收入、订单额和已确认收入桥接 |
| 按收入流拆分毛利率 | 未披露 | 重大 | 硬件 vs 软件组合是估值核心驱动 | 要求按产品 / 服务类别拆分毛利率 |
| 账面现金和跑道 | 未披露 | 重大 | 无法检验资本充足性 | 要求最新现金余额、烧钱速度和跑道情景 |
| 客户集中度 | 未披露 | 重大 | 一个锚定试点可能掩盖覆盖面薄弱 | 要求按客户、行业和阶段拆分管线 |
| 合同结构 | 未披露 | 重大 | 里程碑、订金和验收条款决定收入质量 | 要求样本 MSA / SOW 和验收政策 |
| 制造 capex 计划 | 未披露 | 重大 | 全栈硬件路径可能在收入规模化前吃掉现金 | 要求 capex 预算、外包计划和供应商承诺 |
| 服务 / 维护负担 | 未披露 | 中等至重大 | 支持负载会侵蚀毛利率 | 要求人员配比、保修假设和 SLA 模型 |
| 人员规模和研发节奏 | 未披露 | 中等至重大 | 人才烧钱往往先于收入推高现金需求 | 要求人员计划、算力支出和里程碑路线图 |
这些不是表面披露缺口。它们是判断 Bulage 究竟是软件型期权,还是资本密集型硬件项目所需的最低数据。
[CI001, CI021, CI024, CI027, CI032, CI035]映射全栈具身 AI 初创公司在稳定收入质量被证明前必须融资覆盖的主要现金消耗。
[CI022, CI023, CI027, CI028, CI035, CI038]4.5 图表
05产品与技术
5.1 产品定义与公开界面
相比估值,Bulage 的公开产品定义仍薄得多。官网没有可读产品文档、模块列表、信任页面、API 或开发者入口,也没有路线图。主要公开描述来自媒体报道:Lin Junyang 的新公司聚焦世界模型和具身大脑。这是有意义的战略方向,但还不是产品地图。它没有说明 Bulage 销售的是一体化人形机器人、第三方机器人的控制栈、训练数据管线、开发者工具,还是定制部署。因此,正确的产品技术框架要从缺口开始:Bulage 有主题,但没有可公开检查的产品。 周边赛场说明,公开产品定义通常可以具体得多。Figure 展示面向家庭的人形机器人界面和很深的 Helix 技术叙事。AGIBOT 发布多种机器人形态和开放数据集计划。UBTECH 发布工业产品页、年报中的架构语言,以及开源 Thinker 模型。NVIDIA、DeepMind 和 Physical Intelligence 都展示了行业如何收敛到显式栈:推理模型、行动策略、仿真、数据管线、硬件和安全基础设施相结合。Bulage 还没有揭示自己位于这套栈的哪个位置。[CE001, CE002, CE003, CE021, CE030, CE034]
| 模块 / 资产 | 用户 | 状态 / 成熟度 | 差异化信号 | 尽调缺口 |
|---|---|---|---|---|
| Bulage 具身大脑 / 世界模型命题 | 未知;可能是 OEM 或企业运营方 | 仅见媒体说法 | 贴合前沿具身 AI 方向 | 没有公开模块、基准或架构细节 |
| 官网 / 文档界面 | 研究者、合作伙伴、客户 | 基本缺席 | None | 没有可读的产品、文档、信任或路线图页面 |
| 机器人本体 / 硬件平台 | 运营方和集成商 | 未披露 | Unknown | 没有自有机器人、合作伙伴机器人或本体规格的公开证据 |
| 训练数据管线 | ML 和机器人团队 | 未披露 | 如果跨本体且任务足够丰富,可能成为护城河 | 没有公开数据集、采集策略或标注细节 |
| 仿真 / 数字孪生层 | 模型和控制团队 | 未披露 | 对放大迭代很重要 | 没有公开仿真栈或数字孪生证据 |
| 安全 / 信任层 | 终端用户和企业买家 | 未披露 | 真实世界使用的关键 | 没有公开质量、隐私、可靠性或安全材料 |
这些行反映可信具身 AI 技术栈需要什么。对 Bulage 而言,多数行仍未验证,而不是已完成模块。
[CE001, CE002, CE020, CE021, CE030, CE031]分层展示 Bulage 若要让其「具身大脑」命题落地,需要披露或搭建的产品架构。
这是根据公开同行披露综合出的概念栈。Bulage 尚未公开确认任何层的实现选择。
[CE001, CE003, CE015, CE018, CE020, CE026]5.2 参考架构与具身模型栈
2025–2026 年具身 AI 披露中最清晰的技术模式,是高层推理与低层控制分层。Figure 的 Helix 把较慢的推理 VLM 与快速视觉运动控制器分开。DeepMind 的 Gemini Robotics 1.5 把具身推理模型与 VLA 执行器配对。Physical Intelligence 的 π0 使用预训练 VLM,加上在广泛机器人数据上训练的 flow-matching 行动头。NVIDIA 的 GR00T 平台再把这种模式变成通用生态供给:基础模型、开放数据、仿真、中间件和部署硬件。这些架构重要,因为它们说明“世界模型”或“具身大脑”在真实出货系统中可能到底意味着什么。 它们也让 Bulage 当前披露缺口更刺眼。如果 Bulage 做的是模型层公司,行业已经期待它回答跨具身迁移、推理延迟、遥操作或模仿数据、仿真和硬件抽象。如果它做的是全栈机器人,期待会进一步扩大到机械操作、电池与热管理、现场可靠性,以及对人安全的运行。早期保持安静本身没有错,但公开资料也没有基础假定 Bulage 已经把这套栈中的任何具体环节做得好过如今已有充分文档的参考组。[CE003, CE004, CE005, CE006, CE007, CE008]
| 用户任务 | 当前工作流 | 公开解决方案信号 | 可衡量收益 | 局限 |
|---|---|---|---|---|
| 一般家务任务 | 杂乱环境中的人力劳动 | Figure 03、1X NEO、π0、Gemini 演示 | 可能带来更广泛自主性和便利 | 安全、支持和泛化仍难 |
| 工业物料搬运 | 人工 + 固定自动化 | Agility Digit、UBTECH Walker、Unitree 硬件 | 替代人工,提高在线率 | 定制集成和安全负担 |
| 机器人技能训练 | 遥操作、脚本演示、人工数据标注 | AGIBOT World、Isaac Lab、GR00T 等平台 | 更可扩展的学习和 sim-to-real 迁移 | 数据质量和迁移仍是瓶颈 |
| 具身推理 / 规划 | 特定任务管线 | Gemini ER、Helix S2、Thinker 规划 | 更好的长时程规划和物理推理 | 需要落到真实控制回路 |
| 跨本体策略迁移 | 逐台机器人调参 | Gemini 1.5、π0、GR00T | 降低跨机器人本体适配成本 | 仍早期,尚未普遍证实 |
| Bulage 目标工作流 | 未披露 | 仅有世界模型 / 具身大脑叙事 | 可能成为 OEM 中立智能层 | 没有公开用例、ROI 或工作流定义 |
本表用公开参考对象界定 Bulage 可能瞄准的工作流;Bulage 自身尚未说明具体要替用户完成什么任务。
[CE004, CE005, CE006, CE007, CE012, CE014]| 层级 / 组件 | 作用 | 依赖 | 风险 |
|---|---|---|---|
| 推理模型 / 世界模型 | 高层任务理解和规划 | 大规模多模态预训练 | 没有具身落地,可能仍过于抽象 |
| 视觉-语言-动作策略 | 把指令和感知转成运动命令 | 具身数据和低延迟控制 | 长时程、未见过或安全关键任务上可能失效 |
| 仿真 / 数字孪生 | 放大训练和评估 | 准确物理、场景和机器人模型 | sim-to-real 差距可能吞掉收益 |
| 机器人本体 / 执行机构 | 执行操作和移动 | 机械设计、动力、传感器、控制器 | 硬件约束可能盖过模型本身的卖点 |
| 数据管线 | 收集并标注经验数据 | 遥操作、标注、质控 | 数据质量差会削弱泛化能力 |
| 安全 / 信任控制 | 保护人员、财产和数据安全 | 标准、监控、支持流程 | 缺少安全层会卡住部署 |
| 部署 / 集成工具 | 嵌入客户实际工作流 | API、中间件、车队管理、支持 | 运营摩擦可能扼杀采用,即便模型很强 |
这张架构表有意把软件、硬件、数据和运营放在一起看,因为公开发声的技术负责人已不再把它们拆成彼此独立的问题。
[CE003, CE007, CE008, CE009, CE015, CE017]根据 Helix、Gemini Robotics、π0、GR00T 和 AGIBOT World 披露综合出的公开运营流程。
[CE004, CE005, CE006, CE007, CE008, CE012]5.3 数据、仿真、安全与依赖
2025–2026 年公开技术语料给出的最醒目教训是:具身智能既是模型问题,也是数据和安全问题。AGIBOT World 2026 通过突出异构真实世界数据、自由式遥操作、全身控制、多模态感知和 1:1 数字孪生,把这一点讲得很直接。NVIDIA Isaac Lab 和 GR00T 则从平台侧做了同样的事,强调可规模化机器人学习工作流和 sim-to-real 迁移。DeepMind 的 Gemini Robotics 1.5 和 π0 都显示,更宽的预训练只有经过大量具身数据适配后才变得有用。UBTECH 的 Thinker 和年报语言展示了类似的中国框架:空间智能、时间理解、世界模型和工业机器人协作,都离不开数据质量与全栈执行。 安全和信任位于这张依赖图之上。NVIDIA Halos 把物理 AI 描述为需要全栈安全 OS、检查制度、标准对齐和外部监控。SCIO 的标准工作和 MERICS 的政策研究说明了这在中国为何重要:机器人正在进入工厂、门店、医院和家庭,可靠性与安全不再是可选附加项。Bulage 目前没有公开展示自己的安全、隐私、质量或支持控制,这是一项产品技术风险,而不只是披露脚注。[CE009, CE010, CE011, CE012, CE018, CE019]
| 控制 / 质量要素 | 状态 | 范围 | 缺口 | 影响 |
|---|---|---|---|---|
| Bulage 公开安全文档 | 未披露 | 公司整体 | 没有公开的信任或安全披露面 | 企业买家无法公开尽调 |
| Bulage 隐私 / 安全态势 | 未披露 | 公司整体 | 没有公开隐私、安全或事件材料 | 削弱面向人的机器人的信任基础 |
| 中国人形机器人标准框架 | 已发布 | 全行业 | 落地仍处早期 | 提升可比性,但不能立刻形成证明 |
| NVIDIA Halos 功能安全栈 | 生态内可用 | 机器人安全平台 | 非 Bulage 专属 | 显示安全预期正在快速上升 |
| Gemini 安全框架 | 技术文档已发布 | 语义安全和政策对齐 | 合作伙伴可用性有限 | 显示安全已成模型的一等问题 |
| UBTECH 工业质量 / 支持信号 | 公开但不完整 | 工业部署 | 未披露完整可靠性指标 | 好过沉默,但仍不是完整证明 |
本表把生态层面的安全进展和 Bulage 自身披露分开,因为后者目前缺位。
[CE018, CE019, CE020, CE029, CE032, CE035]有向依赖图,展示具身产品交付如何依赖数据、仿真、硬件、安全和标准层。
这张图是分析性归纳,但直接基于保留的公开技术披露。Bulage 没有公开说明自身依赖的责任方或合作伙伴。
[CE007, CE008, CE009, CE012, CE018, CE029]5.4 成熟度、路线图与技术结论
公开成熟度在赛场内并不均匀。Figure、DeepMind、NVIDIA、Physical Intelligence、AGIBOT、UBTECH 和 Unitree 都至少展示了一层可检查内容:产品页、技术报告、数据集发布、仿真框架、开放 repo,或与标准相连的安全材料。其中一些主张可能仍有愿景成分,但它们给尽调留下了可挑战的界面。相比之下,Bulage 几乎没有提供这些东西。没有公开模块图、公开发布里程碑、公开开发者社区,也没有公开集成、可靠性或支持证据。 这并不意味着 Bulage 的策略无效。事实上,行业技术方向说明,如果一家公司能在不同机器人身体和任务之间泛化,同时把推理、安全和集成保持在可管理范围内,OEM 中立具身大脑仍有空间。但这仍是假设,不是已展示产品。因此,技术结论是有条件的:Bulage 的方向与前沿演进一致,但在公开透明度和证明标准上,明显落后于同行已经设定的基线。在公司披露栈选择、合作伙伴、安全控制和若干任务级表现证据前,应把它的产品技术案例视为有前景但未验证。[CE013, CE014, CE024, CE025, CE026, CE027]
| 日期 / 阶段 | 功能 / 里程碑 | 状态 | 影响 | 来源 |
|---|---|---|---|---|
| 2024-10 | π0 通用策略发布 | 已发布 | 显示多机器人基础策略方向已经公开且在推进 | Physical Intelligence |
| 2025-02 | Figure Helix 官宣 | 已发布 | 全上半身 VLA 成为具体产品主张 | Figure |
| 2025-09 | Gemini Robotics 1.5 博客与报告 | 已发布 | 推理加动作架构和跨本体迁移成为公开基准 | Google DeepMind |
| 2026-01 | UBTECH Thinker 开源与年报技术栈表述 | 已发布 | 中国同行正在打开模型和世界模型界面 | UBTECH |
| 2026-01 | AGIBOT 出货和生态规模叙事 | 已发布 | 中国同行把数据和部署叙事串在一起 | AGIBOT |
| 2026-07 | AGIBOT WORLD 2026 分阶段发布 | 已发布 | 大型开放数据集和数字孪生策略变得可见 | AGIBOT |
| 2026-07-22 | Bulage 公开技术栈披露 | 仍缺失 | Bulage 在公开尽调中仍主要依赖投资论点 | Bulage 官网与媒体 |
路线图条目混合了产品、数据集、模型和安全相关里程碑,因为具身 AI 真正发布的单位就是这些。
[CE004, CE005, CE010, CE012, CE022, CE024]围绕 Bulage 投资命题最关键的模块,对比各方能力与成熟度。
[CE001, CE007, CE009, CE010, CE012, CE013]5.5 图表
06客户
6.1 公开客户界面与证明缺口
Bulage 的公开客户记录实际上仍是空白。官网没有客户页面、案例库、合作伙伴故事、定价或试用流程,也没有部署参考。2026 年 6 月把 Bulage 头条估值立起来的融资报道,也没有填上这个缺口;它描述的是创始人背景和世界模型 / 具身大脑论点,但没有识别出具名客户、试点地点、设计伙伴或付费工作流。也没有买方证言、采购线索或用户引语,能说明 Bulage 已经改善任何生产任务。这意味着客户尽调的起点结论是负面但重要的:目前尚无公开证据表明 Bulage 已把技术故事转化为真实买方关系。 这个缺口在 2026 年比一年前更重要,因为参照组已经具体得多。Agility 和 GXO 公开描述从试点到付费部署的路径。BMW 公开报告其 Figure 试点的运行小时数和组件数量。Hyundai 公开列出分阶段 Atlas 部署计划。UBTECH 公开点名多家汽车和制造业客户。AGIBOT 和 KEENON 展示出更清晰的买方分层和商业化界面。Bulage 可能仍太年轻,不适合广泛披露,但投资人不应把这种可能性误认为实际客户证明。[CU001, CU002, CU003, CU004, CU006, CU030]
| 细分市场 | 买方 / 用户 / 付款方 | 用例 | 规模 / 证明质量 | 收入 / 战略价值 | 缺口 |
|---|---|---|---|---|---|
| Bulage 未披露目标客户 | 买方未知、操作者未知、预算负责人未知 | 未披露具身 AI 工作流 | 无公开具名客户证明 | 若绑定 OEM 或大型企业,可能具备战略价值 | 未公开披露客户分层 |
| 物流 / 仓库运营商 | COO、自动化负责人、站点经理 | 料箱搬运、交接、重复性仓储作业 | Agility 与 GXO 提供最强公开证明 | 验证付费部署及工作流软件和服务 | Bulage 没有可比仓储证明 |
| 汽车 / 工业 OEM | 工厂运营、制造工程、质量负责人 | 排序、搬运、装配、检测 | BMW、Hyundai 和 UBTECH 工厂案例提供强证明 | 大体量客户可成为锚定参考和训练场 | Bulage 没有具名工厂客户 |
| 服务 / 医疗 / 酒店运营商 | 场馆运营商、医疗管理者、服务经理 | 接待、配送、医疗辅助、饮品服务 | KEENON 提供早期场景证明 | 拓宽 TAM,但证据更偏演示 | Bulage 尚未展示服务业切入口 |
| 直销 / 开发者 / 产消者买家 | 个人开发者、实验室、SMB 集成商 | 机器人评估、早期访问、实验 | Unitree 和 1X 可见 | 最快形成漏斗顶部需求和社区信号 | Bulage 没有公开购买或候补路径 |
| 中国工业生态买方 | OEM 合作伙伴、供应商、地方政府背书运营商 | 场景部署和扩张 | AGIBOT 和 UBTECH 商业化叙事中可见 | 可在单一地理区域内快速形成部署密度 | Bulage 没有公开地理客户足迹 |
各行把当前人形机器人市场可见的买方形态,与 Bulage 实际披露内容分开;后者仍几乎为零。
[CU003, CU007, CU014, CU017, CU021, CU023]同业公开可见的市场旅程已经清楚;Bulage 从发现到复用部署之间的公开步骤仍然缺失。
这是公众可见证明的旅程图,描述投资人能从外部观察到什么,并非 Bulage 未披露的内部 CRM 阶段。
[CU001, CU002, CU003, CU006, CU030]6.2 同行买方与采用模式
同行证据显示,人形机器人客户形成正分成几条不同路径。一条是工业自动化,物流运营商和汽车 OEM 愿意忍受集成复杂度,换取劳动力替代、安全改善或流程一致性。Agility / GXO、Figure / BMW、Hyundai / Boston Dynamics,以及 UBTECH 的汽车案例都在这个桶里。第二条是服务与酒店场景,KEENON 展示了医疗、休息室和品牌饮品服务场景。第三条是直接销售或开发者式界面,Unitree 和 1X 即便没有完全披露企业留存指标,也让买方定位变得可读。 对 Bulage 来说,这张同行地图有用,因为它揭示了缺失项。没有公开声明说明 Bulage 正追逐仓储运营商、汽车 OEM、零售连锁、医院、机器人 OEM,还是软件授权伙伴。没有已披露渠道策略、生产地域,也没有公开工作流适配证明。换句话说,品类已经开始呈现真实买方形状,而 Bulage 仍主要像一个等待首个可公开检查客户动作的论点。[CU007, CU008, CU011, CU014, CU017, CU023]
| 指标 | 数值 | 日期 | 来源 | 置信度 | 影响 | 缺失分母 |
|---|---|---|---|---|---|---|
| Bulage 具名客户 | 公开具名为 0 | 2026-07-22 | 官网与融资报道 | 高 | 尚无公开采用证明 | 私下管线可能存在,但未披露 |
| Bulage 公开部署阶段 | Null | 2026-07-22 | 官网与融资报道 | 高 | 无法判断是试点还是生产 | 无站点数、账户数或工作流范围 |
| Agility 到 GXO 转化 | 2023 年末试点到 2024 年多年期部署 | 2024-06-27 | GXO 与 Agility | 高 | 展示了从评估到付费铺开的可信路径 | 机器人数量和实际续约经济性仍未披露 |
| BMW 与 Figure 运营证明 | 支持生产 >30,000 辆 BMW X3;搬运 >90,000 个部件;累计运行 ~1,250 小时 | 2026 | BMW | 中 | 同组中最好的公开流程级产出证明 | 未披露合同价值或长期续约 |
| Hyundai / Atlas 推出 | 2028 年起在 HMGMA 分阶段部署目标 | 2026-01-06 | Hyundai | 中 | 显示买方兴趣,也显示商业化周期很长 | 当前已部署数量未披露 |
| AGIBOT 部署叙事 | 到 2026 年 3 月第 10,000 台机器人下线;七套标准化解决方案 | 2026 | AGIBOT | 中 | 显示商业化触面广、场景包装清晰 | 具名客户列表和经常性收入组合仍不完整 |
| UBTECH 具名工业案例 | 多个汽车和工业客户公开具名 | 2026 | UBTECH 与 HKEX | 高 | 中国工厂商业化正在越过概念阶段 | 单账户收入和留存仍未披露 |
null 表示该指标未公开披露。同行指标用于校准 Bulage 尚未拿出的证明水平。
[CU003, CU004, CU005, CU008, CU010, CU012]| 客户 / 账户 | 细分市场 | 部署 / 用例 | 生产 / 试点 | 结果 | 局限 |
|---|---|---|---|---|---|
| Bulage | 未披露 | 未发现公开客户证明 | Unknown | 无公开证据 | 未披露具名账户、工作流或用户结果 |
| GXO 与 Agility | 物流运营商 | Digit 料箱搬运和仓库工作流集成 | 试点转为多年期付费部署 | 人形机器人从试点走向客户现场收入的最佳公开证明 | 机器人数量、合同价值和长期续约条款仍未公开 |
| BMW 与 Figure | 汽车 OEM | Spartanburg 车身车间钣金搬运 | 有具体运营产出的试点,并继续评估更多用例 | 支持生产超过 30,000 辆 X3,搬运超过 90,000 个部件 | 公开经济性和扩张速度仍未披露 |
| Hyundai 与 Boston Dynamics | 汽车 OEM | HMGMA / Georgia 工厂的 Atlas 排序和未来装配工作 | 真实工作测试加分阶段推出计划 | 来自母公司整合的强战略客户信号 | 主要扩张排到 2028 年及以后 |
| UBTECH 汽车客户 | 汽车 / 工业 OEM | 搬运、装配、检测和工厂物流 | 多个具名工业应用案例 | 所审阅样本中最广的中国工厂具名客户清单 | 单账户利用率、留存和经济性未公开 |
| KEENON 服务场景 | 医疗 / 酒店 / 服务运营商 | 医疗站、休息室酒吧和品牌饮品服务 | 以演示为主的场景证明 | 显示买方分层和面向人的工作流定位 | 仍弱于已披露且长期运行的生产部署 |
每一行都至少基于两条保留来源,或一组保留来源加一个佐证官方背景页。Bulage 仍是唯一没有正面公开证明的行。
[CU003, CU007, CU008, CU011, CU012, CU014]公开证明从宽泛的商业化表述急剧收窄,最后只剩少数真正披露具名客户、运营产出或连续性的案例。
计数指已审阅来源集中的证明类别,不代表市场上机器人或客户的完整数量。
[CU006, CU010, CU014, CU017, CU031, CU037]工业部署目前拿出的公开证明质量最强;Bulage 仍没有可见的买方级证据。
单元格表达证据质量,不代表收入贡献或产品质量。
[CU003, CU010, CU014, CU017, CU023, CU030]6.3 留存、扩张与集中度
整个人形机器人领域的留存证据仍然稀少,Bulage 更是一片空白。已审阅来源没有披露 Bulage 的 NRR、GRR、流失率、续约率、合同期限、扩张收入,甚至没有第一个公开客户 logo。即便是证据更强的同行,公开客户材料仍偏向首次部署公告,而不是长期账户经济性。 Agility 的 GXO 案例值得注意,正是因为它披露了 2023 年底试点后签下多年协议;BMW 的 Figure 披露也值得注意,因为它发布了运营产出,而不是证明长期续约。 UBTECH 具名工厂案例强于多数同行,但即便如此,公开队列耐久度仍弱于部署广度。已审阅来源没有显示 Bulage 在客户现场的上线、培训、支持或服务责任归属。 Bulage 成立时间太短,这会放大客户集中问题。最早一批客户很可能会过度影响融资叙事、背书质量和产品迭代。如果首批账户集中在某个创始人网络、某个 OEM 或某个地区,公司在更广泛买方出现前,可能看起来比真实情况更强。 因此,严谨判断应先要求客户验证里程碑,再为采用买单:具名试点、部署范围、任务频次、正常运行时间、用户责任归属,以及至少一些重复预算或扩张信号。落到执行,就是要材料,不要口号。[CU015, CU016, CU031, CU032, CU036, CU037]
| 指标 | 数值 | 细分市场 | 置信度 | 尽调要求 |
|---|---|---|---|---|
| NRR | Bulage 所有客户 | 低 | 一旦存在任何付费账户基础,索取按队列拆分的 NRR | |
| GRR / 客户流失 | Bulage 所有客户 | 低 | 索取试点到生产转化,以及任何已流失评估客户清单 | |
| 合同期限 | Bulage 企业账户 | 低 | 索取标准试点期限、付费期限和续约机制 | |
| 重复购买证据 | Bulage 具名账户 | 低 | 索取任何首个客户的第二张 PO 或扩张范围 | |
| 同行连续性证明 | GXO 试点到多年期部署 | 工业人形机器人参考集 | 中 | 要求 Bulage 展示任何类似转化路径 |
| 公开纵向客户指标 | 即便在同行中也很薄 | 人形机器人行业 | 中 | 用最强同行披露给 Bulage 定标,而不是泛泛的机器人乐观叙事 |
null 表示未公开披露。所审阅样本中唯一可见的连续性信号,是试点推进到部署的叙事证明,而不是完整 SaaS 式留存报告。
[CU008, CU010, CU015, CU031, CU036, CU037]| 扩张驱动因素 | 集中度风险 | 影响 | 尽调路径 |
|---|---|---|---|
| 首个旗舰工业客户 | 单一客户叙事依赖 | 一个可见客户就可能主导 Bulage 的可信度和产品路线图 | 索取按垂直行业拆分的完整管线,而不只是旗舰名称 |
| 创始人网络或投资人介绍的试点 | 关系集中 | 可快速带来早期胜利,但会高估广泛市场拉力 | 按投资人、创始人和合作伙伴映射介绍来源 |
| 中国优先制造业部署 | 地理集中 | 可能加速学习,但海外需求、合规和服务能力仍未验证 | 索取客户地域、出口计划和服务足迹 |
| OEM 中立的具身大脑策略 | 集成集中 | Bulage 可能依赖少数硬件或 OEM 合作伙伴触达用户 | 索取已签署 OEM 或集成商关系以及排他条款 |
| 类别热度和商业化时点 | 采购摩擦和扩张停滞 | 机器人热情可能跑在真实重复预算和安全准备之前 | 把任何签约额叙事与实际任务频率、正常运行时间和扩张 PO 对齐 |
本表拆出首客故事看起来强于真实需求耐久度的主要路径。
[CU013, CU019, CU027, CU029, CU030, CU040]多数人形机器人买方只公开到入门队列可见度;Bulage 目前连初始公开队列都没有。
百分比代表公开连续性证据的可见度,不代表实际客户留存率。0 表示未发现公开复用信号。
[CU003, CU008, CU012, CU014, CU017, CU036]6.4 图表
07风险
7.1 按严重度排序的风险图景
Bulage 的风险画像不寻常:几乎所有重大尽调问题都同时悬而未决。公司成立不到一年,公开叙事绑定在雄心勃勃的具身大脑论点上;在展示具名客户、公开安全控制或运营指标之前,报道估值已达数十亿美元。 这并不证明业务弱,但意味着外部证据还无法给下行封顶。真正的投资判断难题,是不确定性叠加:合规负担、算力访问、客户证明和烧钱速度各自重要,又会相互强化。 行业背景让这组风险更难忽视。同行现在披露分阶段工厂部署、工业案例页面、交易级融资,以及越来越正式的合规语境。与此同时,分析师和行业观察者提醒,商业化热度可能跑在持久价值前面。 因此,投资者应把 Bulage 风险章节当成排序练习,而不是泛泛的免责声明列表。严重度最高的风险,是那些会同时伤害商业化、融资和估值叙事的风险:控制面缺失、出口管制与工具链依赖、资金需求强烈,以及证据落后于同行。[CR001, CR002, CR003, CR017, CR023, CR026]
| 失效模式 | 公开信号 | 可能性 | 严重性 | 缓解成熟度 | 剩余暴露 | 未解缺口 |
|---|---|---|---|---|---|---|
| 控制面薄于估值面 | bulage.cn 没有公开的信任、隐私、事件或客户证明页面 | 高 | 高 | 低 | 高 | 没有公开证据说明谁负责合规、支持或安全 |
| 模型 / 机器人范围混乱 | 公开材料从未说清 Bulage 卖模型、机器人、服务,还是三者都卖 | 高 | 高 | 低 | 高 | 没有模块或工作流地图 |
| 安全性与可靠性未知 | 没有公开的可用时间、事故、MTBF 或任务成功率指标 | 高 | 高 | 低 | 高 | 没有外场表现或保障人员安全的部署控制证据 |
| 工具链或算力中断 | 具身 AI 技术栈依赖和出口管制暴露仍未解决 | 中 | 高 | Unknown | 高 | 没有公开的备选或本地化策略 |
| 采购证明滞后 | 同业已披露部署参考,Bulage 仍未披露 | 高 | 中-高 | 低 | 高 | 没有具名试点、生产场地或扩展订单 |
运营项聚焦外部投资者仍无法用公开证据框定的内容。
[CR001, CR003, CR017, CR031, CR032, CR035]残余风险最高的,是公开证据最薄、向估值传导最快的领域。
[CR003, CR017, CR023, CR026, CR036, CR037]Bulage 的主要风险相互作用,并非彼此隔离;因此,一个缺失的控制面会很快变成融资问题。
[CR017, CR023, CR035, CR037, CR038, CR039]7.2 监管、法律和地缘政治敞口
Bulage 所在的 AI 栈位置,法律和地缘政治敞口可能比收入扩张更快。中国已经围绕数据来源、个人信息、内容处理、透明度和安全评估,监管面向公众的生成式 AI 服务。 欧洲正在落地基于风险分级的 AI Act;系统一旦触及安全、劳动或受监管产品场景,义务就会加重。欧盟和美国的出口管制框架同样重要,因为具身 AI 公司依赖软件、技术协助、芯片和算力基础设施;即便机器人本体在本地制造,这些要素也可能跨司法辖区流动。 这些框架都不说明 Bulage 眼下已被卡住。风险在于,公司公开合规控制面很少,可能只有在尝试大规模销售、境外融资、跨境合作或接入客户运营时,才发现这些义务。 法律先例也有参考价值。专业律所的行业 M&A 材料反复强调机器人交易中的监管审批。这并不是说 Bulage 现在有交易麻烦,而是说明周边行业只是转移资产和伙伴关系,就已经需要法律基础设施。 对一家有前沿 AI 野心的上海创业公司来说,这是实打实的提醒:跨境增长可能比种子轮叙事暗示的更慢,也更贵。[CR004, CR005, CR006, CR007, CR009, CR010]
| 风险 | 司法辖区 / 触发条件 | 当前证据 | 可能性 | 严重性 | 缓释成熟度 | 剩余敞口 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| 中国 AI 服务合规缺口 | 中国面向公众的 AI 或 API 服务 | CAC 措施要求数据合法、保护个人信息、稳定运行,并在部分情形下要求安全评估 | 中 | 高 | Unknown | 高 | 索取法律负责人、备案状态、训练数据来源和服务治理控制 |
| 欧盟 AI Act 与合格评定负担 | 欧盟客户、受监管产品,或欧洲部署 | AI Act 义务、透明度规则和标准工作已经推进 | 中 | 高 | Unknown | 高 | 要求提供产品和市场地图,标出哪些系统可能属于高风险或受透明度要求覆盖 |
| 出口管制 / 工具访问摩擦 | 美国来源工具、芯片、软件、技术协助;欧盟两用物项 | EAR 和欧盟两用规则管控软件、技术和协助,而具身 AI 技术栈仍依赖外部技术 | 中 | 高 | Unknown | 高 | 要求提供工具链依赖地图,以及任何产品级出口分类或受限供应商暴露 |
| 跨境审批节奏风险 | 海外合作、并购、分销,或战略交易 | 行业法律顾问反复提示,机器人交易可能需要监管审批 | 低-中 | 中 | Unknown | 中 | 要求律师备忘录说明计划中的跨境架构和审批顺序 |
| 标准与文档滞后 | 工业部署和跨国客户 | 中国标准建设和欧盟合格评定标准都指向更重的文档要求 | 中 | 中 | 发展中 | 中 | 要求按目标用例提供标准矩阵、测试计划和认证路线图 |
各行按严重程度排序,判断依据是公开监管文本和法律先例,而不是 Bulage 内部备忘录。
[CR004, CR005, CR006, CR007, CR008, CR009]在内部缓释措施可见之前,Bulage 的公开证据已经显示出对外部监管、外国工具和未来客户的依赖。
[CR007, CR011, CR017, CR033, CR035, CR037]7.3 运营、商业和人才风险
运营层面,最大挑战是 Bulage 还没有发布足以让投资者判断公司类型的材料:它究竟是模型层公司、机器人产品公司,还是重集成的部署业务。 公开可靠性记录、公开安全论证、公开质量看板、隐私姿态和公开伙伴图谱都不存在。也就无法区分可管理的早期不透明和真正的控制缺口。商业侧又放大了问题。 证据更充分的同行显示,即便成功部署,仍然是分阶段、烧资本、服务强度高。UBTECH 的公开业绩仍在亏损。Agility 即便有订单,也需要巨额融资。Figure 和 Hyundai 说明,工业落地需要时间。 AGIBOT 和 UBTECH 还说明,竞争压力现在不只来自产品 demo,还来自公开部署证据。 人才风险同样重要。Bulage 的公开身份仍集中在 Lin Junyang 和创始人故事上。公司如果不披露运营、合规、制造、销售和客户成功梯队,商业化复杂度上升时,就可能过度依赖一个小核心团队。 年轻机器人公司最先遇到的隐性瓶颈,往往就是执行深度。[CR018, CR019, CR020, CR021, CR022, CR024]
| 依赖项 | 交易对手 / 资产 | 角色 | 集中度 | 失效情景 | 严重性 | 缓解措施 | 剩余暴露 |
|---|---|---|---|---|---|---|---|
| 先进算力和海外工具 | 美国来源芯片、框架或技术协助 | 模型训练和迭代 | 未知但可能很高 | 许可、供应或访问摩擦拖慢进展 | 高 | 中国政策强调国产替代和共享算力资源 | 高 |
| 首个旗舰客户 | 未披露的初始设计伙伴或买家 | 收入叙事和产品反馈闭环 | 初期可能较高 | 单一客户主导路线图和标杆质量 | 高 | 无公开披露 | 高 |
| 硬件 / OEM 集成伙伴 | 未披露的机器人本体或部署伙伴 | 具身落地和客户触达 | Unknown | 没有真实集成渠道,OEM 中立论点就站不住 | 高 | 无公开披露 | 高 |
| 监管机构和标准组织 | CAC、中国工业标准生态,后续可能还有欧盟监管机构 | 许可和合规背景 | 中 | 销售或扩张被治理工作拖慢 | 中 | 公开层面只看得到行业级政策认知 | 中 |
