初创公司尽调
尽调报告 Humanoid robotics / embodied AI private, post-Series B 2026-08-01

Zhiyuan Robot

中国出货领先和真实部署证据让 Zhiyuan 具备投资价值,但公开经济性仍太薄,难以支撑激进定价。

观察:Zhiyuan 是中国人形机器人赛道可信的前排玩家,已有真实部署证据;但当前公开估值给出的确定性,已经高于公开经济数据能支撑的程度。

封面要素

估值锚点 01
2070 USD M [CV004]
投资建议 02
track [CV015]
估值立场 03
stretched [CV016]
成立时间锚点 04
2023-02 [CO001]
公开点名客户证据 05
Fulin / China Mobile / Longcheer [CU004, CU008, CU012]
出货领先信号 06
China leader cluster [CV008]

公司概况

Zhiyuan Robot 的国际品牌是 AGIBOT,是一家位于上海的具身 AI 机器人公司,公开画像已远超单一人形机器人原型。官方材料和本报告审阅的报道支撑其全栈打法:工业人形机器人、偏交互系统、清洁和工作流产品、面向开发者的资产,以及由伙伴驱动的商业化。公开部署证据在中国制造业和战略采购中最强,但经济性披露仍落后于公司叙事和估值规模。

官网
www.zhiyuanrobot.com
成立时间
2023-02-01
创始人
Deng Taihua, Peng Zhihui
创立地点
Shanghai, China
总部
Shanghai, China
产品
组合覆盖工业和轮式人形机器人、服务与交互系统、清洁设备、面向开发者的开源资产,以及具身 AI 模型 / 数据基础设施。
客户
公开证据集中在制造业、战略采购和伙伴主导的企业商业化上;Fulin Precision、China Mobile、Chery 以及 Longcheer 是最清晰的点名信号。
商业模式
模式混合了机器人硬件、部署与集成、维护与支持,以及未来靠数据、工具和伙伴生态扩张沉淀的长期平台价值。
阶段
private, post-Series B
融资情况
公开报道指向 2025 年多轮战略融资:Tencent 参与早期轮次,随后出现 JD 支持或与 JD 有关的投资方;估值锚点约 RMB15 billion / 约 US$2.07 billion,具体轮次机制仍未公开。
[CO001, CO005, CO006, CO016, CO017, CO018, CO020, CI001]

执行摘要

主要优势

  • 公开证据显示,公司已有真实出货和商业化势头,包括在中国出货叙事中的领先位置,以及多个具名企业案例。
  • 公司做的不是单一机器人本体,而是更宽的平台:产品线、开放技术资产、模型、数据和伙伴生态一起铺开。
  • Fulin Precision、China Mobile、Longcheer 等具名部署和采购案例,让 Zhiyuan 比许多仍停留在叙事阶段的人形机器人创业公司有更扎实的采用证据。
  • 战略投资者和生态伙伴提高了制造、渠道、商业化支持持续叠加的概率。

主要风险

  • 公开披露的收入、毛利率、烧钱速度、债务和股权结构仍过于不透明,难以干净地做估值承销。
  • 客户证据仍集中在少数旗舰账户,叙事集中和收入集中风险都不低。
  • 硬件加服务的资本强度看起来是该品类的结构性问题;如果部署经济性成熟慢,后续稀释和现金跑道压力会升高。
  • 公司已经进入真实工业场景,但运营可靠性、安全和支持负担仍只披露了一部分。
  • 国际扩张和伙伴主导部署会在海外付费客户深度跑出来之前,先带来合规、质量控制和治理复杂度。

未决问题

  • 经审计的 2025-2026 年收入、毛利率,以及出货到收入的桥接
  • 当前现金余额、债务结构、月度烧钱速度和下行情景下的现金跑道计划
  • 按账户类型列示的客户队列表:在线机器人数量、扩张、流失和服务负担
  • 2025 年融资条款、股权结构表和清算优先权堆叠的精确细节
  • 头部客户集中度、海外付费客户数量和现场支持 KPI
  • 大型企业买家或 IPO 投资者会要求的安全、隐私、责任和事件控制材料包

目录

Chapter 01

01公司概览

1.1 身份与使命

理解 Zhiyuan Robot,更合适的框架是上海具身 AI 机器人公司,而不是只做单一人形机器人演示的团队。官方中英文材料持续把中文公司身份 智元创新 与国际品牌 AGIBOT 放在一起,并把使命表述为「用智能机器创造无限生产力」。这些页面重要,因为它们也界定了公司的经营范围:全栈机器人、智能算法、开放平台,以及更大的应用生态。官方简介给出了目前公开证据中最清晰的成立时间线:2023 年 2 月公司注册,2023 年 8 月首款 Expedition A1 发布。这条时间线比种子轮语境更精确,也说明公司从注册到公开硬件发布的速度异常快。公开产品面已经覆盖 Expedition 人形机器人、Lingxi 与 Genie 平台、数据服务和工具链,因此后续章节应把 Zhiyuan 当作平台加产品公司来分析,而不只是一个拿着单一人形机器人原型的研究团队。[CO001, CO002, CO003, CO004, CO010, CO033]

KPI 快照表
指标数值或状态日期置信度缺口或备注
注册 / 创立锚点公司注册信息显示为 2023 年 2 月2023-02官方时间线比泛泛媒体说法更精确。
国际品牌AGIBOT当前官方英文材料使用 AGIBOT,中文材料使用 智元创新。
总部 / 基地上海;公开材料提到临港制造布局当前仍缺少一份权威总部事实表。
使命用智能机器创造无限生产力当前官方公司使命表述。
运营模式全栈具身 AI 机器人公司当前机器人本体、算法和平台都在公开叙事里。
最新可辩护估值锚点截至 2025 年 3 月,约 US$2.07B / RMB 14.7-15.0B2025-03公开媒体区间一致,但精度还不到一手条款清单。
最新融资状态2025 年多轮融资,Tencent 参与,随后 JD / 上海具身智能基金加入2025具体轮次规模和是否含老股转让仍未披露。
规模验证官方时间线称 2025 年 1,000 台机器人;Forbes 称截至 2026 年 3 月 10,000 台;官方发布称 2026 年中达到 15,000 台2025-2026出货里程碑不能等同于收入。
商业验证China Mobile、Chery 等已披露订单2025是有用的牵引信号,但没有经审计的合同收入桥。
当前员工数null当前公开招聘强度可见;干净的员工数不可见。

本快照有意把公共证据和待解私有指标拆开;null 表示审阅过的公开证据不支持。

[CO001, CO002, CO004, CO018, CO026, CO029]
FO002: 公司快照逻辑

产品宽度、AI 栈、投资方、客户和供应链如何拼成 Zhiyuan 的公开公司故事。

这是基于公开证据重建的分析性运营模型,不是公司发布的流程图。

[CO002, CO003, CO016, CO020, CO026, CO027]

1.2 创始人与领导层

官方领导层披露实质改变了创始人叙事。公司现在公开点名 Deng Taihua 为创始人、董事长兼 CEO,Peng Zhihui 为联合创始人、总裁兼 CTO。相比只围绕 Peng 互联网名气展开的流行说法,这给出了更正式的控制图景。Peng 仍然极其重要:独立报道把他与 Huawei Genius Youth 计划、早期 OPPO 经历和庞大的创作者关注度联系起来,这些给 Zhiyuan 带来工业初创公司少见的品牌触达和人才磁力。但已披露的团队显示,公司运营面更宽,覆盖市场、具身业务运营、综合业务、科学和 HR 等高管角色。公开报道还显示,高管会细致回应定价、供应链选择和商业化节奏,说明领导层优化的不只是机器人 R&D,还包括规模化部署。主要治理缺口不在于高管是谁,而在于董事会控制权、优先权和创始人投票权在幕后如何分配。[CO005, CO006, CO007, CO008, CO009, CO027]