| 资本市场和未来投资者 | 后续私募资本或战略投资者 | 现金跑道和选择权 | 高 | 证明里程碑到来前,市场情绪先转向 | 高 | 大额种子轮买来时间,但买不来免疫力 | 高 |
关键依赖风险并不是 Bulage 已披露的某个单一伙伴;真正的问题是目前根本没有披露依赖地图。
[CR017, CR023, CR026, CR035, CR036, CR039]| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重性 | 缓解措施 | 尽调路径 |
|---|---|---|---|---|---|
| 创始人 / 研究负责人 | 公开身份围绕 Lin Junyang 展开 | 中 | 高 | 创始人履历强,投资人背书足 | 要求完整组织架构图、决策权划分和二层负责人 |
| 合规 / 法务负责人 | 未公开披露负责人 | 高 | 高 | 无公开信息 | 要求说明隐私、数据、出口和客户安全对应负责高管 |
| 商业化和客户成功 | 没有销售或部署团队厚度的公开证据 | 高 | 中-高 | Unknown | 要求提供垂直负责人、解决方案工程师和支持人员配置计划 |
| 运营 / 质量 / 制造 | 没有现场支持或生产运营的公开证据 | 高 | 中-高 | Unknown | 要求提供部署、支持和质量管理架构 |
早期机器人公司往往先撞上组织瓶颈,技术瓶颈才会在公开层面显形。
[CR002, CR003, CR037, CR038]7.4 缓释因素、监控和止损标准
明显的缓释因素也存在。Bulage 拿到了规模很大的种子轮弹药,站上及时的市场主题,创始人叙事也足够强,能快速吸引顶级资本。 更大的行业也在验证几件事:具身 AI 确实有客户,工厂可以成为标杆现场,标准工作正在成熟。这些都是实打实的正面因素。但它们还不是 Bulage 自身的缓释因素,因为公司还没有展示自己真正掌握哪些控制、客户或工作流。 现阶段的缓释故事,更像期权价值,而不是已验证的流程纪律。 因此,监控负担必须说清楚。投资者应要求具名客户证明、合规负责人和路线图、算力与工具链依赖的清晰说明,以及一些公开或私下的安全与可靠性证据。 如果这些材料迟迟不出现,正确反应不是无限期等市场继续兴奋,而是用严得多的标准重做估值判断。下面的止损标准不是泛泛的谨慎信号;它们是让 Bulage 继续处在可投区间的最低里程碑。[CR003, CR017, CR023, CR026, CR032, CR035]
| 风险 | 可监控触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 客户证明缺口 | 具名客户或试点披露 | 下一个主要尽调周期内仍没有具名账户或限定范围的试点 | 重新核保采用率假设,并下调估值溢价 |
| 合规不透明 | 合规路线图和负责人 | 数据、隐私、AI 治理和安全义务没有明确负责人 | 治理材料可用前暂停尽调 |
| 算力 / 工具链依赖 | 依赖地图和备选方案 | 严重依赖受限或难替代的外国工具,且没有应急方案 | 将路线图视为易受地缘政治摩擦冲击 |
| 资本强度 | 现金跑道和烧钱可见度 | 下一次融资需求出现前,看不到通向证明里程碑的路径 | 假设稀释或估值下调轮风险急升 |
| 运营安全 / 可靠性 | 试点指标和事故报告 | 真实部署没有任务成功率、可用时间或安全证据 | 不要用演示或融资标题核保规模化主张 |
终止标准要具体到投资者可跟踪,而不是模糊到可以忽略。
[CR023, CR026, CR032, CR038, CR039, CR040]7.5 图表
08估值
8.1 价格与证据
Bulage 报道中的约 $2B 估值,只能理解为给期权价值定价,而不是给已披露运营证据定价。公开来源仍没有说明收入、具名客户、产品架构、任务级部署指标,或重复使用的可靠证据。 它们能确认的是创始人履历、大额种子轮、诱人的市场主题,以及一个投资者仍在竞相下注人形机器人和具身 AI 叙事的赛道。这组要素可以产生真实价值,但也意味着当前价格更多由可能性支撑,而不是可验证牵引力。 这个区别很重要,因为同行数据现在设下了比一年前更高的门槛。Figure 把估值与 BMW 和 Helix 绑定。Agility 把估值与订单和客户数量绑定。 Apptronik 把大额融资与 Mercedes-Benz、GXO、Jabil 以及生产型机器人叙事绑定。UBTECH 和 AGIBOT 说明,中国同行至少愿意发布收入、出货或部署信号。 按这个标准看,Bulage 的账面估值并非离谱,但它显然跑在了公开证据曲线前面。[CV001, CV002, CV003, CV005, CV006, CV007]
| 建议 | 置信度 | 风险评级 | 估值立场 | 决策含义 |
|---|---|---|---|---|
| 继续研究 | 中 | 高 | 偏高 | 在目前公开记录之外拿不到客户、产品和治理证明前,不要按传闻中的 $2B 估值核保 |
该判断对证据和价格都敏感:证明更好或价格更低,都可能改变立场;但现有公开记录不足以支撑买入。
[CV022, CV031, CV032, CV033, CV034]| 论点 | 哪些证据会改变判断 |
|---|---|
| 大市场、强创始人履历和超大种子轮,可能在公开收入出现前先孵出真正的具身 AI 胜者。 | 发布一个具名锚定客户、一张产品地图,以及一套证明材料,显示公司在泛化能力或部署经济性上领先。 |
| 人形机器人私募资本市场仍会在盈利前奖励上行空间,因此 Bulage 即使处在证明构建期,也可能维持价值。 | 证明后续资本只是选项而非结构性必需,或现有资金足以支持公司跨过商业里程碑。 |
| 当前估值偏高,因为公开证明比 Figure、Agility 或 Apptronik 更薄。 | 要么降低入场价格,要么用客户产出、技术材料和重复预算证据提高证明标准。 |
| 公开市场可比公司展示了长期上行空间,但不能验证今天的价格。 | 拿出具备收入质量的牵引力,开始像真实装机基础,而不是研发期权。 |
反向论点不是 Bulage 不能赢,而是当前公开证据还无法判断:现价是否已经过度计入胜利。
[CV004, CV005, CV006, CV007, CV013, CV014]当前判断沿着一条简单链条展开:大市场和强融资都真实存在,但按据报价格买入,证明仍太薄。
[CV001, CV004, CV005, CV006, CV013, CV022]8.2 可比公司组及其含义
最相关的直接估值参照,是私有人形机器人融资轮和交易估值锚点,而不是成熟工业上市公司。Figure 和 Agility 有参考意义,因为它们在野心和市场热度上足够接近,能说明投资者愿意为 AI 叙事、机器人叙事和客户证明的组合支付多少。 Apptronik 有参考意义,因为它显示,一旦真实企业关系可见,战略资本可以进入到什么规模。1X 是有用的低证据可比对象,尽管它的家庭场景角度不同。 这些可比对象说明,私募资本仍愿意在盈利前为巨大上行提供资金。 上市公司组的价值在另一个层面。ABB、Rockwell、FANUC、Teradyne 和 Intuitive 展示了机器人或自动化公司达到持久收入、装机基础、监管披露和投资者报告深度之后,估值可以长成什么样。 因此,这些公司是天花板参照,不是入场价格辩护。若只指向 $40-180B 的上市市值、却不承认这些数字下面压着几十年的证据,就无法为 Bulage 当前估值辩护。[CV005, CV006, CV007, CV008, CV009, CV013]
| 可比对象 | 指标 | 倍数 / 估值 / 状态 | 参考价值 | 局限 |
|---|---|---|---|---|
| Bulage | 传闻种子轮估值 | ~$2.0B 私募估值标记 | 当前入场参考 | 公开证明非常薄 |
| Figure | 私募融资估值标记 | $2.6B 估值,伴随 BMW 和 Helix 叙事 | 最接近的高溢价直接私募可比对象 | 仍是私有公司,财务并不完全透明 |
| Agility | 公开交易估值 | ~$2.5B 合并估值,带有订单和客户基础 | 证明支撑较强的人形机器人可比对象 | SPAC 式结构可能扭曲标题估值 |
| Apptronik | 私募融资背景 | $403M Series A 总额;估值未披露 | 有具名客户、由证明支撑的私募可比对象 | 无公开估值数字 |
| UBTECH | 公开证明背景 | 上市公司;2025 年收入 RMB2.0B,但仍亏损 | 显示商业化并不消除烧钱风险 | 阶段和业务组合不同于 Bulage |
| ABB / Rockwell / FANUC / Intuitive / Teradyne 等上市公司 | 公开市场市值 | 2026 年 7 月区间为 $39B-$180B | 为已验证机器人和自动化业务设定天花板 | 不能直接锚定种子期估值 |
该表用于锚定判断,并不暗示今天的 Bulage 存在单一、完全可比的倍数。
[CV001, CV005, CV006, CV007, CV010, CV013]IC 风格评分卡,聚焦截至 runDate 公开证据能确认的内容。
[CV001, CV002, CV005, CV006, CV007, CV022]8.3 情景分析和预期价值
Bulage 这个阶段,基于里程碑的情景模型比收入倍数模型更站得住脚,因为公开收入仍未知。乐观情形下,Bulage 证明具身大脑论点能跨机器人本体运行,拿下至少一个高质量工业锚点,并开始像一个值得向上重估的中国本土模型层或部署平台。 基准情形下,公司用种子资本建立技术可信度和早期试点,但在商业杠杆清晰前仍需要更多资本。悲观情形下,客户证明仍薄,竞争和热度压缩削弱投资者胃口,下一轮融资按更低实际估值发生。 在这个框架下,当前报道估值仍显得要求很高。最关键的敏感项是具名客户证明、重复预算证据,以及合规或算力依赖不会拖慢商业化的证据。 这些事实一旦出现,支撑区间可以很快上移。但在它们出现前,预期价值逻辑更该服务于保护下行,而不是追逐上行。[CV023, CV024, CV025, CV026, CV027, CV028]
| 情景 | 假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 乐观 | Bulage 证明 OEM 中立的具身大脑切入点可行,拿下锚定工业部署,并发布可信产品栈,同时资本消耗可控。 | 可支持区间约为 $2.4B-$3.2B;上行来自证明质量向 Figure / Agility / Apptronik 靠近,同时保留前沿叙事溢价。 | 执行仍取决于客户转化、算力访问和安全部署。 | 低-中 |
| 基准 | Bulage 将论点转成试点和技术可信度,但仍未规模化,广泛商业化前可能还需要后续资本。 | 可支持区间约为 $1.2B-$1.8B;只有下一轮融资前里程碑进展变得可见,价格才撑得住。 | 证明来得比市场预期慢;资本强度仍然不低。 | 中-高 |
| 悲观 | Bulage 拿不出清晰客户证明,热度压缩,下一轮因谈判筹码变弱而重新定价。 | 可支持区间约为 $0.6B-$1.2B;下行来自估值下调轮或平轮动态,而不只是业务失败。 | 竞争替代、证明滞后和融资市场纪律同时收紧。 | 中 |
由于公开收入仍未知,区间是基于里程碑的启发式判断,不是收入模型输出。
[CV023, CV024, CV025, CV029, CV035, CV036]有支撑的估值最受真实客户证明拉动;单靠泛泛的市场热情,推动要小得多。
[CV023, CV024, CV025, CV026, CV027, CV035]当前据报估值高于基准区间中点,只低于由里程碑驱动的牛市结果。
[CV022, CV023, CV024, CV025, CV035, CV036]8.4 建议和最后尽调问题
证据支持的建议是继续研究。若要给出买入判断,要么入场价格显著更低,要么当前价格下证据显著更强。Bulage 完全可能变得重要;问题是公开尽调还无法判断,公司只是前景诱人,还是已经开始复利增长。 因此,建议质量必须对价格敏感。估值更低时,同一组证据或许足以支持投机性跟踪。按当前报道价格,更干净的立场是:估值偏紧、信心中等、风险高。 要改变这个判断,需要补的工作也很清楚。投资者需要具名客户或伙伴部署、任务级运营指标、产品范围清晰度,以及后续资本究竟是可选项还是必然项的判断。 这些不是细枝末节的缺口,而是世界模型故事与可投投资论证之间缺失的桥。在 Bulage 搭起这座桥之前,最聪明的姿态是有纪律的好奇,而不是确信。[CV031, CV032, CV033, CV034, CV039, CV040]
| 触发项 | 阈值 | 如何冲击论点 | 行动含义 |
|---|---|---|---|
| 没有具名客户证明 | 到下一个主要尽调周期仍没有具名试点或部署 | 当前价格仍只能靠叙事支撑 | 维持继续研究,或大幅下调定价 |
| 产品范围仍不清晰 | 管理层仍说不清卖的是模型、机器人,还是服务很重的部署 | 经济性和竞争护城河无法核保 | 暂停上调任何建议 |
| 后续资本看起来不可避免 | 证明里程碑落地前现金跑道先到头 | 稀释或平轮 / 估值下调轮纳入基准情景 | 向下重切可支持估值区间 |
| 合规 / 算力脆弱性 | 明显依赖难替代的受限工具,或没有治理负责人 | 路线图风险上升,客户尽调负担加重 | 视为乐观情景受损 |
| 试点经济性薄弱 | 任何首批客户若成本过高、定制过重或无法复制 | OEM 中立或软件溢价论点明显削弱 | 转向悲观情景估值框架 |
这些触发项供投委会跟进,必须可监控,而不是泛泛的风险提示。
[CV026, CV027, CV028, CV036, CV037, CV040]| 主题 | 缺失证据 | 重要性 | 负责人或尽调路径 |
|---|---|---|---|
| 具名客户 | 没有公开锚定账户、试点或生产场地 | 当前价格与证据之间最大的缺口 | 要求提供客户推荐、场地清单和限定范围部署结果 |
| 产品范围 | 没有公开说清模型层与全栈机器人策略的取舍 | 毛利、护城河和可比对象选择都无法判断 | 要求提供产品架构、BOM 责任边界和工作流地图 |
| 部署经济性 | 没有公开的任务成功率、在线稳定性或复购预算证据 | 证据不足,估值仍由叙事驱动 | 索取试点指标、支持负担和扩张订单 |
| 资本充足性 | 未披露与里程碑计划匹配的烧钱速度或现金续航 | 后续融资需求会实质改变当前公允价值 | 索取走到下一轮融资或盈亏平衡里程碑的经营计划 |
| 治理与合规 | 隐私、安全和 AI 治理没有公开责任人或路线图 | 监管和客户尽调最终都会要求这些材料 | 索取治理材料包和辖区合规地图 |
上述要求是最低证据集;只有拿到它们,才能在接近报道估值的水平上,从好奇推进到有信心。
[CV002, CV003, CV031, CV034, CV040]8.5 图表
免责声明
本报告是独立尽调摘要,仅基于截至 2026-07-22 已审阅的公开信息。Bulage 是一家公开披露异常有限的私营公司,因此,缺少证据不能误读为事实不存在;同样,也不能用猜测填补。报告不构成投资建议,任何资本决策前都应配合一手尽调。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Reviewed 2026 sources identify the company as Shanghai Bulage Technology Co., Ltd. / 上海卜拉格科技有限公司 rather than using a stable official English legal name page from the company itself. | 中 | SO012, SO015, SO016 |
| CO002 | The assignment brief’s Chinese rendering 布拉格 does not match the Chinese characters used in the fetched public source set, which overwhelmingly uses 卜拉格. | 中 | SO012, SO013, SO014, SO015, SO016, SO017 |
| CO003 | Bulage was reported as established on 2026-05-27. | 中 | SO016 |
| CO004 | Reviewed coverage places Bulage in Shanghai. | 高 | SO003, SO016 |
| CO005 | bulage.cn resolved on 2026-07-22 but returned no readable corporate content through the fetch-url workflow. | 中 | SO001 |
| CO006 | bulage.com was not an official corporate site for this company on 2026-07-22 and instead served unrelated online game content. | 中 | SO002 |
| CO007 | Because the reported founding date is 2026-05-27, Bulage was less than two months old on the run date 2026-07-22. | 中 | SO016 |
| CO008 | Lin Junyang, also rendered as Justin Lin in English-language reporting, is the founder of Bulage and is described as the former core or technical head of Alibaba’s Qwen/Tongyi Qianwen model effort. | 高 | SO003, SO005, SO012, SO015 |
| CO009 | Lin Junyang was born in 1993. | 中 | SO010, SO012, SO019 |
| CO010 | Public biographies say Lin Junyang studied English at the University of International Relations and completed graduate work in linguistics or foreign languages at Peking University. | 中 | SO010, SO012, SO019 |
| CO011 | Lin joined Alibaba’s DAMO Academy in 2019. | 高 | SO005, SO009, SO010, SO019 |
| CO012 | By late 2022 or 2023, Lin had become the formal technical lead or central leader of Alibaba’s Qwen team. | 高 | SO005, SO007, SO009, SO010 |
| CO013 | Public sources report that Lin formed a robotics and embodied-intelligence team inside Qwen in 2025. | 中 | SO008, SO010, SO011 |
| CO014 | Lin publicly announced in early March 2026 that he was stepping down from Qwen. | 高 | SO005, SO006, SO008 |
| CO015 | Several reports tie Lin’s March 2026 departure to disagreements around Alibaba team restructuring rather than to a formally choreographed transition. | 中 | SO011, SO018 |
| CO016 | The fetched public record did not disclose a Bulage cofounder slate, board composition, or independent governance structure. | 中 | SO001, SO012, SO013, SO014, SO015, SO016 |
| CO017 | June 2026 reporting said Lin directly controlled Yuyong (Shanghai) Technology, held 99% of Shanghai Bulage Technology, and used Bulage plus a 1% direct stake to control Shanghai Gewuzhiyong Management Consulting Partnership. | 中 | SO012, SO013, SO014 |
| CO018 | Multiple outlets explicitly interpret “Bulage” as a transliteration of “Pragmatics,” connecting the name choice to Lin’s linguistics background. | 中 | SO012, SO013, SO014, SO017 |
| CO019 | In May 2026, pre-close reporting already valued Lin’s new lab at roughly $2 billion while fundraising was still underway. | 中 | SO004, SO011 |
| CO020 | By mid-June 2026, multiple outlets reported that Bulage had completed its first financing at about a $2 billion post-money valuation. | 高 | SO003, SO012, SO015, SO018, SO019 |
| CO021 | Crunchbase News classified Bulage’s June 2026 financing as a $220 million seed round led by Gaorong Capital and HSG. | 中 | SO003 |
| CO022 | Chinese-language reports described the same financing as a first round or angel round totaling several hundred million dollars rather than always calling it a seed. | 中 | SO015, SO016, SO018, SO019 |
| CO023 | The publicly named checks of $100 million from Gaorong, $100 million from Sequoia/HongShan, and $20 million from Tencent sum to $220 million, matching Crunchbase’s figure while still allowing additional unnamed investors. | 高 | SO003, SO015, SO018, SO019, SO020 |
| CO024 | Fetched coverage uses HSG, HongShan, and Sequoia China interchangeably for the same co-lead investor franchise in Bulage’s round. | 中 | SO003, SO015, SO021 |
| CO025 | Tencent was repeatedly described as a $20 million participant in Bulage’s first financing rather than as a co-lead. | 中 | SO015, SO018, SO019 |
| CO026 | Several outlets said Bulage began exploring another financing round immediately after closing the first one. | 中 | SO004, SO012, SO013 |
| CO027 | Reviewed sources describe Bulage’s technical focus as world models and embodied brains or embodied intelligence. | 中 | SO011, SO012, SO015, SO016, SO017 |
| CO028 | Bulage’s stated goal is to push AI from virtual interaction toward physical-world perception, understanding, and action. | 中 | SO016, SO019 |
| CO029 | Public reporting emphasized that Bulage had no public product catalog, official website content, or disclosed revenue at the time of its debut financing. | 中 | SO001, SO004 |
| CO030 | Several reports characterized a roughly $2 billion valuation for a newly formed Chinese AI lab with no disclosed products or revenue as unprecedented or nearly unprecedented. | 中 | SO004, SO013, SO014 |
| CO031 | The available source set implies investors underwrote Bulage primarily on founder track record and frontier technical direction rather than on disclosed commercial traction. | 中 | SO004, SO011, SO013, SO023 |
| CO032 | Crunchbase placed Bulage among the new June 2026 unicorn cohort even though the company was less than one year old. | 中 | SO003, SO007 |
| CO033 | Yicai reported that China’s embodied-AI sector drew CNY93.5 billion (USD13.8 billion) across 322 deals in H1 2026. | 中 | SO023 |
| CO034 | ChinaBiz Insider reported that at least 25 Chinese embodied-intelligence startups were already valued above RMB10 billion and that many carried only 18 to 24 months of cash runway. | 中 | SO024 |
| CO035 | Unite.AI reported that China’s top economic planning agency had warned in late 2025 that more than 150 humanoid-robot companies were flooding the market and creating bubble risk. | 中 | SO025 |
| CO036 | As of the run date, the fetched public record did not disclose Bulage’s revenue, ARR, customer count, or employee count. | 中 | SO001, SO003, SO012, SO016 |
| CO037 | Bulage’s external diligence surface is unusually fragile because its .cn domain lacks readable content and the matching .com domain is controlled by an unrelated operator. | 高 | SO001, SO002 |
| CO038 | The co-lead investors named in Bulage’s round are established venture franchises: HSG describes itself as a long-running global venture and private-equity firm, and Gaorong’s homepage presents it as one of China’s active venture investors founded in 2014. | 中 | SO021, SO022 |
| CO039 | Reporting around Bulage repeatedly frames the startup thesis as a move from reasoning-oriented models toward agentic systems acting in the physical world. | 中 | SO011, SO013, SO014, SO017 |
| CO040 | The path from Qwen departure in early March 2026 to completed first financing by mid-June 2026 shows that Bulage went from founder breakaway to unicorn-scale financing in roughly one quarter. | 中 | SO005, SO012, SO015, SO016 |
| CM001 | Bulage's publicly described thesis places it in embodied intelligence and world models rather than generic SaaS AI or conventional robotics. | 高 | SM024, SM025, SM026 |
| CM002 | The most relevant included spend for Bulage covers embodied models, corpora, simulation and training systems, and downstream general-purpose embodied robots used in pilots or deployment. | 中 | SM002, SM017, SM020 |
| CM003 | The relevant market should exclude fixed industrial robots, warehouse-only automation, autonomous driving, surgical robots, and consumer robot appliances that do not depend on embodied-AI control stacks. | 中 | SM002, SM004, SM017 |
| CM004 | Status-quo substitutes for embodied-intelligence systems remain human labor, conventional task-specific automation, and internal robotics programs at large enterprises. | 中 | SM004, SM006, SM023 |
| CM005 | QbitAI states that Bulage's创业方向聚焦世界模型和具身大脑, directly linking the company to the embodied-AI market. | 中 | SM024 |
| CM006 | China had 2,027,000 industrial robots working in factories in 2024 and represented 54% of global industrial-robot demand. | 中 | SM010 |
| CM007 | Chinese robot suppliers reached 57% share of the domestic market in 2024, overtaking foreign manufacturers at home. | 高 | SM010, SM004 |
| CM008 | TrendForce forecasts China humanoid output will grow 94% in 2026. | 中 | SM001 |
| CM009 | TrendForce says the second half of 2026 is the key commercialization period for the global humanoid industry. | 中 | SM001 |
| CM010 | Morgan Stanley forecasts 50,000 China humanoid shipments in 2026. | 中 | SM007 |
| CM011 | Morgan Stanley forecasts China's humanoid robot market will reach about $2 billion in 2026. | 中 | SM007 |
| CM012 | Morgan Stanley forecasts China's humanoid robot market will reach $15 billion and 446,000 annual shipments by 2030. | 中 | SM007 |
| CM013 | China Economic Net cites IDC estimates that user spending on embodied intelligent robots in China exceeded $1.4 billion in 2025 and could reach $77 billion by 2030. | 中 | SM003 |
| CM014 | China Economic Net cites a Development Research Center estimate that China's embodied-intelligence industry could reach RMB 400 billion in 2030 and exceed RMB 1 trillion in 2035. | 中 | SM003 |
| CM015 | Omdia-derived reporting puts global general-purpose embodied intelligent robot shipments at 13,318 units in 2025. | 高 | SM011, SM012 |
| CM016 | Omdia-derived reporting projects global annual shipments of general-purpose embodied intelligent robots will reach 2.6 million units by 2035. | 高 | SM011, SM012 |
| CM017 | Interact Analysis defines eight application segments for humanoid robots including academic R&D, robot training, entertainment, manufacturing, warehouse, public service, and household use. | 中 | SM002 |
| CM018 | KrASIA reports that the main buyers used to be academic research labs. | 中 | SM006 |
| CM019 | KrASIA reports that a newer customer profile is state-owned enterprises placing robots in lobbies for display. | 中 | SM006 |
| CM020 | Analyst and company sources show embodied robots are now being trialed across manufacturing, warehouse, public-service, hospitality, education, and security scenarios rather than only research demos. | 中 | SM002, SM012, SM020 |
| CM021 | Shanghai's embodied-intelligence plan explicitly prioritizes logistics and assembly, industrial manufacturing, commercial retail, healthcare and rehabilitation, and household services. | 高 | SM008, SM017 |
| CM022 | Shanghai's implementation plan offers support of up to 30% of approved project investment for key R&D efforts, capped at RMB 50 million per project. | 中 | SM017 |
| CM023 | Shanghai's implementation plan creates five public platforms for computing power, digital-twin training, pilot testing, investment, and leasing. | 高 | SM008, SM017 |
| CM024 | The likely near-term payer for embodied-AI deployments is an OEM, factory owner, SOE, municipal platform, or large enterprise rather than a household consumer. | 中 | SM006, SM017, SM020 |
| CM025 | The operational buyer is usually a lab head, automation lead, operations executive, or innovation office rather than an individual end user. | 中 | SM002, SM006, SM020 |
| CM026 | The Hangzhou national pilot base operates more than 130 robots across over 30 vocational scenarios, providing evidence of real training and pilot infrastructure. | 中 | SM020 |
| CM027 | Shanghai targets 100 leading enterprises, 100 application scenarios, 100 competitive products, and more than RMB 50 billion of core industry output by 2027. | 高 | SM008, SM017 |
| CM028 | SCIO reports Shanghai has set a target of deploying 100,000 humanoid robots in factories by 2030. | 中 | SM018 |
| CM029 | SCIO reports embodied-intelligence sector revenue grew 22.4% year on year from January to May 2026 and industrial purchases more than tripled. | 中 | SM018 |
| CM030 | Official and media sources state that China had over 140 humanoid robot manufacturers and more than 330 models released over the prior year. | 高 | SM003, SM019, SM016 |
| CM031 | China's first national humanoid-robot standard system was built with participation from more than 120 institutions and enterprises. | 中 | SM019 |
| CM032 | The standards push is explicitly designed to reduce coordination costs, accelerate modularization, and move humanoids from demonstration to larger-scale deployment. | 中 | SM019 |
| CM033 | National and local policy support for embodied AI has been visible since at least 2023 and was elevated through the 2025 government work report and 2026 planning cycle. | 高 | SM004, SM005, SM018 |
| CM034 | MERICS links China's embodied-robotics push to labor shortages, rising labor costs, manufacturing-chain depth, and state policy. | 高 | SM004, SM022, SM023 |
| CM035 | MERICS finds China's humanoids still lack precision and dexterity and are mostly deployed in limited tasks and site-specific trials. | 中 | SM004 |
| CM036 | MERICS reports current industrial humanoids in China average roughly CNY 300,000 to CNY 500,000 each, while commercial viability may require a threshold around CNY 160,000. | 中 | SM004 |
| CM037 | MERICS says China remains dependent on Nvidia's AI stack and on foreign firms for several high-end components even as hardware localization advances. | 中 | SM004 |
| CM038 | KrASIA says most institutional investors and academic researchers expect the sector to reach maturity over the next five to ten years. | 中 | SM006 |
| CM039 | KrASIA reports that some automotive-factory humanoid deployments are viewed by insiders as strategic tie-ups or demos rather than true sales. | 中 | SM006 |
| CM040 | Embodied Global identifies scarce real-world data and poor cross-factory generalization as the hardest bottlenecks for embodied-AI progress. | 中 | SM015 |
| CM041 | Unite.AI reports that China's NDRC warned about bubble risk in a market with more than 150 companies and immature business models and application scenarios. | 中 | SM016 |
| CM042 | Yicai reports China's embodied-AI sector drew CNY 93.5 billion of financing across 322 deals in H1 2026, underscoring how quickly capital formation is outrunning proof of revenue. | 中 | SM013, SM014 |
| CP001 | Bulage's relevant competitor set includes direct humanoid OEMs, embodied-model labs, industrial incumbents, and status-quo substitutes rather than one simple peer bucket. | 高 | SP020, SP021, SP022, SP024, SP025 |
| CP002 | Figure positions Figure 03 as a general-purpose home humanoid and explicitly ties it to Helix for operation in unpredictable home environments. | 高 | SP001, SP002 |
| CP003 | Helix is described by Figure as an onboard VLA that controls the full upper body at high rate and generalizes to thousands of unseen household objects from natural language prompts. | 中 | SP002 |
| CP004 | Figure raised $675 million at a $2.6 billion valuation and said the capital would be used for AI training, robot manufacturing, headcount, and commercial deployment. | 中 | SP003 |
| CP005 | Figure has a publicly disclosed, milestone-based commercial agreement to deploy humanoids with BMW Manufacturing in Spartanburg. | 高 | SP004, SP030 |
| CP006 | 1X markets NEO as a home robot, publishes a $200 deposit signal, and says unknown chores can be completed through a 1X Expert-guided path while the robot learns. | 高 | SP005, SP007 |
| CP007 | 1X said it raised a $100 million Series B and more than $125 million in less than 12 months to bring NEO to market and support enterprise clients. | 中 | SP006 |
| CP008 | Agility sells more than a robot body: it pairs Digit with the Arc workflow platform and ongoing service and support. | 中 | SP008 |
| CP009 | Amazon began testing Digit for tote recycling and Agility said RoboFab would start at hundreds of units per year with a path to more than 10,000 annually. | 中 | SP009 |
| CP010 | Agility publicly claims active commercial deployments with major enterprises and expects more than $620 million of gross proceeds from its announced Churchill transaction. | 高 | SP008, SP010, SP011 |
| CP011 | Schaeffler publicly said it sees potential to deploy a significant number of humanoids across its global network of 100 plants by 2030. | 高 | SP010, SP008 |
| CP012 | Boston Dynamics' productized Atlas is industrial-first, with 56 degrees of freedom, a 110-pound lift capacity, four-hour battery life, and all 2026 units reportedly already committed. | 中 | SP012 |
| CP013 | Boston Dynamics' existing revenue base and 500-plus prior robot deployments give it a service and reliability reference point that younger humanoid labs lack. | 中 | SP012 |
| CP014 | Boston Dynamics explicitly argues that home deployment is the wrong first strategy today because costs, capabilities, and safety standards are still insufficient. | 中 | SP012 |
| CP015 | Unitree openly lists G1 pricing from $13,500 and describes a roughly 35kg humanoid with 23-43 degrees of freedom and light arm payload. | 中 | SP013 |
| CP016 | Unitree's H1 and H1-2 pages show a more capable full-size platform with 3.3m/s mobility, 47-70kg weight classes, and expanded degrees of freedom. | 中 | SP014 |
| CP017 | Unitree says it vertically integrates key robot components and algorithms and has filed more than 200 patents, over 180 of them authorized. | 中 | SP015 |
| CP018 | TrendForce reported that Unitree's prospectus showed humanoid revenue exceeded quadruped revenue in 2025, combined gross margin reached 60%, and capacity is being expanded toward 75,000 humanoids annually. | 中 | SP020 |
| CP019 | UBTech positions Walker S as an industrial humanoid with 41 servo joints, multimodal LLM-based decision making, and direct factory-system integration. | 中 | SP016 |
| CP020 | UBTech says Walker S2 can swap batteries autonomously within three minutes, handle 15kg payloads, and has entered mass production and delivery. | 中 | SP017 |
| CP021 | AGIBOT publicly combines a multi-robot product portfolio, the AGIBOT World dataset ecosystem, and commercial mass-production messaging in its market story. | 中 | SP018 |
| CP022 | AGIBOT cited Omdia as ranking it No.1 globally in 2025 shipments and market share, with 5,168 humanoids shipped and 39% share. | 中 | SP019 |
| CP023 | TrendForce, CNBC, and AGIBOT's Omdia release collectively support the view that China leads current humanoid shipment momentum, with Unitree and AGIBOT the clearest scale leaders. | 高 | SP019, SP020, SP021 |
| CP024 | Morgan Stanley's research cited by CNBC said Chinese companies held the top five shipment positions in 2025 while Figure ranked seventh. | 中 | SP021 |
| CP025 | China's national standard system both validates sector seriousness and highlights fragmentation, with 140-plus domestic manufacturers, 330-plus models, and ongoing concerns around cost, generalization, and imported components. | 中 | SP023 |
| CP026 | MERICS argues that China humanoids are still mostly in limited, site-specific trials and that commercial viability will likely require costs to fall by at least half. | 中 | SP022 |
| CP027 | China has a hardware cost and localization advantage in humanoids, but still depends meaningfully on Nvidia software and foreign high-end components. | 中 | SP022 |
| CP028 | KrASIA's adverse reporting says many investors still view humanoids as a five-to-ten-year maturity story and suspect that many factory tie-ups are strategic signals more than proven revenue. | 中 | SP024 |
| CP029 | Yicai reported that embodied-AI financing in China rose to CNY93.5 billion across 322 deals in H1 2026, even as industry sources warned that mass deployment remains some way off. | 中 | SP025 |
| CP030 | Adverse evidence from Unite.ai and SCIO supports the view that the category faces bubble risk, with more than 150 companies, many homogeneous products, and immature business models. | 高 | SP023, SP026 |
| CP031 | Unite.ai reported that UBTech had secured more than RMB800 million of Walker orders, shipped several hundred units, and planned higher annual capacity by 2026-2027. | 中 | SP026 |
| CP032 | Public price discovery remains highly uneven across the field: Unitree posts a low headline price and 1X posts a deposit, while most enterprise vendors still sell through opaque custom agreements or RaaS-style packaging. | 高 | SP005, SP008, SP010, SP012, SP013, SP016 |
| CP033 | The field separates into three main go-to-market archetypes: home-first labs, industrial-first deployers, and China scale-first manufacturers. | 中 | SP001, SP007, SP008, SP012, SP018, SP020 |
| CP034 | Bulage has public funding scale and a world-model narrative, but no public product, pricing, or deployment documentation on its official surface as of 2026-07-22. | 高 | SP027, SP028, SP029 |
| CP035 | Bulage's currently visible differentiation is thesis-level — world models and embodied brains — rather than market-validated customer or shipment proof. | 高 | SP028, SP029, SP027 |
| CP036 | Industrial-first commercialization appears more mature than home-first commercialization because the public evidence is concentrated in factories, warehouses, and narrow enterprise workflows rather than scaled household use. | 高 | SP007, SP008, SP010, SP012, SP022 |
| CP037 | China's policy support, standards push, and supply-chain depth are likely to compress hardware price umbrellas and make it harder for capital alone to function as a durable moat. | 高 | SP020, SP022, SP023, SP025, SP026 |
| CP038 | Service, maintenance, workflow integration, and safety proof are meaningful competitive moats for Agility and Boston Dynamics that Bulage has not yet publicly demonstrated. | 高 | SP008, SP010, SP012, SP027 |
| CP039 | Even the strongest public model-first players still rely on curated data collection, teleoperation, or expert-guided learning paths, suggesting embodied intelligence remains bottlenecked by real-world data acquisition. | 高 | SP002, SP005, SP007, SP024 |
| CP040 | The competitive verdict for Bulage is not that it lacks promise, but that its moat is unproven versus peers already showing price anchors, named deployments, service layers, or shipment scale. | 高 | SP020, SP021, SP027, SP028, SP029 |
| CI001 | Bulage discloses no public revenue, price, customer, gross-margin, or headcount data on its official surface as of 2026-07-22. | 高 | SI001, SI002, SI003 |
| CI002 | Public sources do not identify whether Bulage plans to monetize first through robot sales, software/model licensing, integration services, or a bundled deployment model. | 高 | SI001, SI002, SI003 |
| CI003 | Public humanoid monetization evidence is typically bundled across hardware, software, integration, maintenance, and support rather than presented as a simple SaaS price list. | 高 | SI009, SI015, SI018, SI020 |
| CI004 | Figure said its $675 million Series B would fund AI training, robot manufacturing, engineering headcount, and commercial deployment. | 中 | SI004 |
| CI005 | 1X said its $100 million Series B would bring NEO to market and support enterprise clients in logistics and guarding. | 中 | SI005 |
| CI006 | Agility's announced public transaction is expected to provide more than $620 million of gross proceeds for orders, deployments, production scale-up, and platform investment. | 高 | SI008, SI010 |
| CI007 | Agility publicly disclosed more than $300 million of multi-year Digit v5 orders and a pipeline of over 30 customers. | 高 | SI007, SI010 |
| CI008 | Unitree's G1 public sticker price of $13,500 is the clearest low-end humanoid price anchor in the retained source set. | 中 | SI011 |
| CI009 | Research and Markets' 2026 China report describes low-tier humanoids at RMB100,000-200,000 and mid-tier humanoids at RMB250,000-500,000, with broader market value rising further once software, integration, and maintenance are included. | 中 | SI015 |
| CI010 | The public market framing implies that realized monetization often depends on bundled solution value, not just the robot ASP. | 高 | SI009, SI015, SI020 |
| CI011 | UBTECH reported FY2025 revenue of RMB2,001.0 million, up 53.3% year over year. | 高 | SI020, SI021, SI022 |