领导层与创始人表
人员职务公开依据重要性未决尽调点
Deng Taihua创始人 / 董事长 / CEO官方领导页;公开访谈界定实际控制权和商业化姿态,不只停留在以 Peng 为中心的公共叙事。厘清董事会控制权、投票权和既往关联方安排。
Peng Zhihui联合创始人 / 总裁 / CTO官方领导页;独立媒体画像核心技术品牌和人才磁石,兼具异常高的零售 / 创作者可见度。厘清正式董事会角色、长期留任机制和授权结构。
Yao Maoqing合伙人 / SVP / 具身业务总裁官方领导页;媒体访谈商业化、模型路线图和供应链经济性的公开代表。按业务线索取运营指标。
Wang Chuang合伙人 / SVP / 通用业务总裁官方领导页;客户订单报道客户场景、定价和工业部署上的关键发言人。按场景索取订单管线和转化率。
Luo Jianlan合伙人 / SVP / 首席科学家官方领导页传递的信号是内部科研深度,而不只是系统集成。厘清论文、专利和模型治理产出。
Jiang Qingsong合伙人 / 联席总裁 / 营销与服务总裁官方领导页说明公司在明确搭建商业化和服务层。厘清渠道、服务和部署后组织规模。
Niu Jia合伙人 / VP / CHRO官方领导页暗示公司在推进规模化招聘和组织搭建。索取当前员工数、流失率和城市分布。

覆盖范围有限,仅限公司公开点名或经审阅独立报道引用的领导者。

[CO005, CO006, CO007, CO008, CO009, CO029]

1.3 融资与投资方

即使具体轮次经济性仍不完整,公开融资证据的方向性已经很强。独立 2025 年报道称 Tencent 领投了 2025 年 3 月轮次,JD 以及 Shanghai Embodied Intelligence Fund 参与了后续融资;Zhidx 则描述 2025 年 3 月估值约 US$2.07 billion,或约 RMB 14.7-15.0 billion。这个区间足够一致,可以把 Zhiyuan 视为独角兽,但还不够精确,不能假装公开记录已经给出完整股权结构真相。投资方构成比单个头条数字更关键。公开报道显示,Zhiyuan 同时拿到财务 VC 和战略产业资本,包括 Tencent、JD、SAIC、LG Electronics、Mirae Asset,以及 Baidu Ventures、Dinghui、C Capital 等较早投资方。这样的组合支撑一个判断:Zhiyuan 搭建的不只是融资栈,也是在搭建分销、制造和生态联盟。IPO 讨论仍需谨慎处理:当前公开报道把香港上市计划描述为 2026 年目标或情景,而不是已经完成的融资事件。[CO016, CO017, CO018, CO019, CO020, CO037]

利益相关方或投资人图谱
利益相关方角色公开证据重要性尽调问题
Tencent2025 年 3 月轮次领投方36Kr 和 Zhidx增加资本、信号效应和生态邻近性。索取董事会席位、持股比例和任何战略商业权利。
JD / JD 关联资本2025 年后续投资方36Kr除资本外,还可能撬动物流和零售场景。厘清是否存在商业试点或采购关联(如有)。
上海具身智能基金2025 年后续投资方36Kr国资背书和本地生态支持。厘清支持是否包括采购、设施或政策资源。
SAIC、BYD 系资本等产业投资方战略投资人层Zhidx 和独立报道能压缩供应链、制造和客户导入周期。确认战略投资人是实际付费客户,还是仅为股东。
LG Electronics / Mirae Asset后续报道中的跨境战略-财务参与方Zhidx支撑国际化和全球品牌野心。索取条款,并确认是否暗含产品、渠道或工厂合作。
Baidu Ventures、Hillhouse、Dinghui 与 C Capital公开报道提到的更早期财务投资方36Kr 和 Zhidx显示资本结构是密集联合投资,而不是单一金主叙事。索取逐轮持股比例和清算优先权。
China Mobile / Chery / 其他已披露客户商业利益相关方,而非股权投资人Tencent News 客户报道重要,因为这一品类里部署验证可能比资本更关键。索取已签订单金额、复购率和部署成熟度。

这是公开利益相关方图谱,不是股权结构表。经济权利、董事会控制和二级市场流动性仍未披露。

[CO016, CO017, CO018, CO019, CO020, CO026]
FO003: 快照 KPI

精选最能概括 Zhiyuan 当前成熟度和披露画像的公开标志。

这些是公开尽调标志,不是经审计的财务 KPI。

[CO001, CO018, CO023, CO026, CO028, CO029]

1.4 规模与牵引力

在私人化的人形机器人公司中,Zhiyuan 的公开牵引力证据异常强,但其中混合了出货、试点、订单和营销里程碑,不能混为一谈。官方材料显示,公司在 2025 年累计通用机器人突破 1,000 台,随后声称到 2026 年中具身机器人达到 15,000 台;Forbes 报道称,2026 年 3 月前总出货 10,000 台,其中单季度出货 5,000 台。TrendForce 另称 AgiBot 和 Unitree 可能是中国 2026 年人形机器人产量领先者。客户证据也更具体:China Times/Tencent News 的画像提到 China Mobile 和 Chery 订单,以及其他数百万元级合同;管理层还公开讨论过某一型号低至 RMB 98,000 的产品定价。即便如此,公开记录仍无法把这些事实换算为经审计收入、装机利用率或清晰员工数。活跃招聘和出货里程碑证明了动能,但还没有证明成熟披露质量。[CO013, CO021, CO022, CO023, CO024, CO025]

里程碑表
日期事件类型金额 / 状态参与方含义
2023-02公司注册创立注册于上海Zhiyuan / AGIBOT最好的公开创立锚点。
2023-08发布 Expedition A1产品首款产品发布Zhiyuan显示从概念到发布的速度异常快。
2024-01制造工厂落地上海规模工厂布局确立Zhiyuan显示公司很早就愿意搭建实体规模。
2024-08发布 Expedition A2 和 Lingxi X1产品产品组合扩张Zhiyuan走出单一发布平台。
2024-12宣布通用机器人商业化量产规模通用机器人进入量产AGIBOT商业化叙事的重要转折。
2025-03发布具身基础模型和 Lingxi X2产品GO-1 / X2 里程碑Zhiyuan显示 AI 栈野心,而不只是硬件迭代。
2025-03Tencent 领投融资轮获报道融资公开报道中的约 US$2.07B 估值区间Tencent + 既有投资人坐实独角兽身份叙事。
2025-04发布 Genie Studio产品开发者平台上线Zhiyuan把平台策略拓展到开发者。
2025-07 至 2025-08China Mobile / Chery 等订单公开披露合作已点名商业订单客户 + Zhiyuan增加真实客户验证。
2026-03Forbes / 2026 年领导层叙事称已出货 10,000 台机器人规模大规模出货里程碑Zhiyuan / 独立媒体显示快速放量,但不是经审计收入。
2026-07A3 在 WAIC 展示自主乒乓球演示产品旗舰公开演示Zhiyuan + WAIC 观众展示具身控制野心和营销触达。
2026-07宣布第 15,000 台具身机器人下线规模官方新生产纪录Zhiyuan暗示制造加速仍在持续。
2026-04著作权侵权听证获报道不利浦东法律案件Zhiyuan + 原告提醒投资人,规模不会抹掉法律执行风险。

这条时间线是本章唯一记录口径,纳入公开来源可见的积极规模事件和不利法律信号。

[CO001, CO010, CO011, CO012, CO013, CO014]
FO001: 公司里程碑时间线

公开时间线覆盖从注册到 2026 年规模、产品和法律事件。

时间线只纳入本轮审阅过的有日期公开事件;私人或未公开融资可能缺失。

[CO001, CO010, CO011, CO012, CO013, CO014]

1.5 里程碑与未解缺口

浮现出的战略图景,是一家公司正在从成立快速迈向平台化,但仍要求外部投资者跨过几个关键披露缺口。官方发布给出了可信的里程碑序列:早期注册、数月内推出首个产品、2024 年上海制造落地、2025 年发布具身模型和工具、2026 年提出更大的公开出货或展示主张。独立报道又补上了融资、工业订单和商业化压力的证据。但一些最重要的尽调事实仍顽固地停留在私域,或彼此矛盾。公开来源没有给出经审计收入、当前债务、准确员工数或最终轮次文件;2026 年 IPO 叙事仍是计划,而非交易所申报文件;版权案件和全栈模式争议等反向信号也说明,公司不会因为可见度高就已经去风险。后续章节的正确结论既不是「只有泡沫」,也不是「赢家已经证明」,而是「真实的类别领导者,但投资判断仍有重大盲区」。[CO018, CO019, CO031, CO032, CO036, CO037]

Chapter 02

02市场分析

2.1 市场定义

Zhiyuan Robot 的正确市场边界,比「所有机器人」窄得多,又比「人形机器人硬件」略宽。官方材料显示,公司销售的不只是机器人本体,还包括具身 AI 工具、数据服务和开发者基础设施。因此,这里的市场应是一组通用人形机器人系统,加上让机器人可部署的周边具身软件和服务层。同时,公司并没有理由把无关的工业机械臂、纯软件助手或通用自动化都纳入同一个分母。边界应包含人形机器人以及直接附着的栈——模型、工具、数据和运营服务——并排除没有产品重叠的相邻自动化品类。这一点重要,因为宽泛的机器人 TAM 可以让任何初创公司看起来很大;纪律化的具身人形机器人市场框架,才让 Zhiyuan 的真实可触达市场和举证负担能与 Figure、Unitree、UBTECH、Agility、Apptronik 等同行比较。[CM001, CM002, CM003, CM004, CM027, CM032]

市场定义表
品类纳入支出排除支出买方 / 付款方重要性
通用人形机器人机器人硬件、核心控制、具身软件、部署服务传统固定式工业机器人和通用自动化企业运营方、实验室、服务场馆最贴合 Zhiyuan 当前公开足迹。
具身 AI 平台工具开发者工具、数据采集、模型训练、集成不绑定机器人部署的横向企业 AI开发者、研究实验室、机器人 OEM 团队匹配 Genie Studio 和研究定位。
工业部署预算工位劳动力替代、搬运、物流、维护消费电子需求工厂和物流运营方可能是近期最清晰的预算池,ROI 逻辑也最明确。
服务 / 接待机器人引导、交互、网点或场馆自动化通用智能设备营销支出电信运营商、场馆、企业证明负担低于家用机器人。
家用人形机器人愿景长期家庭协助和陪伴当前仅企业部署消费者 / 家庭战略空间大,但今天证据薄。

边界逻辑把公司真实公开产品足迹和过宽的机器人 TAM 定义区分开。

[CM001, CM002, CM003, CM004, CM020]

2.2 市场规模

公开市场规模估算在方向上一致,在量级上分歧很大。IDC 的 2026 年人形机器人研究是当前最有用的锚点,因为它把品类落在 2025 年的实际出货和收入上:约 18,000 台、约 US$440 million 收入,中国厂商领先。随后,预测机构把空间放大到远高于这批装机基数:Mordor 估计 2026 年市场为 US$3.93 billion,到 2031 年升至 US$17.8 billion;Global Market Insights 则估出更大的 2026 年 US$10.9 billion,以及 2035 年 US$192.7 billion。这些数字都不能不经转换就当成 Zhiyuan 的近期可触达市场。更合适的解释是三层镜头:长期广义 TAM 很大,近期 SAM 小得多,主要是中国工业 / 商业部署;当前已证明的 SOM 更小,因为只有少数用例有清楚的公开证据。[CM005, CM006, CM007, CM008, CM009, CM010]

TAM / SAM / SOM 测算表
发布方 / 视角年份 / 周期地域数值方法或局限
IDC 出货与收入视角2025 年实际值全球约 18k 台 / 约 $440M 收入当前最佳商业化锚点,但偏后视镜。
Mordor 自上而下市场预测2026-2031全球2026 年 US$3.93B,2031 年 US$17.8B预测覆盖的品类范围更宽。
Global Market Insights 自上而下预测2026-2035全球2026 年 US$10.9B,2035 年 US$192.7B多年期 TAM 视角非常宽。
TrendForce 部署视角2026中国预计产出增长 94%以中国为中心的供应和商业化指标,不是完整 TAM。
Zhiyuan 近期已验证 SOM 代理指标2025-2026中国优先null没有公开分母能把当前部署转成可辩护的 SOM。

本表有意混合自上而下和商业化视角;null 标记公开证据尚不支持的公司特定分母。

[CM005, CM006, CM007, CM008, CM009, CM010]
FM001: 市场规模测算视角

在证据约束下,用金字塔把长期大 TAM 和近期小得多的已验证需求分开。

各层数值不可相加,因为单位和范围不同;这座金字塔是决策视角,不是数学拆解。

[CM005, CM008, CM009, CM036]
FM002: 市场估计区间

将公开市场规模估计统一为十亿美元后的低 / 基准 / 高视角。

低值和高值是围绕外部点估计画出的可视区间,不是公司指引。

[CM005, CM006, CM007, CM008, CM009]

2.3 买方分层

买方地图让公司自身故事变得更具体。Zhiyuan 的产品表面和媒体访谈暗示至少五个需求池:工业制造、物流和仓储、服务 / 接待、科研教育,以及娱乐或品牌激活。每个池子的买方、用户和付款方都不同。工厂和物流运营方掌握劳动生产率预算,关心正常运行时间、载荷和 ROI。服务场所和电信网点购买的是接待、引导和客户互动。科研买方买的是灵活性和开发者入口,而非近期生产率。娱乐或品牌客户为新奇感、互动和媒体价值付费。家用机器人是最大的长期叙事市场,但近期公开市场最弱,因为它要求最高的安全、可靠性和成本证明。就 Zhiyuan 而言,最清晰的公开预算证据落在结构化工业和商业部署中,而不是品类故事承诺的广阔家庭市场。[CM011, CM012, CM016, CM017, CM018, CM019]

细分市场 / 买方图谱
细分市场买方用户付款方 / 预算负责人采用触发因素
制造 / 工厂自动化工厂或运营负责人产线操作员和维护团队工业资本开支 / 运营开支负责人劳动力替代、一致性、安全、产出效率。
物流 / 仓储仓库总经理或自动化负责人搬运人员和主管运营和自动化预算结构化重复移动和包裹流。
服务 / 接待网点、场馆或零售运营方前台员工和客户服务人工或客户体验预算客户交互和信息采集。
科研 / 教育实验室负责人或高校 PI研究人员和学生科研经费或机构预算灵活实验和开发者接入。
娱乐 / 品牌活动品牌营销方或活动运营方表演者和访客营销或活动预算新奇感、PR 触达和互动价值。