| CI012 | UBTECH's full-size embodied intelligent humanoid products and services generated RMB820.6 million in FY2025, becoming its largest revenue segment. | 高 | SI020, SI021, SI022 |
| CI013 | UBTECH reported FY2025 gross margin of 37.7%, gross profit of RMB753.8 million, and net loss of RMB789.8 million. | 高 | SI020, SI021, SI022 |
| CI014 | UBTECH's FY2025 R&D expenditure reached RMB507.5 million and cash plus equivalents reached RMB4,887.9 million. | 高 | SI020, SI022, SI027 |
| CI015 | UBTECH sold 1,079 full-size humanoids in 2025, reached annualized capacity above 6,000 units, and pushed Walker S2 into 1,000-unit-level small-batch mass production. | 高 | SI020, SI021, SI022 |
| CI016 | UBTECH says its Walker series has moved from field trials into commercial applications across automotive, electronics, semiconductor manufacturing, and logistics. | 高 | SI018, SI020, SI021 |
| CI017 | MERICS says China humanoids still average roughly RMB300,000-500,000 each and cites a commercial-viability threshold around RMB160,000 for a two-year ROI case. | 中 | SI013 |
| CI018 | Even well-funded peers like Figure, 1X, Agility, and AGIBOT still do not publicly disclose full realized ASPs or margin structures for their flagship humanoid offerings. | 高 | SI004, SI005, SI007, SI023 |
| CI019 | Public humanoid sales motions appear consultative and milestone-based, especially for industrial deployments, rather than low-friction recurring subscriptions. | 高 | SI007, SI008, SI009, SI010 |
| CI020 | Consumer-facing commercialization in China is still early and partly awareness-driven, with stores and deposits used to cultivate market understanding rather than prove immediate scaled demand. | 中 | SI016, SI017 |
| CI021 | Bulage should be treated as pre-revenue or financially unproven in public diligence because no revenue, pricing, or customer data is disclosed. | 高 | SI001, SI002, SI003 |
| CI022 | Financing dependency is category-wide: Figure raised $675 million, Agility is pursuing over $620 million of proceeds, 1X raised $100 million, and UBTECH still reported losses despite meaningful revenue. | 高 | SI004, SI005, SI008, SI013, SI020 |
| CI023 | Bulage's $220 million seed is large enough to fund research and pilots, but may still be insufficient for a fully scaled hardware-plus-model manufacturing path. | 高 | SI002, SI004, SI008, SI020 |
| CI024 | The most important public tests of revenue quality in this category are repeat orders, service attach, task reliability, and deployment utilization rather than one-time valuation headlines. | 高 | SI007, SI010, SI015, SI025 |
| CI025 | Humanoid cost structure likely includes expensive components, compute and data collection, safety engineering, field support, and manufacturing setup before revenue quality is stable. | 高 | SI009, SI013, SI014, SI020 |
| CI026 | Public evidence suggests early real revenue is more likely to come from industrial deployments than from broad household adoption. | 高 | SI007, SI010, SI016, SI021 |
| CI027 | Bulage's next financing decision should depend on first deployment and packaging proof because public evidence does not yet support a financial underwrite. | 高 | SI001, SI007, SI010, SI025 |
| CI028 | Public financial gaps at Bulage include revenue, gross margin, burn, cash, contract structure, customer concentration, manufacturing capex, and support burden. | 高 | SI001, SI002, SI003 |
| CI029 | Yicai's broader financing coverage shows embodied-AI capital has surged ahead of mass commercial deployment in China. | 中 | SI026 |
| CI030 | KrASIA's adverse reporting argues valuations across humanoids are ballooning faster than revenue and product-market fit. | 中 | SI025 |
| CI031 | Offline stores, deposits, and showcase retail are useful demand signals but not substitutes for sustained order flow or margin disclosure. | 高 | SI006, SI017 |
| CI032 | Bulage cannot currently be underwritten on CAC, payback, or gross margin using only public evidence. | 高 | SI001, SI002, SI003 |
| CI033 | Standardization and modularization could improve unit economics over time by reducing coordination and adaptation costs across the supply chain. | 中 | SI014, SI016 |
| CI034 | The category's public revenue architecture spans hardware, integration, maintenance, support, and sometimes software or data layers, which means Bulage's eventual margin profile depends heavily on packaging choice. | 高 | SI009, SI015, SI020 |
| CI035 | The defensible public verdict is that Bulage's revenue quality is unproven, its margin path is unknown, and its financing dependency is likely still high. | 高 | SI001, SI021, SI022, SI025, SI026 |
| CI036 | UBTECH's materially stronger cash balance was achieved only after repeated share placements, underscoring how dependent even scaled humanoid companies remain on external capital markets. | 高 | SI022, SI027 |
| CI037 | Large TAM and commercialization headlines do not substitute for company-specific evidence on Bulage's contract model, unit economics, or repeat demand. | 高 | SI015, SI024, SI026 |
| CI038 | Bulage's $2 billion valuation therefore reads primarily as option value on founder quality and category timing, not as a valuation anchored in disclosed operating metrics. | 高 | SI002, SI003, SI025, SI026 |
| CE001 | Bulage's public product definition is still limited to a world-model / embodied-brain thesis rather than a disclosed module, SKU, or product stack. | 高 | SE001, SE002, SE003 |
| CE002 | Bulage's official site provides no readable product docs, trust pages, roadmap, or technical surface. | 中 | SE001 |
| CE003 | The public frontier stack for embodied AI now visibly spans data, simulation, reasoning, action policies, robot hardware, and safety infrastructure. | 高 | SE005, SE006, SE008, SE009, SE010, SE011 |
| CE004 | Figure's Helix describes a two-system architecture that couples slower semantic reasoning with faster visuomotor control and runs onboard low-power GPUs. | 中 | SE005 |
| CE005 | Gemini Robotics 1.5 publicly pairs an embodied reasoning model with a VLA model and explicitly targets complex multi-step tasks plus cross-embodiment learning. | 高 | SE006, SE007 |
| CE006 | π0 combines internet-scale VLM pretraining with robot data from eight distinct robots and uses a flow-matching action architecture for dexterous control. | 中 | SE011 |
| CE007 | NVIDIA GR00T packages open data pipelines, an open robot foundation model, simulation frameworks, middleware, runtime libraries, and deployment hardware into one reference platform. | 中 | SE008 |
| CE008 | Isaac Lab is a GPU-accelerated simulation and training framework that underpins the GR00T workflow and supports imitation plus reinforcement learning at scale. | 中 | SE009 |
| CE009 | NVIDIA Halos frames safe physical AI as a full-stack system spanning hardware, OS, inspection, standards alignment, and outside-in monitoring. | 中 | SE010 |
| CE010 | UBTECH's open-source Thinker project claims future-state prediction, spatial intelligence, temporal understanding, and visual grounding as core embodied-VLM capabilities. | 中 | SE012 |
| CE011 | UBTECH's annual results describe BrainNet 2.0, Co-Agent, humanoid brain/cerebellum language, and multi-robot industrial collaboration as part of its full-stack architecture. | 高 | SE016, SE017 |
| CE012 | AGIBOT WORLD 2026 publicly describes heterogeneous real-world data, free-form teleoperation, digital twins, whole-body control, multimodal sensing, and phased releases. | 高 | SE013, SE014 |
| CE013 | Unitree's public technology surface emphasizes hardware performance and vertical integration more than detailed public model-stack disclosure. | 高 | SE018, SE019 |
| CE014 | The most publicized customer workflows today are still narrow automation, data collection, or structured deployment tasks rather than fully general consumer autonomy. | 高 | SE021, SE023, SE024, SE025 |
| CE015 | Simulation and digital twins are now central technical dependencies for scaling embodied learning and pre-deployment validation. | 高 | SE009, SE013 |
| CE016 | Cross-embodiment transfer is a core public ambition across Gemini Robotics, π0, and GR00T-like platform narratives. | 高 | SE006, SE007, SE008, SE011 |
| CE017 | Low-power onboard or deployable inference is emerging as an explicit design requirement rather than an afterthought. | 高 | SE005, SE008, SE010 |
| CE018 | Trust and safety controls in public frontier systems now include semantic-safety reasoning, functional-safety operating systems, inspection labs, and standards mapping. | 高 | SE006, SE010, SE020 |
| CE019 | Bulage discloses no public trust, safety, privacy, security, or quality controls of its own. | 高 | SE001, SE020 |
| CE020 | No public source identifies Bulage's robot body, actuator stack, inference hardware, deployment tooling, or support model. | 高 | SE001, SE002, SE003 |
| CE021 | Bulage has no public developer surface comparable to UBTECH Thinker or AGIBOT WORLD. | 高 | SE001, SE012, SE013 |
| CE022 | Chinese peers are now open-sourcing datasets and models, which can accelerate ecosystem learning and weaken any moat based only on secrecy. | 高 | SE012, SE013 |
| CE023 | Public frontier stacks still depend heavily on curated data collection, adaptation, and validation despite advances in foundation models. | 高 | SE006, SE009, SE011, SE013 |
| CE024 | The peer roadmap shows rapid 2025-2026 movement from demos toward structured industrial and partner-gated deployment, but not a fully mature end-state. | 高 | SE006, SE010, SE017, SE025 |
| CE025 | If Bulage can generalize across robot bodies, an OEM-neutral embodied-brain layer is still a plausible technical wedge. | 高 | SE005, SE006, SE008, SE011 |
| CE026 | That wedge would still require credible answers on data, simulation, control, hardware abstraction, and safety integrations, none of which Bulage has disclosed publicly. | 高 | SE008, SE009, SE010, SE013, SE001 |
| CE027 | Bulage is at the earliest public maturity tier of the reference set because it lacks disclosed product, dataset, developer, and safety surfaces. | 高 | SE001, SE005, SE010, SE012, SE013 |
| CE028 | China's policy and research discourse treats embodied AI as strategically important but still technically incomplete and dependent on key external technologies. | 高 | SE021, SE022 |
| CE029 | Standardization can reduce integration and adaptation cost over time, but it also raises the bar for explicit safety and interoperability controls. | 高 | SE020, SE021 |
| CE030 | The defensible product-tech verdict is that Bulage is directionally aligned with frontier embodied-AI themes but publicly unverified on architecture, integration, and support. | 高 | SE001, SE002, SE003, SE021 |
| CE031 | No public source confirms Bulage's deployment, reliability, or support readiness. | 高 | SE001, SE024 |
| CE032 | Public product pages and filings from UBTECH show that industrial embodied systems are being wrapped in full-stack operating, coordination, and uptime narratives rather than single-model claims alone. | 高 | SE015, SE016, SE017 |
| CE033 | AGIBOT and UBTECH both connect technical assets to wider scenario portfolios, underscoring how much public workflow evidence Bulage still lacks. | 高 | SE014, SE017, SE025 |
| CE034 | The transparency gap itself is material: peers expose inspectable tech artifacts while Bulage mostly exposes valuation coverage and strategic slogans. | 高 | SE001, SE005, SE010, SE012, SE013 |
| CE035 | Bulage's missing trust and quality surface is not a cosmetic issue because humanoid products operate in physical environments where safety and support failures are product failures. | 高 | SE010, SE020, SE021, SE001 |
| CU001 | Bulage's official site provides no customer page, case study archive, product documentation, or deployment reference surface. | 中 | SU001 |
| CU002 | Public Bulage coverage around the June 2026 financing highlights founder pedigree, world models, and valuation rather than customers, pilots, or deployments. | 中 | SU002, SU003 |
| CU003 | As of runDate, no reviewed public source names a Bulage customer, partner deployment site, or paying design partner. | 高 | SU001, SU002, SU003 |
| CU004 | No reviewed public source distinguishes whether Bulage is in proof-of-concept, pilot, or production deployment with any account. | 高 | SU001, SU002, SU003 |
| CU005 | No reviewed public source discloses Bulage account counts, robot counts, locations, utilization, or repeat-order metrics. | 高 | SU001, SU002, SU003 |
| CU006 | Named customer proof in humanoids usually appears through deployment releases, customer-quoted announcements, or official industrial case-study pages rather than through generic branding. | 高 | SU004, SU005, SU007, SU009, SU012 |
| CU007 | Agility says Digit entered commercial operations at a GXO facility near Atlanta in June 2024. | 中 | SU004 |
| CU008 | GXO says its multi-year Agility agreement followed a late-2023 proof-of-concept pilot. | 中 | SU005 |
| CU009 | The GXO deployment includes Digit robots, Agility Arc fleet software, and workflow integration with existing warehouse automation. | 中 | SU005 |
| CU010 | Agility frames the GXO installation as revenue-generating, customer-site work rather than a lab or showroom demo. | 高 | SU004, SU005 |
| CU011 | The Robot Report says Atlas was doing real work at Hyundai's Georgia plant and that this was the first time Atlas had been out of the laboratory doing real work. | 中 | SU006 |
| CU012 | Hyundai says Atlas is planned for HMGMA by 2028 for sequencing tasks, with later expansion to assembly and heavier repetitive work. | 中 | SU007 |
| CU013 | Hyundai's AI robotics strategy frames robot deployment as a phased rollout validated process by process rather than an instant plant-wide launch. | 中 | SU007 |
| CU014 | BMW says a Figure pilot at Spartanburg supported production of more than 30,000 BMW X3, moved more than 90,000 components, and logged roughly 1,250 operating hours. | 中 | SU009 |
| CU015 | BMW says the Spartanburg pilot produced enough learning for the company and Figure to evaluate additional Figure 03 use cases. | 中 | SU009 |
| CU016 | Figure's first BMW agreement was milestone-based, which means even strong proof still came with staged gates rather than blanket production commitment. | 中 | SU008 |
| CU017 | UBTECH's industrial case page names BYD, NIO, Foxconn, Zeekr, Dongfeng Liuzhou Motor, and FAW-Volkswagen Qingdao as application cases for its Walker series. | 中 | SU012 |
| CU018 | UBTECH's own Walker S1 product page says the robot has already been introduced into vehicle manufacturing assembly lines. | 中 | SU013 |
| CU019 | UBTECH's Walker S2 page positions the robot for 24/7 industrial operation, implying a buyer set that values uptime and service continuity rather than one-off demos. | 中 | SU014 |
| CU020 | AGIBOT's global store creates a direct commercial surface with listed products and visible sticker prices for some non-humanoid and peripheral hardware. | 中 | SU015 |
| CU021 | AGIBOT says 2026 is "Deployment Year One" and claims seven standardized productivity solutions across manufacturing, logistics, retail, security, and cleaning. | 中 | SU016 |
| CU022 | AGIBOT also says it had rolled out its 10,000th robot by March 2026, signaling a much wider commercial surface than Bulage currently discloses. | 中 | SU016 |