仅为代表性买方图谱;买方、用户和付款方角色会随地域和用例显著变化。

[CM004, CM016, CM017, CM018, CM019, CM020]
FM003: 买方 / 细分市场地图

买方就绪度矩阵,显示哪些细分市场今天最接近有预算的部署。

单元格是基于保留来源集做出的定性判断,不是调研测量值。

[CM016, CM017, CM018, CM019, CM020, CM027]
FM004: 采用漏斗 / 价值链地图

公开证据从广泛市场热度收窄到可重复部署经济性,落差很大。

数值代表相对证明阶段,不是内部 CRM 转化率。

[CM011, CM012, CM021, CM024, CM033, CM034]

2.4 增长驱动与约束

最强的市场驱动因素已经熟悉,并且在公开研究中越来越可见:人口老龄化、结构化物理工作流里的劳动力短缺、组件成本下降、具身模型表现提升,以及人形机器人无需彻底重建基础设施就能使用人类设计的环境。中国政策也是实实在在的顺风,不只是叙事装饰。机器人计划、人形机器人指引和 2026 年标准活动都显示,监管者希望品类走出演示。但市场约束仍然很重。可靠性、泛化、安全、算力依赖、场景集成和客户经济性证明,仍主导采用风险。管理层自己也承认,广泛部署取决于性价比提升和更强模型能力。换句话说,增长故事真实存在,但品类仍在从试点阶段的兴奋,跨向部署阶段的纪律。投资者需要判断这次跨越的速度,而不只是最终 TAM 的大小。[CM013, CM014, CM015, CM021, CM022, CM023]

增长驱动与约束表
驱动因素 / 约束方向时间窗口Zhiyuan 启示尽调追问
中国政策和标准支持正向当前至中期加速本地生态搭建,也提高采购正当性。跟踪标准如何改变真实招标要求。
劳动力压力和老龄化正向中期支撑工业和服务场景的自动化预算。索取带基准人工经济账的客户 ROI 案例。
出货提升带动成本曲线下行正向但不确定当前至中期只有可靠性同步提升,才能放大总可用市场(TAM)。索取物料清单和毛利率变化路径。
可靠性、安全和集成负担负向当前压缩可服务市场(SAM),拖慢大规模部署。索取正常运行时间、事故和服务成本证据。
公开 ROI 分母薄弱负向当前让依赖估值的 TAM 论证更脆弱。索取部署层面的回本周期和利用率数据。

时间窗口和方向同样重要;若产品可靠性和部署经济账不能同步改善,若干“驱动因素”也帮不上忙。

[CM013, CM014, CM015, CM021, CM022, CM023]

2.5 尽调缺口

本章的主要尽调问题不是缺少乐观,而是缺少分母。公开来源可以展示品类扩张、中国政策支持、同行商业化方向,甚至广泛出货增长。它们还不能展示清晰的 Zhiyuan 专属 SAM 或 SOM、按细分稳健计算的 ROI,或按客户类型划分的渗透率。预测机构分歧也太大,任何单一报告都不应主导估值讨论。最安全的分析姿态,是保留彼此冲突的估算,而不是把它们平均成虚假的精确。因此,后续章节要把市场规模按区间处理,买方准备度按细分判断,采用时点受举证负担和标准落地约束。Zhiyuan 显然参与了未来十年最重要的硬件 AI 品类之一,但当前公开记录仍支持一条谨慎、受证据约束的市场论点,而不是简单的「市场巨大,所以赢家必然」结论。[CM010, CM024, CM031, CM034, CM035, CM036]

Chapter 03

03竞争者

3.1 竞争格局

Zhiyuan 所在的同行集合正在快速成形,但并不是每家机器人公司都同样重要。当前最值得比较的一组包括 Unitree、Figure AI、UBTECH、Agility Robotics 和 Apptronik,因为它们都公开表达了通用人形机器人野心,并且至少有部分企业商业化路径。格局也要放在现状替代品旁边看:固定自动化、轮式物流机器人,以及更简单的工业系统,在不需要人形灵活性时,往往能更便宜地解决很多任务。这个区分重要,因为人形机器人初创公司不只是在和其他人形机器人竞争,也在和「够用就好」的自动化竞争。在直接同行中,中国玩家目前在硬件规模化节奏和供应链密度上更强,美国同行则在资本市场叙事和高端品牌上更强。因此,核心比较不只是机器人对机器人,而是在不同举证负担下,运营模式对运营模式。[CP001, CP002, CP003, CP004, CP022, CP031]

竞品画像表
竞品类别规模 / 融资信号目标细分市场差异化局限
Zhiyuan / AGIBOT中国人形机器人全栈平台2025 年公开估值锚点 ~US$2.07B;10k-15k 出货叙事中国工业、服务和生态牵引型部署产品宽度、中国规模、投资方生态、模型 / 工具栈公开收入和部署分母透明度弱。
Unitree中国硬件牵引型人形机器人同业TrendForce 显示其与 AgiBot 同为出货份额领先者;H1 价格可谈通用人形机器人和零部件硬件透明度和零部件广度留存来源里的企业品牌验证较少。
Figure AI美国高端人形机器人龙头Series C 轮 >$1B,估值 $39B汽车和企业自动化BMW 验证、高端资本入口、Helix 叙事高估值跑在公开收入披露前面。
UBTECH中国工业人形机器人同业上市公司画像,Walker S 聚焦工业工业多任务场景工业流程清晰度和上市公司成熟度前沿估值热度不如 Figure 或 Zhiyuan 明确。
Agility Robotics美国仓储 / 企业同业2026 年公开合并和设施扩建叙事仓储和企业物流商业化纪律和基础设施公开单量规模可见度较低。
Apptronik美国 Apollo 人形机器人同业据 CNBC,Series A 轮 >$935M;估值 $5B制造、3PL、企业部署资本入口和商业化合作伙伴公开单量规模仍不如中国龙头可见。

所选公司覆盖最显眼的直接同业,但排除了许多更小的新进入者和非人形机器人替代方案。

[CP001, CP004, CP005, CP009, CP011, CP012]
FP001: 竞争定位图

按公开部署证明和资本 / 生态力量做相对定位。

坐标轴只使用从已审阅公开证据合成的 1-10 定性评分。

[CP004, CP005, CP007, CP011, CP013, CP016]

3.2 竞争者画像

公开画像显示,估值领先、部署领先和定价透明度之间分裂明显。Figure 完成估值 $39 billion 的 C 轮融资后,是最显眼的估值异类,也通过 BMW 拿到异常强的点名企业证据。Unitree 在硬件和组件表面更透明,外部分析师把它与 AgiBot 一起放在中国出货曲线顶端。UBTECH 是严肃的工业竞争者,因为 Walker S 明确围绕工厂场景定位,而不是新奇互动。Agility 和 Apptronik 还没有展示出与中国领导者相同的公开台数规模,但两者都投入大量精力建设商业化基础设施、设施和伙伴关系,后续可能变得重要。Zhiyuan 处在中间位置:比多数美国同行拥有更可见的中国规模和生态密度,但融资与收入质量在全球投资者眼中还不够清晰。[CP004, CP005, CP006, CP007, CP008, CP009]

功能 / 能力矩阵
标准ZhiyuanUnitreeFigureUBTECHAgility / Apptronik
出货 / 规模可见度中国叙事里高中国叙事里高中低
工业流程聚焦
定价透明度中低
开发者 / 工具叙事
独立企业验证中低

单元格仅概括已审阅的公开证据;低可见度不等于低能力。

[CP004, CP006, CP007, CP009, CP011, CP015]

3.3 能力对比

从买方视角看,最重要的维度不是抽象的「AI 领先」主张,而是形态适配、部署证明、定价透明度、供应链韧性、灵巧操作和可维护性。Zhiyuan 最清晰的比较优势,是产品组合宽度、中国生态密度和快速规模化叙事。Figure 最清晰的优势,是 BMW 带来的高端企业验证,以及无人匹敌的头条融资。Unitree 在硬件透明度和定价线索上得分较高。UBTECH 的工业工作流框架较强;Apptronik 和 Agility 则在商业化基础设施和企业伙伴纪律更重要的场景下最强。值得注意的是,灵巧手和具身模型闭环正在成为明确的竞争变量。Gasgoo 对 Zhiyuan 灵巧手剥离的反向解读也说明,同一个能筑护城河的全栈野心,如果焦点碎片化,也会制造战略拉扯。因此,买方不应把宽度直接等同于持久优势,除非宽度真正转化为部署和服务表现。[CP009, CP010, CP015, CP019, CP020, CP021]

定价 / 包装对比
公司公开价格 / 包装信号包含内容未知项含义
Zhiyuan2025 年一个公开低端标价信号约 RMB 98k入门级人形机器人硬件参考点实际 ASP、折扣、服务、企业打包有参考价值,但不适合作跨厂商对比。
UnitreeH1 商店显示需联系获取真实价格人形机器人硬件,按协商报价真实企业折扣和服务条款比多数同业更透明,但仍未标准化。
Figure留存来源未见公开标准化机器人定价企业部署和平台价值主张ASP、租赁、服务、支持经济账估值故事目前跑在价格透明度前面。
UBTECH / Agility / Apptronik留存来源未见清晰公开的标准化标价企业部署和合作伙伴包装真实合同结构和全生命周期服务该品类销售更看解决方案经济账,而不是目录标价。

同业公开价格披露普遍薄弱,因此包装方式和部署验证比标价对比更重要。

[CP009, CP018, CP019, CP020]
FP002: 功能宽度 / 能力图

压缩矩阵,展示各主要同行在公开证据中看起来最强的领域。

基于公开证据的定性视角,不是实验室基准测试。

[CP009, CP010, CP015, CP016, CP018, CP020]

3.4 护城河评估

这个品类中最强的护城河,可能不是某一次演示、某个执行器或某个模型检查点,而是数据采集、部署、服务反馈、供应链深度、制造执行和融资耐力之间的闭环。按这个定义,Zhiyuan 在中国拥有一个可信护城河,因为它把广泛的公开产品、产业投资方、可见订单和快速放量制造叙事组合在一起。但护城河尚未定型。Figure 拥有更多资本和一个全球共振的企业旗舰客户。Unitree 的硬件价格可读性可能相当甚至更强。UBTECH 具备上市公司成熟度和工业焦点。Apptronik 与 Agility 的商业化基础设施,未来可能转化为更强的企业耐久性。因此,Zhiyuan 的护城河论点真实存在,但仍有条件:按中国生态执行衡量时最好看,按全球可比披露质量和重复付费部署衡量时最弱。[CP015, CP016, CP021, CP022, CP027, CP028]

护城河耐久性 / 竞争风险清单
护城河主张威胁严重性重要性尽调追问
中国规模和生态密度同业追上,或本地需求碎片化没有耐久服务闭环的规模可能很快商品化。索取重复部署和支持指标。
全栈集成层级过多导致聚焦承压Gasgoo 显示,同一套护城河叙事也会制造运营拖累。索取业务线盈利能力和组织聚焦图。
工业投资方网络投资方没有转化为需求中高战略资本不等于付费客户。索取来自股东关联客户的收入。
产品组合广度SKU 过多稀释执行中高广度有助于总可用市场(TAM)叙事,但可能伤害聚焦。索取 SKU 级利用率和利润率。
中国出货领先全球信任和披露落后同业国际客户可能高度看重透明度。索取经审计收入和海外部署数量。

本清单聚焦护城河耐久性,而非一般公司风险;不少威胁来自成功后的拉扯,不只是显性失败。

[CP015, CP021, CP023, CP024, CP026, CP027]
FP003: 护城河 / 就绪度 KPI

紧凑的公开指标,用来判断 Zhiyuan 当前相对同行是否具备防守性。

这些是定性尽调 KPI,不是经审计的运营指标。

[CP015, CP016, CP017, CP018, CP025, CP033]

3.5 竞争风险

主要竞争风险,是软件和服务护城河完全成形前,硬件先走向商品化。如果多家厂商都能在结构化任务中做到「够用」的移动、操作和可靠性,那么资本、分销和服务经济性可能比任何单一技术突破都更重要。因此,公开定价透明度弱并不是旁注,而是护城河争论的核心。另一个风险是,高额私人估值会遮蔽其他玩家更低调但更持久的客户证据。最后一个风险是战略过度延展。Gasgoo 对 Zhiyuan 灵巧手策略的反向分析,凸显了全栈公司可能赢得注意力,却同时面对焦点拉扯。实际结论是,投资者不应只按演示或估值给同行排序,还要看谁正在搭建可重复的部署闭环,谁仍主要依赖品牌热度。按这个标准,赛场仍足够流动,今天的排行榜可能很快改写。[CP018, CP019, CP024, CP031, CP035, CP036]

Chapter 04

04财务

4.1 收入模式

Zhiyuan 的公开记录支撑一条多层收入故事,但还不是一份清晰财务报表。官方产品、研究和解决方案页面显示,公司销售的不止一个人形机器人 SKU:公开产品面包括全尺寸和轮式机器人、数据采集与训练基础设施,以及面向具体场景的解决方案。这意味着变现架构很可能覆盖硬件销售、部署与集成、维护,最终还会叠加软件或工具。企业证据也强化了这一点。China Mobile 招标显然是项目型收入;Fulin Precision 工厂证据则显示,公司先把机器人嵌入工作流,再扩大量。因此,收入质量很可能取决于验收里程碑、部署服务和特定站点工作,而不只是出货台数。投资者应把出货头条当作领先指标,而不是收入报表本身。[CI001, CI002, CI003, CI004, CI005, CI014]

收入流表
收入流机制单位当前价值 / 状态质量尽调追问
机器人硬件销售向企业买方销售人形或轮式机器人按机器人或项目包公开产品和部署记录清楚显示该收入流存在按型号拆出实际 ASP,并拆出交付时确认收入与验收时确认收入的占比。
部署 / 集成服务场地适配、流程设计、安装和调试按部署或场地Fulin 和解决方案页面强烈指向该收入流量化集成费、实施毛利和单场地人工负担。
运营 / 维护支持技术支持和安装后优化按合同收费,或嵌入项目经济账公开部署报道明确把它列为成本驱动项中低披露年度服务附着率、续约条款和每台活跃机器人的支持人员配置。
数据 / 训练 / 工具围绕具身智能的开发者、数据或训练基础设施按平台、服务或打包产品收费D1 Ultra、研究和解决方案页面可见,但未做财务量化披露付费客户、定价模型,以及收入是经常性还是项目制。
战略招标 / 大项目大型企业或准采购合同按招标 / 项目包China Mobile 和 Fulin 显示这是有分量的渠道把招标金额映射到交付排期、付款节点和可复用性。