| CU023 | KEENON's WAIC 2025 demonstration focused on service scenarios such as healthcare assistance, lounge service, and branded beverage service rather than factory labor. | 中 | SU017 |
| CU024 | KEENON's official commercial contact flow explicitly spans restaurant, hotel, medical, and industrial-transport categories, showing buyer segmentation more clearly than Bulage does. | 中 | SU018 |
| CU025 | 1X and Unitree provide public buyer-facing surfaces — early-access home positioning in 1X's case and public G1 hardware pricing in Unitree's case — that make customer targeting more legible. | 高 | SU010, SU011 |
| CU026 | TrendForce says the humanoid market is expected to enter a commercialization phase in H2 2026, with Unitree and AGIBOT leading China's market. | 中 | SU019 |
| CU027 | Research and Markets' 2026 China report treats software, system integration, and maintenance as part of the humanoid market opportunity, implying that enterprise buyers care about full-stack deployment not just robot ASP. | 中 | SU020 |
| CU028 | UBTECH's 2025 annual results provide public evidence that humanoid commercialization in China is moving into real industrial accounts rather than staying at pure concept stage. | 高 | SU021, SU012 |
| CU029 | KrASIA's 2026 reality-check article argues that China's humanoid race still faces a gap between technical spectacle and proven commercial value. | 中 | SU022 |
| CU030 | Bulage's lack of customer proof is especially notable because peers now disclose pilots, deployments, storefronts, or segmented target buyers on their public surfaces. | 高 | SU001, SU004, SU007, SU009, SU012, SU015, SU018 |
| CU031 | Agility's investor materials say the company has more than 30 customer relationships and over $300 million in multi-year Digit v5 orders. | 中 | SU023 |
| CU032 | BMW's press material shows that credible humanoid customer proof can include concrete outputs like units touched, operating hours, and supported production volume. | 中 | SU009 |
| CU033 | The BMW Spartanburg plant page confirms the site is a real operating automotive facility, helping ground the Figure case in a specific production environment. | 中 | SU024 |
| CU034 | Unitree's about page presents a scale manufacturing posture that is more compatible with direct or channel hardware sales than Bulage's currently opaque surface. | 高 | SU011, SU025 |
| CU035 | AGIBOT's Omdia-linked shipment claim reinforces that at least some China players now pair deployment rhetoric with volume narratives, even if full customer economics remain undisclosed. | 高 | SU016, SU026 |
| CU036 | No public retention metric such as NRR, GRR, churn, renewal rate, or contract duration was found for Bulage. | 高 | SU001, SU002, SU003 |
| CU037 | Even among better-documented peers, public repeat-order and renewal data remain thinner than initial deployment announcements. | 高 | SU004, SU005, SU009, SU021, SU022 |
| CU038 | The best public production-grade customer proofs in the reviewed set are Agility/GXO, Figure/BMW, and UBTECH's named automotive accounts. | 高 | SU004, SU005, SU009, SU012, SU013 |
| CU039 | Service and consumer-facing proof in the reviewed set remains earlier-stage and more demo- or channel-oriented than the strongest industrial proofs. | 高 | SU017, SU018, SU010, SU015 |
| CU040 | For Bulage, customer underwriting still depends more on future diligence asks than on current public adoption evidence. | 高 | SU001, SU002, SU022 |
| CR001 | Bulage's official site provides no public privacy, trust, compliance, incident, or customer-safety surface. | 中 | SR001 |
| CR002 | Public Bulage coverage emphasizes founder pedigree, world models, and valuation rather than operating controls, customer proof, or compliance readiness. | 中 | SR002, SR003 |
| CR003 | No reviewed public Bulage source discloses a formal compliance roadmap, customer-safety program, or legal owner for data, privacy, or AI governance. | 高 | SR001, SR002, SR003 |
| CR004 | China's interim generative-AI measures require lawful data and foundation-model sources, respect for intellectual property, and lawful handling of personal information. | 高 | SR028, SR029 |
| CR005 | The same China rules require providers to offer safe, stable, and sustained services, maintain complaint channels, and address illegal content promptly. | 高 | SR028, SR029 |
| CR006 | Providers with public-opinion or social-mobilization characteristics may need security assessments and algorithm-filing procedures in China. | 高 | SR028, SR029 |
| CR007 | If Bulage eventually exposes public model services, APIs, or broad user-facing AI tools in China, data and governance compliance will be a first-order operating burden rather than a back-office task. | 高 | SR001, SR028, SR029 |
| CR008 | China has already moved to formalize a national humanoid-robotics standards system, which raises the bar for explicit interoperability and safety work. | 中 | SR008 |
| CR009 | The EU AI Act creates a risk-based legal framework with high-risk obligations around risk management, dataset quality, logging, documentation, human oversight, cybersecurity, and accuracy. | 高 | SR025, SR027 |
| CR010 | The Commission's AI Act implementation page says transparency rules reach generative AI and continue coming into force through 2026 and beyond. | 中 | SR027 |
| CR011 | CEN-CENELEC says harmonized AI standards are being developed to support legal conformity under the AI Act and can create a presumption of conformity once published. | 高 | SR027, SR032 |
| CR012 | If Bulage ever sells into Europe, integrates with regulated products, or serves multinational industrial customers, the AI Act and its standardization trail can add real documentation and testing cost. | 高 | SR025, SR027, SR032 |
| CR013 | The EU export-control page states that dual-use rules govern goods, software, technology, brokering, transit, and technical assistance. | 中 | SR024 |
| CR014 | BIS frames the EAR as the governing U.S. export-control regime for software and technology, meaning cross-border access to U.S.-origin tooling can carry licensing and screening constraints. | 中 | SR026 |
| CR015 | MERICS argues China's embodied-AI push remains dependent on key external technologies even as domestic ambition accelerates. | 中 | SR006 |
| CR016 | CSET likewise frames embodied AI as a strategic path for China but one that still relies on major inputs across chips, software, and ecosystem infrastructure. | 中 | SR007 |
| CR017 | Together, export-control frameworks and foreign-tool dependence make compute, chips, software frameworks, and cross-border collaboration a strategic risk for Bulage. | 高 | SR024, SR026, SR006, SR007 |
| CR018 | TrendForce says the humanoid market is entering a commercialization phase in H2 2026, which raises time pressure on startups to show proof before capital sentiment shifts. | 中 | SR009 |
| CR019 | Morgan Stanley's higher China shipment forecast, as reported by CNBC, shows how heated 2026 expectations have become. | 中 | SR010 |
| CR020 | Yicai reported that embodied-AI financing in China jumped nearly fivefold, which increases crowding and talent competition even before durable economics are proven. | 中 | SR011 |
| CR021 | KrASIA's 2026 reality-check article argues that China's humanoid race still faces a gap between technical spectacle and commercial value. | 中 | SR004 |
| CR022 | Unite.AI reported that China itself warned of bubble risk as roughly 150 companies crowded into the humanoid field. | 中 | SR005 |
| CR023 | Bubble and crowding risk matters for Bulage because a rich seed valuation can compress quickly if proof milestones arrive slower than the market narrative expects. | 高 | SR004, SR005, SR010, SR011 |
| CR024 | UBTECH's 2025 annual results show that even a scaled humanoid player with rising revenue and a booming humanoid segment still carries large losses and heavy R&D spend. | 高 | SR012, SR013 |
| CR025 | Agility's public-market transaction materials show that category leaders still pursue hundreds of millions in capital even after customer orders and deployment proof emerge. | 高 | SR014, SR015, SR016 |
| CR026 | Capital intensity is therefore a structural risk for Bulage rather than a hypothetical one, especially if the company is building both AI models and deployment-grade robotics. | 高 | SR012, SR014, SR015, SR016 |
| CR027 | BMW says its Figure pilot supported more than 30,000 BMW X3, moved more than 90,000 components, and still remained a pilot with additional use cases under evaluation. | 高 | SR017, SR018 |
| CR028 | Hyundai's Atlas plan is phased into 2028 and beyond, while The Robot Report said Atlas was only beginning real factory work and was not yet at mass deployment. | 高 | SR019, SR020 |
| CR029 | UBTECH publicly names many factory accounts, but those materials still do not disclose cohort durability or full account economics. | 高 | SR021, SR012 |
| CR030 | AGIBOT's 2026 deployment story combines seven standardized solutions, a 10,000th robot milestone, and shipment-share claims, which intensifies competitive risk for younger peers. | 高 | SR022, SR023 |
| CR031 | Competitive risk for Bulage is not just model quality; it is the possibility that better-funded peers arrive first with public deployment proof, services infrastructure, and buyer references. | 高 | SR017, SR018, SR021, SR022 |
| CR032 | Bulage still lacks equivalent public deployment, customer, or safety proof on its own surface. | 高 | SR001, SR002, SR003 |
| CR033 | Specialist legal advisers on both sides of a robotics-sector asset sale described the transaction as subject to regulatory approvals, highlighting how industrial-robot expansion can face legal timing friction. | 高 | SR030, SR031 |
| CR034 | Cross-border partnerships, acquisitions, or overseas sales can therefore introduce approval and review timing risk even before product-market risk is solved. | 高 | SR024, SR026, SR030, SR031 |
| CR035 | Bulage's public materials reveal no disclosed OEM, supplier, compute, or deployment-partner diversification. | 高 | SR001, SR002, SR003 |
| CR036 | Because the first public customer or partner proof is still absent, Bulage likely faces elevated first-account concentration risk once commercialization begins. | 高 | SR001, SR002, SR003, SR021 |
| CR037 | Key-person concentration is also meaningful because public company identity is centered on founder Lin Junyang and not a publicly described wider operating bench. | 中 | SR002, SR003 |
| CR038 | The absence of public privacy, safety, reliability, and support metrics leaves Bulage's operational downside effectively unbounded for outside investors. | 高 | SR001, SR027, SR029 |
| CR039 | Compliance burden, export-control friction, customer-proof delays, and capital intensity can transmit into one another and hit financing, valuation, and hiring simultaneously. | 高 | SR017, SR023, SR024, SR026, SR032 |
| CR040 | The minimum kill criteria for the Bulage thesis are named customer proof, a real compliance roadmap, clarity on compute/tool access, and some safety or reliability evidence. | 高 | SR003, SR017, SR024, SR027, SR032 |
| CV001 | Bulage is publicly described as a Shanghai embodied-intelligence startup with a reported roughly $220 million seed round at roughly a $2 billion valuation in June 2026. | 高 | SV002, SV003, SV004 |
| CV002 | Bulage is less than a year old and still lacks public revenue, customer, and deployment proof. | 高 | SV001, SV002, SV003 |
| CV003 | Bulage also lacks a public product or developer surface strong enough to justify a premium entirely on disclosed technical proof. | 中 | SV001, SV019 |
| CV004 | The category opportunity is large and 2026 commercialization sentiment is strong, but public evidence still shows uneven customer proof and heavy execution risk. | 高 | SV012, SV013, SV019, SV020 |
| CV005 | Figure raised $675 million at a $2.6 billion valuation and paired that financing with OpenAI collaboration and a public BMW commercial agreement. | 中 | SV005, SV006 |
| CV006 | Agility's public-market transaction materials cite a roughly $2.5 billion merger valuation alongside more than $300 million of multi-year Digit v5 orders. | 高 | SV007, SV008, SV009 |
| CV007 | Apptronik closed a $403 million Series A in March 2025 after adding strategic investors and publicly naming Mercedes-Benz, GXO, and Jabil commercial engagements. | 中 | SV021 |
| CV008 | Apptronik's Mercedes-Benz agreement shows that private-market valuation support improves when humanoid companies can point to a named enterprise deployment path. | 高 | SV021, SV022 |
| CV009 | 1X raised $100 million and publicly frames NEO around home access rather than industrial proof at Figure or Agility scale. | 高 | SV010, SV011 |
| CV010 | UBTECH reported RMB2.0 billion of 2025 revenue, a rapidly growing humanoid segment, and still remained loss-making, which illustrates the category's capital intensity. | 高 | SV015, SV016 |
| CV011 | AGIBOT says it led 2025 humanoid shipment share and had rolled out its 10,000th robot by March 2026, indicating that some China players already pair scale narrative with deployment narrative. | 高 | SV017, SV018 |
| CV012 | TrendForce says commercialization enters a more concrete phase in H2 2026, which supports upside narratives for the whole category but also intensifies milestone pressure. | 中 | SV012 |
| CV013 | Public market-cap anchors in July 2026 span roughly $39.3 billion for FANUC, $51.0 billion for Rockwell, $57.8 billion for Teradyne, $120.7 billion for Intuitive Surgical, and $179.8 billion for ABB. | 中 | SV023, SV024, SV025, SV026, SV027 |
| CV014 | Those public companies have durable revenue, installed base, and reporting history that Bulage does not yet have, so they are ceiling references rather than direct price anchors. | 高 | SV023, SV024, SV025, SV026, SV028, SV029, SV030, SV031 |
| CV015 | ABB's annual reporting suite shows more than $33.2 billion of 2025 revenue, underlining how much proof sits beneath large public automation valuations. | 中 | SV028 |
| CV016 | Rockwell markets itself as the world's largest pure-play industrial automation company, another reminder that scaled public comps monetize broad installed bases rather than early embodied-AI option value. | 中 | SV029 |
| CV017 | Intuitive's SEC filings and public valuation illustrate the premium markets will award a robot-enabled company once recurring clinical and financial disclosure is mature. | 高 | SV026, SV030 |
| CV018 | FANUC's investor-reporting surface and market cap show that even pure robotics leaders with decades of installed base trade far above Bulage only after long operating proof cycles. | 高 | SV025, SV031 |
| CV019 | Bulage at a reported $2 billion seed valuation sits below Figure and Agility on headline price, but it also sits materially below them on public evidence. | 高 | SV005, SV006, SV007, SV009, SV001, SV002, SV003 |
| CV020 | Compared with Apptronik, Bulage has a richer valuation-to-proof ratio because Apptronik pairs huge funding with named customer engagements and a production robot narrative. | 高 | SV021, SV022, SV001, SV002, SV003 |
| CV021 | Compared with UBTECH and AGIBOT, Bulage has far less public commercialization proof even though Chinese peers already discuss revenue, shipments, or named industrial deployments. | 高 | SV015, SV016, SV017, SV018, SV001, SV002 |