公开证据支持每条收入流存在,但没有披露任何单一收入流的收入结构或确认节奏。

[CI001, CI002, CI003, CI005, CI014, CI015]
定价 / 变现表
价格 / 合同信号标价与实际价格折扣 / 未知项来源含义
一个型号的公开低端信号约 ~RMB98k标价 / 促销信号型号组合、企业选项和实际折扣不清楚Tencent News 访谈有用的底部信号,但不是可信的工业 ASP。
可比工业人形机器人区间 RMB400k-RMB800k估计市场区间A2-W 具体合同价格未披露Yicai Fulin 文章可能更接近当前工业用途的企业部署价格。
China Mobile 针对 AgiBot 全尺寸机器人的 RMB78m 套餐项目合同金额服务、软件和验收阶段是否包含在内未知Yicai China Mobile 招标报道大单很可能不只打包硬件。
Fulin 交易接近 100 台 A2-W项目订单规模最终交付、验收和服务经济账未公开Assembly / Shanghai gov 报道单量标题需要收入桥,才能进入投资测算。
竞品协商定价惯例(Unitree 联系报价、同业企业试点)协商 / 不透明多数厂商不发布标准企业目录Unitree 及同业商业化来源品类定价仍由解决方案牵引,而非目录牵引。

定价证据混合了不同产品、场景和合同结构;应把它看作边界条件,而不是清晰价目表。

[CI011, CI012, CI013, CI014, CI015]
FI001: 收入模式桥接图

公开变现路径从机器人单机,延伸到项目打包、部署工作和更长尾的服务。

基于保留来源的定性瀑布图;不是量化收入桥接。

[CI001, CI003, CI014, CI015, CI019, CI020]

4.2 单位经济性

公开单位经济性证据不完整,但方向上仍有用。披露最强的案例是 Fulin Precision:一支小型现场机器人队伍在三小时内处理了 800 多个箱子,管理层描述的劳动力替代区间,在常规运行下约等于 0.7 名工人,全天候使用时最高可达 1.4 到 2 名工人。这个案例不能证明回本,但与当地每名工人每年约 RMB80,000 的人工成本放在一起,可以给出一阶 ROI 视角。同一篇文章在成本侧同样重要:部署和维护需要技术人员,首次落地可能耗时数月,之后部署才会快得多。公开价格也很混乱。公开资料称 A2-W 可能落在 RMB400,000 到 RMB800,000 的工业区间,而另一次管理层讨论提到另一型号低得多的入门价格。正确结论是,定价取决于产品和场景,支持经济性可能与机器人 BOM 一样重要。[CI006, CI007, CI008, CI009, CI010, CI011]

单位经济模型表
指标数值 / null置信度重要性尽调追问
工业人形机器人价格带可比机器人 RMB400k-RMB800k;A2-W 据称落在该区间为工业部署划出粗略收入分母。索取按合同拆分的实际 ASP,以及硬件与服务占比。
入门级公开标价信号某公开型号信号为 ~RMB98k能说明产品组合广度;若套到企业部署,容易形成错误锚点。索取分型号价格表和附加选项。
劳动力替代代理指标常规情况下等效 0.7 名工人;24/7 利用率下为 1.4-2.0给买方沟通提供第一层 ROI 逻辑。索取各工位实测产出、正常运行时间和人工替代假设。
部署周期代理指标首个项目 3-4 个月;后续部署可大幅压缩意味着前期集成成本重,后续学习曲线会释放收益。索取部署周期中位数和单场地工程小时数。
同业毛利率基准UBTECH 2025 年 37.7%,高于 2024 年的 28.7%这是观察该品类规模化后、扣除运营费用前毛利形态的最佳公开信号。索取 Zhiyuan 按产品和部署类型拆分的毛利率。
同业营运资金压力UBTECH 应收账款 ~RMB1.84bn说明企业人形机器人增长可能伴随大额应收敞口。索取 Zhiyuan 应收账款账龄、坏账准备和付款节点。
Zhiyuan 获客成本(CAC) / 回本周期null没有公开获客成本(CAC)或销售效率披露。按细分市场索取获客成本、销售周期转化率和回本周期。

由于 Zhiyuan 不披露完整单位经济模型包,本表混合使用 Zhiyuan 直接证据和公开同业基准。

[CI008, CI009, CI010, CI011, CI012, CI017]
FI002: 单位经济性桥接图

已观察到的部署证据把机器人价格、劳动力替代和服务负担串到买方 ROI 上。

使用 Fulin 报道中的公开代理值,而不是公司计算的 ROI。

[CI008, CI009, CI010, CI016, CI017, CI018]

4.3 成本结构

成本结构更像资本密集的系统业务,而不像软件公司。Zhiyuan 的公开材料暗示,公司需要同时投入机器人硬件、具身模型研究、数据基础设施、部署团队和方案工程。同行证据让图景更清楚。Figure 的 BotQ 叙事、Apptronik 的制造和伙伴公告,以及 TrendForce 对规模驱动商业化的框架,都指向同一个机制:规模才能压低单位成本,但要达到规模,必须先砸下严肃的工厂、工具和组织投入。Fulin 的部署细节又加了一层成本,因为公司需要场景调试、安全验证和集成支持。营运资金也可能是隐藏压力点。企业和国资客户可能按验收周期付款,而不是订单宣布时付款,这会在利润率成熟前制造应收账款和库存压力。因此,应把这个品类建模为硬件 + 服务 + 数据,多重成本中心都会先于清晰收入能见度爬坡。[CI016, CI017, CI018, CI019, CI020, CI021]

FI004: 资本强度 / 现金流地图

早期商业化人形机器人公司的主要成本层和现金压力层。

基于 Zhiyuan 公开披露和 UBTECH 上市同行证据的定性地图。

[CI016, CI017, CI021, CI022, CI028, CI029]

4.4 资本充足性

公开证据强烈说明,人形机器人公司需要深厚资本池;但对 Zhiyuan 自身资产负债表的说明很弱。Apptronik 的巨额股权融资、Figure 的高产量制造推进,以及 UBTECH 作为上市公司的大额现金和借款披露,都说明严肃竞争者要同时资助硬件放量、软件迭代和部署支持。UBTECH 是最具体的基准:即便 2025 年收入超过 RMB2.0 billion、毛利率改善,它仍报告大额净亏损和有意义的经营现金消耗。这个基准重要,因为 Zhiyuan 还没有以类似细度披露现金、借款或应收账款余额。投资者目前无法直接测算现金续航期,应假设融资依赖会持续,直到商业收入和服务经济性披露得更清楚。下一轮融资的触发点很可能不只是「增长资本」,而是证明部署转化和安装后支持可以扩张,同时不吞掉不成比例的现金。[CI024, CI025, CI026, CI027, CI028, CI029]

资本充足性表
项目公开数值 / 状态置信度重要性尽调要求
Zhiyuan 手头现金null现金余额是衡量跑道的核心分母,但公司未公开披露。要求提供非受限现金、受限现金和月度烧钱额。
Zhiyuan 债务 / 项目融资义务null如果回款延后,债务会放大硬件风险。要求提供借款、担保、供应商融资和租赁义务。
Zhiyuan 外部股权支持2025 年融资记录强劲,且有战略投资方支持说明公司有融资渠道,但不能证明当前跑道足够。要求提供轮后现金衔接和资金用途计划。
同业流动性基准UBTECH 现金及现金等价物约 RMB4.89bn;借款约 RMB1.12bn说明规模化工业人形机器人运营需要怎样的资本结构。将 Zhiyuan 内部资产负债表与上市同业的流动性需求对标。
可能触发下一轮融资的因素在利润率和回款成熟前,公司仍要为制造、部署支持和模型 / 数据迭代提供资金说明融资决策应由商业化进展驱动,而不只是估值。要求提供董事会版下一轮融资触发条件和下行情景流动性方案。

资本充足性仍是公开资料中 Zhiyuan 承销判断最不透明的部分。

[CI028, CI029, CI030, CI032, CI033, CI034]
FI003: 财务估计区间

围绕 Zhiyuan 和该品类少数公开财务分母的区间视角。

混合 Zhiyuan 相关公开信号和上市同行基准;不是管理层预测。

[CI010, CI011, CI014, CI027, CI028, CI029]

4.5 财务结论

公开财务结论是有前景但仍不完整。Zhiyuan 看起来已经从概念期叙事跨入真实商业化,拥有点名企业订单、部署证据,以及比单一机器人销售更宽的变现表面。但公司仍很难用普通成长股权投资框架做投资判断,因为关键分母缺失:按型号实现的 ASP、按部署类型划分的收入、扣除服务负担后的毛利率、应收账款账龄、库存结构和月度现金消耗。公开同行申报和融资公告说明了这为什么重要。人形机器人里,收入可以很快上升,但亏损和资本需求仍然很大,尤其当支持、数据和制造成本并行扩张时。审慎结论是,Zhiyuan 正在形成可信收入引擎,但还没有形成公开市场质量的财务模型。下一轮尽调应少看头条估值,多看现金转化、利润率轨迹和可重复部署经济性。[CI003, CI004, CI014, CI027, CI028, CI032]

公开财务缺口表
缺失的私有指标影响具体尽调路径
按型号和场景拆分的月度收入用来区分试点、招标和可持续部署要求提供月度管理口径 P&L,并附合同和验收瀑布。
扣除服务负担后的毛利率用来检验部署支持是否侵蚀硬件经济性要求按部署类型提供贡献利润率,并计入现场服务人力。
现金、烧钱额和跑道用来评估资本充足性和融资紧迫性要求提供资金计划表、月度现金流和下行情景跑道测算。
应收账款和库存画像用来评估营运资本拖累和客户付款质量要求提供 AR 账龄、库存周转、减值政策和验收条款。
CAC、转化和续约指标用来判断项目销售能否变成可复制增长要求提供销售漏斗、胜率、部署转化、服务续约和分群留存。

这些不是锦上添花的指标;资本密集型硬件公司要做真正承销,至少需要这套材料。

[CI032, CI033, CI036, CI037]
Chapter 05

05产品与技术

5.1 产品概览

Zhiyuan 的公开产品组合已经足够宽,核心尽调任务不是发现,而是分类。官方公司和产品页面展示了面向工业、服务和表演场景的人形机器人平台;一个大场景清洁产品;以及数据或训练基础设施。这种宽度具有战略意义,因为它意味着公司试图掌握端到端工作流和数据闭环,而不只是机器人本体。G1 定位为工业 / 商业主力,主打有意义的操作能力和通道兼容;X2 强调拟人交互和服务式自主;A3 面向表演和多机器人编舞优化;C5 则把技术栈延伸到带自动维护站的清洁运营。即便不看模型层,产品故事已经是多形态。投资者不应像评估单一人形机器人演示公司那样评估 Zhiyuan;它已经选择成为系统家族公司,这同时提高上行空间和执行复杂度。[CE001, CE002, CE003, CE004, CE005, CE006]

产品模块 / 资产矩阵
模块 / 资产用户状态 / 成熟度差异化尽调缺口
G1 人形机器人工业 / 商业运营方公开产品页给出了具体工作流和数据主张面向工厂的操作能力、数据采集和平台视角需要正常运行时间、部署数量和实际客户组合。
X2 人形机器人服务 / 交互场景公开产品页功能丰富,但商业证明较弱多模态交互和仿人行为需要具名生产部署和支持指标。
A3 人形机器人舞台 / 表演 / 协同场景公开产品页声称续航和群控能力编舞和高曝光协同用例需要收入贡献和相邻商业场景的相关性。
C5 清洁系统物业 / 大型场馆运营公开产品页给出了工作流和工作站细节非人形工作流扩展和自动维护站需要装机量、ARR / 服务附加和利润率画像。
开源 X1 / GitHub 资产开发者 / 研究人员 / 合作伙伴公开文档和代码库可见开发者可信度和可检查的架构需要映射到商业支持和发布治理。
WITA / 模型层具身交互和推理栈公开披露了基准和模型定位将差异化上移到多模态推理需要生产指标、评测方法和部署相关性。

该矩阵覆盖已审阅来源中可见的主要公开产品和平台界面,不覆盖每个内部项目或配件。

[CE001, CE003, CE004, CE005, CE006, CE007]
工作流 / 用例表
用户任务当前工作流公司方案可衡量收益限制
工厂搬运 / 分拣人工重复搬运和分拣物料G1 / A2-W 式人形机器人或轮式部署,并做现场适配可能替代人工并连续运行稳定产出价值前,需要集成人力和场景调优。
前台 / 互动服务由人工员工或自助终端提供引导和互动X2 仿人多模态交互更丰富的互动和自主导航可靠性和经济性的公开证明有限。
舞台或活动协同人类表演者或定制机电设备A3 协同人形表演平台高曝光展示和多机控制活动之外的商业耐久性仍不清楚。
大面积设施清洁人工或传统清洁设备C5 清洁机器人,配工作站自动化降低维护负担,并把清洁工作流数字化需要公开证明部署规模和支持负担。
开发者 / 数据工作流内部机器人集成和实验开源 X1,加数据集和平台界面更快实验和生态参与开放资产不等于企业支持保证。

收益按已审阅来源中可推导的工作流结果表述,不是经审计的客户 KPI。

[CE003, CE005, CE006, CE007, CE011, CE014]
FE002: 客户工作流 / 运营流程

公开资料勾勒出 Zhiyuan 把工作流问题落地为运营部署的路径。

该流程把现场和产品证据压缩成一条客户旅程。

[CE003, CE007, CE014, CE015, CE016, CE022]

5.2 架构

最清楚的架构线索来自公司的开源和研究表面,而不只是打磨过的营销页面。X1 文档和 GitHub 仓库暴露出一个使用 AimRT 中间件、以强化学习训练和推理流程为中心的模块化栈;这个技术信号强于笼统的「AI 驱动」定位。AgiBot-World 和 G1 页面强化了数据运营的重要性;WITA 和 WAIC 材料则显示,公司也在向多模态和世界模型式具身推理的更高层推进。正在成形的架构看起来是分层的:机器人本体和末端执行器;感知、控制和中间件;模型和世界模型层;数据采集与训练;以及部署运营。Yicai 报道称,某次部署依赖约 95% 模拟数据和 5% 真实数据,这一点尤其有用,因为它说明 Zhiyuan 正在尝试连接实验室和现场。同一报道也说明,只有架构还不够:现场变量仍会迫使公司适配、处理安全并派驻工程。[CE009, CE010, CE011, CE012, CE013, CE014]