| CV022 | The cleanest price conclusion from public evidence is that Bulage looks stretched relative to its disclosed proof, even if not obviously irrational relative to private humanoid enthusiasm. | 高 | SV005, SV006, SV007, SV009, SV021, SV022, SV019 |
| CV023 | A bull case for Bulage requires the company to prove that an OEM-neutral embodied brain can generalize across robot bodies and quickly land anchor industrial customers. | 中 | SV002, SV003, SV020 |
| CV024 | A base case assumes Bulage spends most of the current round on research, early pilots, and integration work, then still needs follow-on capital before broad deployment economics are visible. | 中 | SV010, SV015, SV016, SV007, SV009 |
| CV025 | A bear case assumes customer proof remains thin, competition compresses attention, and the next financing happens at a materially lower implied price. | 中 | SV019, SV012, SV013, SV018 |
| CV026 | The single most important positive valuation sensitivity would be a named production-grade customer or partner deployment with measurable operating outputs. | 高 | SV006, SV008, SV018, SV022 |
| CV027 | A second major sensitivity would be repeat-budget or expansion evidence proving Bulage can move from technical interest to durable commercial value. | 中 | SV008, SV015, SV016 |
| CV028 | Compliance clarity and toolchain resilience matter because export-control or governance risk can reduce both commercial velocity and valuation appetite. | 中 | SV020, SV019 |
| CV029 | Without public revenue or customer metrics, scenario analysis is more credible when anchored to milestone probability and peer proof than when anchored to fake precision around near-term ARR. | 高 | SV005, SV007, SV021, SV015 |
| CV030 | The mature public-comp set is useful mainly for showing what fully proven robotics and automation businesses can become, not for justifying Bulage's current entry price. | 高 | SV023, SV024, SV025, SV026, SV027, SV028, SV029, SV030, SV031 |
| CV031 | The most defensible current recommendation is research-more rather than buy, because the next decision-driving facts are still unknown rather than merely noisy. | 高 | SV001, SV002, SV003, SV022 |
| CV032 | Confidence in that recommendation should be medium because the proof gap is clear but the company could still surprise positively with private evidence not visible publicly. | 中 | SV001, SV002, SV003, SV021 |
| CV033 | Risk rating should remain high because financing, customer, product, and compliance uncertainty all remain open together. | 高 | SV019, SV020, SV022 |
| CV034 | Valuation stance should be stretched because the public record still shows far more option value than operating evidence. | 高 | SV001, SV005, SV007, SV021 |
| CV035 | A practical bull range would likely require Bulage to approach the public-proof standard already demonstrated by Figure, Agility, or Apptronik rather than merely repeating their narrative themes. | 高 | SV005, SV006, SV007, SV008, SV021, SV022 |
| CV036 | A practical base range should sit below the current reported price unless and until Bulage discloses named customers, clearer product scope, or technical proof. | 高 | SV001, SV002, SV003, SV022 |
| CV037 | A practical bear range can fall well below unicorn level because the category still has down-round and hype-compression risk once capital stops rewarding promise alone. | 中 | SV019, SV012, SV013 |
| CV038 | At current evidence quality, the expected-value logic is closer to downside protection than upside chasing. | 中 | SV022, SV031 |
| CV039 | Bulage's present valuation support is therefore mostly a combination of founder quality, market timing, and investor appetite rather than disclosed traction. | 高 | SV002, SV003, SV012, SV013, SV021 |
| CV040 | The next diligence package that could move the call toward buy is specific: named customer proof, task-level deployment metrics, product architecture clarity, and evidence that follow-on capital is optional rather than required. | 高 | SV001, SV003, SV022 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Bulage | bulage.cn homepage | |
| SO002 | bulage.com | 黑月传奇三职业 | 自由交易 材料保值 装备保值 账号保值 完善打金系统 |
| SO003 | Crunchbase News | Led By DeepSeek, 10 Frontier Labs Rush Onto The Crunchbase Unicorn Board In June | Shanghai-based Bulage, a robotics intelligence company, raised a $220 million seed round led by Gaorong Capital and HSG. The less than 1-year-old company founded by an Alibaba researcher was valued at $2 billion. |
| SO004 | AsiaICT | A $2 Billion Bet and Lin Junyang’s Second Act - AI | A company without an official name, products, revenue, or even a website achieved this valuation in its debut funding round. |
| SO005 | TechCrunch | Alibaba's Qwen tech lead steps down after major AI push | Junyang Lin, a central technical leader on Alibaba’s Qwen team, said in a post on X on Tuesday that he was “stepping down” from the project. |
| SO006 | VentureBeat | Did Alibaba just kneecap its powerful Qwen AI team? Key figures depart in wake of latest open source release | The departure of Junyang “Justin” Lin, the technical lead who steered Qwen from a nascent lab project to a global powerhouse with over 600 million downloads, marks a volatile inflection point for Alibaba Cloud. |
| SO007 | NYU Shanghai RITS | Junyang Lin Steps Down as Qwen Tech Lead in Abrupt Departure | By 2023, he had become the formal tech lead of the Qwen team, steering the project from a nascent lab effort into a global open-source powerhouse. |
| SO008 | NDTV | Alibaba Group's AI Head, Who Warned Of US-China Tech Gap, Steps Down | He had set up a robotics team just last year and been posting updates about Qwen on X just a day before announcing his departure. |
| SO009 | OfficeChai | Alibaba Qwen's Tech Lead Junyang Lin, 2 Other Researchers Step Down | Lin joined Alibaba in 2019 as a Senior Algorithm Engineer working on NLP and multimodal research and, from 2023, served as the formal tech lead of the Qwen team. |
| SO010 | Baidu Baike | 林俊旸 | 2025年10月8日,林俊旸在社交媒体上发文表示,已在通义团队内部建立机器人和具身智能的小组。 |
| SO011 | 36Kr Europe | Lin Junyang Launches Business: New Company Valued at Around $2 Billion | Exclusive from Intelligence Emergence | The directions he is considering include world models and embodied brains. |
| SO012 | 36Kr Europe | Report: Lin Junyang Secures Tencent Investment with First-round Valuation of 13.5 Billion Yuan and Launches New Round of Financing | As the absolute major shareholder (99%), he established Shanghai Bulage Technology Co., Ltd. |
| SO013 | QbitAI | 林俊旸新公司卜拉格亮相!首轮估值135亿,腾讯高榕红杉全投了 | 高榕创投与红杉中国联合领投,各出资1亿美元。腾讯跟投2000万美元。 |
| SO014 | Sina Tech | 林俊旸新公司「卜拉格」亮相!首轮估值135亿,腾讯高榕红杉全投了 | 企查查信息显示,5月到6月短短一个月内,林俊旸密集注册了多家公司实体,包括100%控股的语用(上海)科技有限公司、99%持股的上海卜拉格科技有限公司。 |
| SO015 | Tencent News | 卜拉格科技完成首轮融资,融资总额达数亿美元 | 这轮已完成的融资由高榕创投和红杉中国领投,这两家风投机构各投资了1亿美元,此外,腾讯在此轮融资中投入了2000万美元。 |
| SO016 | NetEase / Yiou-generated account | 卜拉格获红杉中国等天使轮投资 | 成立于2026年5月27日,主要从事世界模型和具身大脑技术研发业务。 |
| SO017 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SO018 | ChainCatcher | Tencent invests in Alibaba's Qwen former head Lin Junyang's AI laboratory, with a valuation of 2 billion dollars | Tencent invested $20 million, with the total amount raised in the first round reaching several hundred million dollars, and a post-investment valuation of approximately $2 billion. |
| SO019 | AsiaICT | Lin Junyang’s Startup Secures $20 Million in First-Round Financing, with Tencent as Investor - AI | This financing was led by Gaorong Capital and Sequoia China, each contributing $100 million, while Tencent invested $20 million. |
| SO020 | MarketScreener | Shanghai Pragmatics Technology Co., Ltd. announced that it has received funding from HongShan Capital Advisors Limited, Gaorong Capital, Tencent Holdings Limited, and other investors | Shanghai Pragmatics Technology Co., Ltd. announced that it has received funding from HongShan Capital Advisors Limited, Gaorong Capital, Tencent Holdings Limited, and other investors. |
| SO021 | HSG | Home - HSG | HSG is a leading venture capital and private equity firm investing globally across technology, healthcare, and consumer sectors. |
| SO022 | Gaorong Capital | 高榕创投 | 高榕创投是中国最活跃的风险投资机构之一,致力于发现优秀创业者,与他们共建长期价值。 |
| SO023 | Yicai Global | China’s Embodied AI Financing Jumps Five Times to Nearly USD14 Billion on IPO Hopes, LLM Progress | China's embodied artificial intelligence sector attracted CNY93.5 billion (USD13.8 billion) of financing in the first half. |
| SO024 | ChinaBiz Insider | China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026 | Most startups in the cohort carry cash runways of only 18 to 24 months, meaning a reckoning is likely to arrive between 2027 and 2028. |
| SO025 | Unite.AI | China Warns of Bubble Risk as 150 Companies Flood Humanoid Robot Market | China’s top economic planning agency issued a rare warning about the risk of a bubble forming in the country’s humanoid robotics industry. |
| SO026 | Embodied Global | Five $1B+ CEOs Debate Embodied AI: Stockpile Ammunition Now, Mass Deployment in 2 Years — BAAI 2026 Forum | If you haven’t secured top-tier funding and valuation this year, next year will be extremely difficult. |
| SM001 | TrendForce | 产业洞察 - TrendForce集邦咨询: 预估2026年中国人形机器人市场产量将年增94% | 预估2026全年中国人形机器人市场产量年增高达94%。 |
| SM002 | Interact Analysis | Humanoid Robots - 2026 | Interact Analysis | The report focuses on the latest humanoid robot market developments, adoption drivers and barriers, form-factor and technology innovation, component-level changes, production plans, demand sectors and 2026-2035 forecasts. |
| SM003 | China Economic Net | China's Humanoid Robot Boom Gains Speed | IDC is even more bullish, estimating that user spending on embodied intelligent robots in China exceeded $1.4 billion in 2025 and is projected to skyrocket to $77 billion by 2030. |
| SM004 | MERICS | Embodied AI: China's Ambitious Path to Transform Its Robotics Industry | China’s humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SM005 | CSET | China's Embodied AI: A Path to AGI | This trend toward embodied AI is backed by policy support at the national and local government levels. |
| SM006 | KrASIA / 36Kr English | Bubble or breakthrough? China’s humanoid robotics race faces reality check | The main buyers used to be academic research labs. Now we have a new customer profile: state-owned enterprises putting them in lobbies for display. |
| SM007 | CNBC | Morgan Stanley doubles China humanoid robot shipment forecast as commercialization accelerates | Morgan Stanley estimated China's humanoid robot market will reach $2 billion this year and grow to $15 billion by 2030. |
| SM008 | Shanghai Municipal Government | Shanghai maps out plan for embodied AI_Policy Insights | The plan outlines three goals: to attract 100 leading enterprises, to launch 100 application scenarios, and to promote 100 globally competitive products. |
| SM009 | China Daily | China's industrial upgrades accelerate as new growth drivers gain strength | China has built more than 56,000 basic-level smart factories, over 9,000 advanced-level smart factories and more than 500 leading-level smart factories. |
| SM010 | International Federation of Robotics | China Tops World Record of 2 Million Factory Robots | China recorded a world record of 2,027,000 industrial robots working in factories. Annual installations hit 295,000 units in 2024. |
| SM011 | Telecoms.com | Humanoid robot production goes exponential, led by China | Total 2,300 units in 2024 and 13,318 units in 2025. Source: Omdia. |
| SM012 | AGIBOT | Omdia Ranks AGIBOT No.1 Worldwide in Humanoid Robot Shipments in 2025 | AGIBOT shipped 5,168 humanoid robots during the year, accounting for 39% of global market share. |
| SM013 | Yicai Global | China’s Embodied AI Financing Jumps Five Times to Nearly USD14 Billion on IPO Hopes, LLM Progress | China's embodied artificial intelligence sector attracted CNY93.5 billion of financing in the first half. |
| SM014 | ChinaBiz Insider | China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026 | Most startups in the cohort carry cash runways of only 18 to 24 months. |
| SM015 | Embodied Global | ¥96 Billion Poured Into Embodied AI — But the Real Answer Still Eludes the Industry | Skills learned in Factory A will almost certainly fail when deployed in Factory B. |
| SM016 | Unite.AI | China Warns of Bubble Risk as 150 Companies Flood Humanoid Robot Market | More than half of China’s 150-plus humanoid robot companies are startups or cross-sector entrants. |
| SM017 | Shanghai Municipal Government | Implementation Plan of Shanghai Municipality for the Development of the Embodied Intelligence Industry | For application projects demonstrating industrial innovation and integration, financial support of up to 20 percent of the approved total investment, not exceeding 10 million yuan, shall be granted. |
| SM018 | State Council Information Office | Embodied intelligence gains traction in factories, homes and stores | Revenue in the embodied intelligence sector grew 22.4 percent year on year from January to May 2026, and industrial purchases more than tripled. |
| SM019 | State Council Information Office | China's first national standard system for humanoid robotics poised to spur industry development | Over 140 domestic manufacturers released more than 330 different models within 12 months. |
| SM020 | State Council Information Office | China launches national vocational training ground for embodied robots | The National Pilot Base for Embodied AI Applications features more than 130 robots operating in over 30 vocational scenarios. |
| SM021 | Shanghai Municipal Government | Shanghai expo charts future of embodied intelligence | The Yangtze River Delta region hosts more than 50 percent of China's embodied intelligence companies and over half of the sector's financing. |
| SM022 | The Next Web | Barclays says humanoid robots can offset 60% of China’s 37M worker shortfall by 2035 | China's deployment of humanoid robots could offset as much as 60% of the country's projected labour-force decline by 2035. |
| SM023 | Barclays Investment Bank | Robots roll out, economies rewire | Physical AI has so far emerged as a productivity and efficiency enhancer across labour-intensive and asset-heavy industries. |
| SM024 | QbitAI | 林俊旸新公司卜拉格亮相!首轮估值135亿,腾讯高榕红杉全投了 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SM025 | AsiaICT | A $2 Billion Bet and Lin Junyang’s Second Act - AI | Today’s investment logic revolves around three pillars: technological paradigms, talent density, and whether a founder can shape the next era of innovation. |
| SM026 | Crunchbase News | Led By DeepSeek, 10 Frontier Labs Rush Onto The Crunchbase Unicorn Board In June | The new unicorn frontier labs are focused on new architectures in AI model development in robotics, physics and self-learning. |
| SP001 | Figure | Figure 03 home page | Figure 03 is a general purpose humanoid robot for every day. |
| SP002 | Figure | Introducing Helix | Helix is the first VLA that runs entirely onboard embedded low-power-consumption GPUs, making it immediately ready for commercial deployment. |
| SP003 | PRNewswire / Figure AI | Figure Raises $675M at $2.6B Valuation and Signs Collaboration Agreement with OpenAI | Figure has raised $675M in Series B funding at a $2.6B valuation. |
| SP004 | PRNewswire / Figure AI | Figure announces commercial agreement with BMW Manufacturing to bring general purpose robots into automotive production | BMW Manufacturing and Figure will pursue a milestone-based approach. |
| SP005 | 1X | NEO home robot | NEO works autonomously by default. For any chore it doesn't know, you can schedule a 1X Expert to guide it. |
| SP006 | 1X | 1X secures $100M in Series B funding | 1X has raised over $125 million in less than 12 months. |
| SP007 | 1X | Introducing NEO Gamma | NEO Gamma's design opens the door to start internal home testing. |
| SP008 | Agility Robotics | Solutions | A fully autonomous tool with proven commercial deployments, Digit is a tireless partner for your team. |
| SP009 | Agility Robotics | Agility Robotics broadens relationship with Amazon | Agility anticipates production capacity of hundreds of Digit robots in the first year, with the capability to scale to more than 10,000 robots per year. |
| SP010 | Agility Robotics | Agility Robotics announces strategic investment and agreement with Schaeffler Group | Schaeffler intends to purchase humanoid robots from Agility Robotics for use across the whole global Schaeffler plant network. |
| SP011 | Agility Robotics | Agility Robotics and Churchill Capital Corp XI announce confidential submission of draft registration statement | The proposed business combination is expected to provide more than $620 million in gross proceeds. |
| SP012 | The Robot Report | Boston Dynamics, Google reunite on next-gen Atlas humanoid | All units for 2026 are already committed, shipping to the RMAC and Google DeepMind, with additional customers planned for 2027. |
| SP013 | Unitree | G1 humanoid robot | G1 ... $13,500 Deposit. |
| SP014 | Unitree | H1 humanoid robot | Moving speed of 3.3m/s(world record). |
| SP015 | Unitree | About Unitree | More than 200 domestic and foreign patent applications have been submitted, including more than 180 authorized patents. |
| SP016 | UBTECH | UBTECH Walker S Industrial Humanoid Robot | With 41 servo joints with force feedback ... Walker S can reliably perceive its surroundings, humans and objects. |
| SP017 | UBTECH | UBTECH Walker S2 Humanoid Robot | Walker S2 is able to swap battery autonomously within 3 minutes. |
| SP018 | AGIBOT | AGIBOT home page | Introducing AGIBOT World, the first Large Scale, Enterprise Quality, Realistic Task Dataset and Ecosystem for Embodied AI. |
| SP019 | AGIBOT | AGIBOT ranked No.1 globally in humanoid robot shipment volume and market share in 2025, according to Omdia | AGIBOT shipped 5,168 humanoid robots during the year, accounting for 39% of global market share. |
| SP020 | TrendForce | Global Humanoid Robot Market to Enter Commercialization Phase in H2 2026, with Unitree and AgiBot Leading China's Market | Unitree Robotics and AgiBot are projected to account for nearly 80% of total shipments. |
| SP021 | CNBC | Morgan Stanley doubles China humanoid robot shipment forecast as commercialization accelerates | Chinese companies dominated the top five positions by shipments, while American rival Figure AI ranked seventh. |
| SP022 | MERICS | Embodied AI: China's Ambitious Path to Transform Its Robotics Industry | China's humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SP023 | State Council Information Office | China's first national standard system for humanoid robotics poised to spur industry development | Over 140 domestic manufacturers released more than 330 different models within the 12 months. |
| SP024 | KrASIA / 36Kr English | Bubble or breakthrough? China's humanoid robotics race faces reality check | Most institutional investors and academic researchers expect the sector to reach maturity over the next five to ten years. |
| SP025 | Yicai Global | China's Embodied AI Financing Jumps Five Times to Nearly USD14 Billion on IPO Hopes, LLM Progress | Industry sources, however, caution that mass commercial deployment of robots remains some way off. |
| SP026 | Unite.AI | China Warns of Bubble Risk as 150 Companies Flood Humanoid Robot Market | UBTech Robotics began mass production and delivery of its Walker S2 industrial humanoid robot in mid-November, shipping several hundred units to factory partners. |
| SP027 | Bulage | bulage.cn homepage | |
| SP028 | Crunchbase News | Led By DeepSeek, 10 Frontier Labs Rush Onto The Crunchbase Unicorn Board In June | Shanghai-based Bulage, a robotics intelligence company, raised a $220 million seed round led by Gaorong Capital and HSG. |
| SP029 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SP030 | BMW Manufacturing | Spartanburg Plant | BMW Manufacturing Co., LLC is located in Spartanburg, South Carolina. |
| SI001 | Bulage | bulage.cn homepage | |
| SI002 | Crunchbase News | Led By DeepSeek, 10 Frontier Labs Rush Onto The Crunchbase Unicorn Board In June | Shanghai-based Bulage, a robotics intelligence company, raised a $220 million seed round led by Gaorong Capital and HSG. |
| SI003 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SI004 | PRNewswire / Figure AI | Figure Raises $675M at $2.6B Valuation and Signs Collaboration Agreement with OpenAI | This new capital will be used strategically for scaling up AI training, robot manufacturing, expanding engineering headcount, and advancing commercial deployment efforts. |
| SI005 | 1X | 1X secures $100M in Series B funding | The company intends to utilize the new capital to bring to market its second generation android NEO. |
| SI006 | 1X | NEO home robot | NEO ... $200 Deposit. |
| SI007 | Agility Robotics | Investor Relations | Transaction expected to create the only U.S. publicly listed pure-play humanoid company with proven commercial deployments. |
| SI008 | SEC / Churchill XI | Exhibit 99.1 - Joint press release of Churchill Capital Corp XI and Agility Robotics | The proposed business combination is expected to provide more than $620 million in gross proceeds. |
| SI009 | Agility Robotics | Solutions | Arc connects Digit to your existing warehouse automation ... Monitor robot workflows, view live metrics, and manage your entire fleet with ease. |
| SI010 | Nasdaq / Business Wire syndication | Agility Robotics to Go Public Through $2.5 Billion Merger with Churchill Capital Corp XI | Agility has already secured more than $300 million of multi-year orders for Digit v5 ... with a growing pipeline of over 30 customers. |
| SI011 | Unitree | G1 humanoid robot | G1 ... $13,500 Deposit. |
| SI012 | TrendForce | Global Humanoid Robot Market to Enter Commercialization Phase in H2 2026, with Unitree and AgiBot Leading China's Market | Unitree ... in 2025, revenue from humanoid robots surpassed that of quadruped robots for the first time ... Combined gross margin across both segments reached 60%. |
| SI013 | MERICS | Embodied AI: China's Ambitious Path to Transform Its Robotics Industry | Guotai Securities calculates humanoid commercial viability threshold at CNY160,000 ... China's average humanoids cost CNY300,000-500,000 each. |
| SI014 | State Council Information Office | China's first national standard system for humanoid robotics poised to spur industry development | The release of this standard system can unify industrial technical specifications ... reduce coordination and adaptation costs across the industrial chain. |
| SI015 | Yahoo Finance / Research and Markets | Chinese Humanoid Robot Industry Development Report 2026 Now Available | The market was segmented into low (RMB 100,000-200,000), mid (RMB 250,000-500,000), and high (above RMB 500,000) price categories. |
| SI016 | State Council Information Office | China advances toward scaled commercialization of humanoid robots | Humanoid robots in China are now transitioning into practical industrial use, and are also poised to move from labs into a phase of mass production and commercial deployment. |
| SI017 | Global Times | China’s humanoid robot industry accelerates commercialization, marked with growing sales, surging investment | Companies are opening offline experience stores intensively, not primarily for short-term sales, but to cultivate market awareness. |
| SI018 | UBTECH | UBTECH Walker S Industrial Humanoid Robot | Walker S can automatically connect to the manufacturing management system to exchange information in real-time. |
| SI019 | UBTECH | UBTECH Walker S2 Humanoid Robot | Walker S2 is able to swap battery autonomously within 3 minutes. |
| SI020 | HKEX / UBTECH | Annual Results Announcement for the Year Ended December 31, 2025 | Revenue increased by 53.3% to RMB2,001.0 million ... revenue from full-size embodied intelligent humanoid robot products and services grew rapidly ... to RMB820.6 million. |
| SI021 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million. |
| SI022 | RobotToday | UBTECH Robotics FY2025 Results: Humanoid Revenue Surges 2,200%, Loss Narrows Sharply | R&D Expenditure RMB507.5M ... Cash & Equivalents RMB4,887.9M. |
| SI023 | AGIBOT | AGIBOT ranked No.1 globally in humanoid robot shipment volume and market share in 2025, according to Omdia | AGIBOT shipped 5,168 humanoid robots during the year, capturing 39% of global market share. |
| SI024 | CNBC | Morgan Stanley doubles China humanoid robot shipment forecast as commercialization accelerates | Morgan Stanley estimated China's humanoid robot market will reach $2 billion this year. |
| SI025 | KrASIA / 36Kr English | Bubble or breakthrough? China's humanoid robotics race faces reality check | What's left is a familiar disconnect: ballooning valuations with little revenue to back them up. |
| SI026 | Yicai Global | China's Embodied AI Financing Jumps Five Times to Nearly USD14 Billion on IPO Hopes, LLM Progress | Mass commercial deployment of robots remains some way off. |
| SI027 | UBTECH | 2025 Annual Report | 2025 Annual Report |
| SE001 | Bulage | bulage.cn homepage | |
| SE002 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SE003 | NetEase / Yiou-generated account | 卜拉格获红杉中国等天使轮投资 | 成立于2026年5月27日,主要从事世界模型和具身大脑技术研发业务。 |
| SE004 | Figure | Figure 03 home page | Figure 03 is a general purpose humanoid robot for every day. |
| SE005 | Figure | Introducing Helix | Helix is the first VLA that runs entirely onboard embedded low-power-consumption GPUs. |
| SE006 | Google DeepMind | Gemini Robotics 1.5 brings AI agents into the physical world | Gemini Robotics 1.5 is our most capable vision-language-action model. |
| SE007 | Google DeepMind | Gemini Robotics 1.5 Technical Report | Official technical report for Gemini Robotics 1.5. |
| SE008 | NVIDIA Developer | NVIDIA Isaac GR00T | NVIDIA Isaac GR00T is an open reference platform for general-purpose humanoid robots. |
| SE009 | NVIDIA Developer | NVIDIA Isaac Lab Open-Source Modular Framework | Isaac Lab is the foundational robot learning framework of the NVIDIA Isaac GR00T platform. |
| SE010 | NVIDIA Technical Blog | Inside NVIDIA Halos for Robotics: A Full-Stack Functional Safety System for Physical AI | NVIDIA Halos OS is a comprehensive, full-stack safety system. |
| SE011 | Physical Intelligence | Our First Generalist Policy | π0 is a prototype model that combines large-scale multi-task and multi-robot data collection with a new network architecture. |
| SE012 | GitHub / UBTECH-Robot | GitHub - UBTECH-Robot/Thinker: Thinker | We are pleased to open-source Thinker, a state-of-the-art vision-language foundation model specifically engineered for embodied intelligence. |
| SE013 | AGIBOT | AGIBOT Open-Sources AGIBOT WORLD 2026 Dataset to Accelerate Embodied AI Development | AGIBOT WORLD 2026 will be released in five phases. |
| SE014 | AGIBOT | AGIBOT home page | Introducing AGIBOT World, the first Large Scale, Enterprise Quality, Realistic Task Dataset and Ecosystem for Embodied AI. |
| SE015 | UBTECH | UBTECH Walker S Industrial Humanoid Robot | Walker S integrates deeply with LLM. |
| SE016 | UBTECH | UBTECH Walker S2 Humanoid Robot | BrainNet 2.0 + Co-Agents Building the Dual-Loop AI System. |
| SE017 | HKEX / UBTECH | Annual Results Announcement for the Year Ended December 31, 2025 | Our full-stack technological advancements enable embodied intelligent humanoid robots to achieve superior task planning, dexterous manipulation, navigation and mobility, and human-computer interaction capabilities. |
| SE018 | Unitree | H1 humanoid robot | The advanced powertrain provides the highest level of speed, power, maneuverability and flexibility. |
| SE019 | Unitree | About Unitree | Unitree fully self-researches key core robot components such as motors, reducers, controllers, LIDAR and algorithms. |
| SE020 | State Council Information Office | China's first national standard system for humanoid robotics poised to spur industry development | The standard system is structured on six pillars ... application, and safety and ethics. |
| SE021 | MERICS | Embodied AI: China's Ambitious Path to Transform Its Robotics Industry | China's humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SE022 | CSET | China's Embodied AI: A Path to AGI | This trend toward embodied AI is backed by policy support at the national and local government levels. |
| SE023 | KrASIA / 36Kr English | Bubble or breakthrough? China's humanoid robotics race faces reality check | The sector's path to commercialization remains murky. |
| SE024 | Agility Robotics | Solutions | Our commitment doesn't stop at deployment. We back every relationship with on-site service, online support, and real-time monitoring. |
| SE025 | AGIBOT | AGIBOT ranked No.1 globally in humanoid robot shipment volume and market share in 2025, according to Omdia | AGIBOT has built a diversified humanoid robot portfolio spanning full-sized humanoids, compact half-sized humanoids, and wheeled embodied robots. |
| SU001 | Bulage | bulage.cn homepage | |
| SU002 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SU003 | NetEase / Yiou-generated account | 卜拉格获红杉中国等天使轮投资 | 成立于2026年5月27日,主要从事世界模型和具身大脑技术研发业务。 |
| SU004 | Agility Robotics | Digit Deployed at GXO in Historic Humanoid RAAS Agreement | For the first time, a humanoid robot was deployed in commercial operations, as Digit stepped into the workforce at a GXO facility near Atlanta. |
| SU005 | GXO Logistics | GXO Signs Industry-First Multi-Year Agreement with Agility Robotics | This agreement, which follows a proof-of-concept pilot in late 2023, is both the industry’s first formal commercial deployment of humanoid robots and first Robots-as-a-Service deployment of humanoid robots. |
| SU006 | The Robot Report | Boston Dynamics shows Atlas humanoid working at Georgia Hyundai plant | This was the first time Atlas has been out of the laboratory doing real work. |
| SU007 | Hyundai Motor Group | Hyundai Motor Group Announces AI Robotics Strategy to Lead Human-Centered Robotics Era at CES 2026 | Atlas to be deployed at HMGMA by 2028 for sequencing tasks, advancing human-centric smart factory innovation. |
| SU008 | PRNewswire / Figure AI | Figure announces commercial agreement with BMW Manufacturing to bring general purpose robots into automotive production | BMW Manufacturing and Figure will pursue a milestone-based approach. |
| SU009 | BMW Group PressClub | BMW Group to deploy humanoid robots in production in Germany for the first time | Within ten months, the robot Figure 02 supported the production of more than 30,000 BMW X3, working ten-hour shifts daily from Monday to Friday. |
| SU010 | 1X | NEO home robot | NEO ... $200 Deposit. |
| SU011 | Unitree | G1 humanoid robot | G1 ... $13,500 Deposit. |
| SU012 | UBTECH Robotics | UBTECH Humanoid Robot Industrial Application Solution | Walker S1 has entered the BYD factory to perform handling tasks and has completed the world's first one-stop autonomous logistics application that humanoid robots coordinate with autonomous logistics vehicles, AMRs/AGVs and intelligent manufacturing management systems. |
| SU013 | UBTECH Robotics | UBTECH Walker S1 Humanoid Robot | For Multi-task Industrial Scenarios | Walker S1 has been introduced into vehicle manufacturing assembly lines to assist in car production, working collaboratively with autonomous logistics vehicles and AMRs/AGVs through smart manufacturing management systems. |
| SU014 | UBTECH Robotics | UBTECH Walker S2 Humanoid Robot | Autonomous Battery Swapping for Mass Production Delivery | Walker S2 is able to swap battery autonomously within 3 minutes. |
| SU015 | AGIBOT | AGIBOT Global Store | AGIBOT Global Store is Now LIVE! |
| SU016 | PRNewswire / AGIBOT | AGIBOT Declares 2026 "Deployment Year One" at APC 2026, Accelerating the Era of Embodied AI Productivity | To accelerate commercialization, AGIBOT introduced seven standardized productivity solutions targeting high-value industry scenarios. |
| SU017 | PRNewswire / KEENON Robotics | KEENON Debuts First Bipedal Humanoid Service Robot at WAIC, Showcasing Role-Specific Embodied AI Solutions | At the medical station, the humanoid XMAN-F1 partners with logistics robot M104 to create a closed-loop smart healthcare solution. |
| SU018 | KEENON Robotics | Smart Service Robots & Solutions | Please select an industry category ... Restaurant ... Hotel ... Medical ... Industrial Transport. |
| SU019 | TrendForce | Global Humanoid Robot Market to Enter Commercialization Phase in H2 2026, with Unitree and AgiBot Leading China's Market | Unitree ... in 2025, revenue from humanoid robots surpassed that of quadruped robots for the first time ... Combined gross margin across both segments reached 60%. |
| SU020 | Yahoo Finance / Research and Markets | Chinese Humanoid Robot Industry Development Report 2026 Now Available | The market was segmented into low (RMB 100,000-200,000), mid (RMB 250,000-500,000), and high (above RMB 500,000) price categories. |
| SU021 | HKEX / UBTECH | Annual Results Announcement for the Year Ended December 31, 2025 | Our full-stack technological advancements enable embodied intelligent humanoid robots to achieve superior task planning, dexterous manipulation, navigation and mobility, and human-computer interaction capabilities. |
| SU022 | KrASIA / 36Kr English | Bubble or breakthrough? China's humanoid robotics race faces reality check | The sector's path to commercialization remains murky. |
| SU023 | Agility Robotics | Investor Relations | Transaction expected to create the only U.S. publicly listed pure-play humanoid company with proven commercial deployments. |
| SU024 | BMW Manufacturing | Spartanburg Plant | BMW Manufacturing Co., LLC is located in Spartanburg, South Carolina. |
| SU025 | Unitree | About Unitree | Unitree fully self-researches key core robot components such as motors, reducers, controllers, LIDAR and algorithms. |
| SU026 | AGIBOT | AGIBOT ranked No.1 globally in humanoid robot shipment volume and market share in 2025, according to Omdia | AGIBOT has built a diversified humanoid robot portfolio spanning full-sized humanoids, compact half-sized humanoids, and wheeled embodied robots. |
| SR001 | Bulage | bulage.cn homepage | |
| SR002 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SR003 | NetEase / Yiou-generated account | 卜拉格获红杉中国等天使轮投资 | 成立于2026年5月27日,主要从事世界模型和具身大脑技术研发业务。 |
| SR004 | KrASIA / 36Kr English | Bubble or breakthrough? China's humanoid robotics race faces reality check | The sector's path to commercialization remains murky. |
| SR005 | Unite.AI | China Warns of Bubble Risk as 150 Companies Flood Humanoid Robot Market | UBTech Robotics began mass production and delivery of its Walker S2 industrial humanoid robot in mid-November, shipping several hundred units to factory partners. |
| SR006 | MERICS | Embodied AI: China's Ambitious Path to Transform Its Robotics Industry | China's humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SR007 | CSET | China's Embodied AI: A Path to AGI | This trend toward embodied AI is backed by policy support at the national and local government levels. |
| SR008 | State Council Information Office | China's first national standard system for humanoid robotics poised to spur industry development | The standard system is structured on six pillars ... application, and safety and ethics. |
| SR009 | TrendForce | Global Humanoid Robot Market to Enter Commercialization Phase in H2 2026, with Unitree and AgiBot Leading China's Market | Unitree ... in 2025, revenue from humanoid robots surpassed that of quadruped robots for the first time ... Combined gross margin across both segments reached 60%. |
| SR010 | CNBC | Morgan Stanley doubles China humanoid robot shipment forecast as commercialization accelerates | Morgan Stanley estimated China's humanoid robot market will reach $2 billion this year. |
| SR011 | Yicai Global | China's Embodied AI Financing Jumps Five Times to Nearly USD14 Billion on IPO Hopes, LLM Progress | Mass commercial deployment of robots remains some way off. |
| SR012 | HKEX / UBTECH | Annual Results Announcement for the Year Ended December 31, 2025 | Our full-stack technological advancements enable embodied intelligent humanoid robots to achieve superior task planning, dexterous manipulation, navigation and mobility, and human-computer interaction capabilities. |
| SR013 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million. |
| SR014 | Agility Robotics | Investor Relations | Transaction expected to create the only U.S. publicly listed pure-play humanoid company with proven commercial deployments. |
| SR015 | SEC / Churchill XI | Exhibit 99.1 - Joint press release of Churchill Capital Corp XI and Agility Robotics | The proposed business combination is expected to provide more than $620 million in gross proceeds. |
| SR016 | Nasdaq / Business Wire syndication | Agility Robotics to Go Public Through $2.5 Billion Merger with Churchill Capital Corp XI | Agility has already secured more than $300 million of multi-year orders for Digit v5 ... with a growing pipeline of over 30 customers. |
| SR017 | PRNewswire / Figure AI | Figure announces commercial agreement with BMW Manufacturing to bring general purpose robots into automotive production | BMW Manufacturing and Figure will pursue a milestone-based approach. |
| SR018 | BMW Group PressClub | BMW Group to deploy humanoid robots in production in Germany for the first time | Within ten months, the robot Figure 02 supported the production of more than 30,000 BMW X3, working ten-hour shifts daily from Monday to Friday. |
| SR019 | Hyundai Motor Group | Hyundai Motor Group Announces AI Robotics Strategy to Lead Human-Centered Robotics Era at CES 2026 | Atlas to be deployed at HMGMA by 2028 for sequencing tasks, advancing human-centric smart factory innovation. |
| SR020 | The Robot Report | Boston Dynamics shows Atlas humanoid working at Georgia Hyundai plant | This was the first time Atlas has been out of the laboratory doing real work. |
| SR021 | UBTECH Robotics | UBTECH Humanoid Robot Industrial Application Solution | Walker S1 has entered the BYD factory to perform handling tasks and has completed the world's first one-stop autonomous logistics application that humanoid robots coordinate with autonomous logistics vehicles, AMRs/AGVs and intelligent manufacturing management systems. |
| SR022 | PRNewswire / AGIBOT | AGIBOT Declares 2026 "Deployment Year One" at APC 2026, Accelerating the Era of Embodied AI Productivity | To accelerate commercialization, AGIBOT introduced seven standardized productivity solutions targeting high-value industry scenarios. |
| SR023 | AGIBOT | AGIBOT ranked No.1 globally in humanoid robot shipment volume and market share in 2025, according to Omdia | AGIBOT has built a diversified humanoid robot portfolio spanning full-sized humanoids, compact half-sized humanoids, and wheeled embodied robots. |
| SR024 | European Commission | Exporting dual-use items | |
| SR025 | EUR-Lex | Regulation - EU - 2024/1689 - EN | |
| SR026 | Bureau of Industry and Security | EAR | |
| SR027 | European Commission | AI Act | |
| SR028 | Cyberspace Administration of China | 生成式人工智能服务管理暂行办法 | 提供者应当依法开展预训练、优化训练等训练数据处理活动。 |
| SR029 | China Law Translate | Interim Measures for the Management of Generative Artificial Intelligence Services | These Measures are drafted on the basis of the Cybersecurity Law of the PRC, the PRC Data Security Law, the Personal Information Protection Law of the PRC... |
| SR030 | ARQIS | ARQIS advises Agile Robots SE on the acquisition of thyssenkrupp Automation Engineering | |
| SR031 | Taylor Wessing | Taylor Wessing advises thyssenkrupp on the sale of Automation Engineering to Agile Robots | |
| SR032 | CEN-CENELEC | Artificial Intelligence | |
| SV001 | Bulage | bulage.cn homepage | |
| SV002 | Phoenix Tech | 林俊旸新公司“卜拉格”亮相,首轮估值135亿 | 而具体的创业方向,则聚焦世界模型和具身大脑。 |
| SV003 | NetEase / Yiou-generated account | 卜拉格获红杉中国等天使轮投资 | 成立于2026年5月27日,主要从事世界模型和具身大脑技术研发业务。 |
| SV004 | Crunchbase News | Led By DeepSeek, 10 Frontier Labs Rush Onto The Crunchbase Unicorn Board In June | Shanghai-based Bulage, a robotics intelligence company, raised a $220 million seed round led by Gaorong Capital and HSG. |
| SV005 | PRNewswire / Figure AI | Figure Raises $675M at $2.6B Valuation and Signs Collaboration Agreement with OpenAI | This new capital will be used strategically for scaling up AI training, robot manufacturing, expanding engineering headcount, and advancing commercial deployment efforts. |
| SV006 | PRNewswire / Figure AI | Figure announces commercial agreement with BMW Manufacturing to bring general purpose robots into automotive production | BMW Manufacturing and Figure will pursue a milestone-based approach. |
| SV007 | Agility Robotics | Investor Relations | Transaction expected to create the only U.S. publicly listed pure-play humanoid company with proven commercial deployments. |
| SV008 | SEC / Churchill XI | Exhibit 99.1 - Joint press release of Churchill Capital Corp XI and Agility Robotics | The proposed business combination is expected to provide more than $620 million in gross proceeds. |
| SV009 | Nasdaq / Business Wire syndication | Agility Robotics to Go Public Through $2.5 Billion Merger with Churchill Capital Corp XI | Agility has already secured more than $300 million of multi-year orders for Digit v5 ... with a growing pipeline of over 30 customers. |
| SV010 | 1X | 1X secures $100M in Series B funding | The company intends to utilize the new capital to bring to market its second generation android NEO. |
| SV011 | 1X | NEO home robot | NEO ... $200 Deposit. |
| SV012 | TrendForce | Global Humanoid Robot Market to Enter Commercialization Phase in H2 2026, with Unitree and AgiBot Leading China's Market | Unitree ... in 2025, revenue from humanoid robots surpassed that of quadruped robots for the first time ... Combined gross margin across both segments reached 60%. |
| SV013 | CNBC | Morgan Stanley doubles China humanoid robot shipment forecast as commercialization accelerates | Morgan Stanley estimated China's humanoid robot market will reach $2 billion this year. |
| SV014 | Yahoo Finance / Research and Markets | Chinese Humanoid Robot Industry Development Report 2026 Now Available | The market was segmented into low (RMB 100,000-200,000), mid (RMB 250,000-500,000), and high (above RMB 500,000) price categories. |
| SV015 | HKEX / UBTECH | Annual Results Announcement for the Year Ended December 31, 2025 | Our full-stack technological advancements enable embodied intelligent humanoid robots to achieve superior task planning, dexterous manipulation, navigation and mobility, and human-computer interaction capabilities. |
| SV016 | Yicai Global | China's Ubtech Pops After Posting 23-Fold Leap in Annual Humanoid Robot Sales | Ubtech sold 1,079 embodied artificial intelligence robots taller than 160 centimeters last year, raking in CNY820.6 million. |
| SV017 | AGIBOT | AGIBOT ranked No.1 globally in humanoid robot shipment volume and market share in 2025, according to Omdia | AGIBOT has built a diversified humanoid robot portfolio spanning full-sized humanoids, compact half-sized humanoids, and wheeled embodied robots. |
| SV018 | PRNewswire / AGIBOT | AGIBOT Declares 2026 "Deployment Year One" at APC 2026, Accelerating the Era of Embodied AI Productivity | To accelerate commercialization, AGIBOT introduced seven standardized productivity solutions targeting high-value industry scenarios. |
| SV019 | KrASIA / 36Kr English | Bubble or breakthrough? China's humanoid robotics race faces reality check | The sector's path to commercialization remains murky. |
| SV020 | MERICS | Embodied AI: China's Ambitious Path to Transform Its Robotics Industry | China's humanoids still lack precision and dexterity and are mostly deployed in limited tasks and in site-specific trials. |
| SV021 | Apptronik | Apptronik Closes Additional Series A Funding, Bringing Total Round to $403M | Apptronik today announced the successful close of an oversubscribed $403M Series A funding round. |
| SV022 | Apptronik | Apptronik and Mercedes-Benz Enter Commercial Agreement That Will Pilot Apptronik’s Apollo Humanoid Robot in Mercedes-Benz Manufacturing Facilities | The partnership represents Apptronik’s first publicly announced commercial deployment of Apollo. |
| SV023 | CompaniesMarketCap | ABB (ABBN.SW) - Market capitalization | As of July 2026 ABB has a market cap of $179.79 Billion USD. |
| SV024 | CompaniesMarketCap | Rockwell Automation (ROK) - Market capitalization | As of July 2026 Rockwell Automation has a market cap of $50.98 Billion USD. |
| SV025 | CompaniesMarketCap | Fanuc (6954.T) - Market capitalization | As of July 2026 Fanuc has a market cap of $39.30 Billion USD. |
| SV026 | CompaniesMarketCap | Intuitive Surgical (ISRG) - Market capitalization | As of July 2026 Intuitive Surgical has a market cap of $120.65 Billion USD. |
| SV027 | CompaniesMarketCap | Teradyne (TER) - Market capitalization | As of July 2026 Teradyne has a market cap of $57.83 Billion USD. |
| SV028 | ABB Group | Annual Reporting Suite 2025 | ABB | 2025 in numbers: Revenues $33,220 mn. |
| SV029 | Rockwell Automation | Investor Relations | Rockwell Automation | US | As the world's largest pure-play industrial automation company, we are building a future where industrial operations are smarter, more efficient, and more sustainable. |
| SV030 | Intuitive Surgical | SEC Filings | Intuitive Surgical | Jul 21, 2026 10-Q Quarterly report which provides a continuing view of a company's financial position. |
| SV031 | FANUC | Integrated Reports - Library - Investors | Conventional "Annual Report" has been redesigned from the 2022 issue, provided as "Integrated Report" including ESG information. |