技术 / 运营架构表
层级 / 流程 / 组件角色依赖风险
机器人本体和移动系统在不同场景中执行物理任务机械可靠性、执行器、电池和传感器硬件复杂度可能跑在服务能力前面。
感知 / 控制 / 中间件将输入转成协同行动AimRT 式中间件、规划和控制逻辑集成缺陷或延迟会削弱工作流信任。
基础 / 世界 / 交互模型任务理解、推理和多模态行为训练数据、算力、基准和持续迭代基准表现可能领先于部署鲁棒性。
数据采集和训练栈采集、标注并回放真实或模拟数据数据集治理、仿真器质量、云工具数据治理薄弱或仿真错配会伤害现场表现。
部署运营场景适配、服务和维护集成商、专家、客户现场人力密集度和支持负担会压缩利润率。

该架构根据公开技术和部署证据重建,并非来自内部工程图。

[CE009, CE010, CE012, CE013, CE014, CE017]
FE001: 产品架构图

从机器人本体到部署运营,公开信息可推断出的架构层。

栈是对已审阅官方文档、GitHub 资产和一线报道的综合,不是公司发布的模块图。

[CE009, CE010, CE012, CE013, CE017, CE018]

5.3 差异化

Zhiyuan 的差异化并不主要落在某一项规格上,而是落在宽度、开放性和生态搭建之间的互动上。开源 X1 资产让开发者和研究者更容易读懂公司,比许多纯营销型同行更透明。APC 2026 和 WAIC 2026 显示,公司有意把这种技术姿态转成伙伴网络;WITA 则暗示公司在模型层也有野心,而不只是停留在本体层。但这些优势都有条件。Figure、UBTECH、Unitree 和 Apptronik 都在营销集成式人形机器人栈;Figure 近期帖子也显示,竞争前沿同样围绕数据、预训练和工作流学习展开。Gasgoo 关于灵巧手剥离的报道,是这里最有用的反向信号:它提醒投资者,全栈范围可能在可靠性和服务执行最关键的时点制造组织碎片化。Zhiyuan 的差异化真实存在,但仍需要转化为可支撑的部署,而不只是更多模块和公告。[CE021, CE022, CE023, CE024, CE031, CE032]

FE003: 关键依赖图

Zhiyuan 的技术栈需要数据、模型、硬件和伙伴运营相互加固。

依赖关系基于已审阅的公开证据推断。

[CE013, CE021, CE022, CE031, CE034, CE036]

5.4 信任与安全

公开来源中的信任证据有正有负。积极一面,一篇现场报道显示,当有人接近工厂地面上的机器人时,机器人会立即停止;中国标准和政策环境也在走向更明确的人形机器人安全、应用和互操作预期。这些都是有用锚点。但本报告审阅的公开记录,对企业买方想看的具体控制披露很薄:正常运行时间历史、正式事故率、已发布 SLA、网络安全架构、环境数据的隐私处理,以及点名认证都披露不足。这不是学术问题。具身系统运行在真实工作空间中,采集运营数据,并且越来越依赖云连接的模型和工具层。因此,公司的技术可信度走在公开信任披露前面。正确尽调姿态,是把政策和基准进展视为顺风,同时在假设企业级准备度之前,要求一套具体得多的控制包。[CE015, CE016, CE025, CE026, CE027, CE028]

信任 / 质量 / 合规表
控制 / 认证 / 质量指标状态范围缺口
工厂现场报道中的即时停止行为已观察到某次部署中的运营安全行为需要更完整的安全论证文档、险情数据和政策。
中国人形机器人标准和政策框架制定中 / 外部行业层面的设计和部署预期需要公司级映射,说明内部控制和合规责任人。
公开可靠性指标未披露应覆盖正常运行时间、MTBF、故障率和事故趋势企业就绪度的重大尽调缺口。
网络安全 / 隐私控制包未披露应覆盖遥测、客户数据、访问控制和留存具身 AI 部署的重大尽调缺口。
具名认证 / 审计 / SLA 条款已审阅来源未明确披露可证明企业级质量流程大型企业推广前的重大尽调缺口。

最重要的信任条目仍近似空白,因为已审阅的公开记录没有给出完整控制包。

[CE015, CE016, CE025, CE026, CE027, CE028]
FE004: 产品成熟度 / 能力图

按主要产品和平台触点梳理的公开成熟度视图。

这是基于公开证据的定性评估,不是内部就绪度计分卡。

[CE003, CE005, CE006, CE007, CE010, CE018]

5.5 路线图

公开路线图信号明显指向扩张。近期发布把新产品页面、开源资产、基准模型公告、全球伙伴会议和实体 AI 思想影响力活动组合在一起。APC 2026 暗示地理和渠道扩张;WAIC 2026 暗示公司想影响技术议程;WITA 加上开源栈,则暗示模型和开发者可信度会继续投入。如果每个新层都让部署更容易、数据更丰富、伙伴采用更快,飞轮就会很强。若组织扩张速度超过可靠性、可支持性和治理成熟度,也会稀释焦点。尽调时应把路线图解读为方向大胆且连贯,但服务、安全和正常运行时间披露仍不足,尚未把宽泛野心转成企业信心。[CE013, CE019, CE021, CE034, CE035, CE036]

路线图 / 发布 / 开发阶段表
日期 / 阶段功能 / 里程碑状态含义来源
2025开源 X1 文档和代码库已公开发布释放开发者界面野心和可检查架构信号AGIBOT 文档 / GitHub
2025-2026WITA 基准和多模态定位已公开宣布暗示栈向具身交互模型扩展AGIBOT WITA 发布
2026-04APC 2026 合作伙伴大会已完成展示生态建设和国际合作伙伴推进APC 2026 概览
2026-07WAIC 物理 AI 论坛已完成将公司放进世界模型 / 物理 AI 叙事中WAIC 论坛发布
持续多形态产品扩展官网可见路线图在硬件、软件和数据界面上持续加层公司 / 产品页面

路线图根据公开发布重建,可能遗漏内部优先事项或已取消项目。

[CE013, CE020, CE021, CE034, CE035, CE036]
Chapter 06

06客户

6.1 客户分层

按工作流而不是只按行业标签,最容易理解 Zhiyuan 的公开客户地图。最强的可见细分是结构化工业制造,机器人可以插入分拣、搬运和检测任务,并有可衡量的劳动力或吞吐逻辑。电信或战略采购是第二个可见桶,China Mobile 更像掌握预算的买方,而不只是演示东道主。第三个桶是伙伴主导的商业化,APC 2026 以及英国、ANZ 后续活动暗示,本地伙伴可能帮助把技术转成部署。公司信息中也出现服务、旅游和交互场景,但第三方客户证据轻得多。这意味着 Zhiyuan 的客户基础并未在各细分之间均衡。目前权重更偏向那些能在结构化企业式工作流中为机器人找到理由的买方,这比消费者炒作更健康,但如果只有少数这类买方真正放量,集中度风险仍会留下。[CU001, CU002, CU003, CU008, CU015, CU017]

客户分层表
客群买方 / 预算负责人工作流公开证明质量含义
工业制造工厂运营方 / 工厂管理层分拣、搬运、检测、产线旁任务最适合近期规模化部署的客群。
电信 / 战略采购大型机构 / 采购团队机器人采购和场景验证中高显示预算意愿,但尚未证明复用经济性。
消费电子制造OEM / 代工厂精密制造和具身 AI 产线部署将证明扩展到汽车相邻工厂之外。
服务 / 交互 / 文旅场馆运营方或推广方引导、交互或展示工作流中低在公司叙事中可见,但第三方客户证据较薄。
海外渠道 / 合作伙伴生态分销商 / 本地合作伙伴本地化、演示和市场进入支持是有用的扩张路径,但付费客户证明仍早期。

客群按价值如何购买和部署来分组,而不只按 NAICS 式行业标签。

[CU001, CU002, CU003, CU008, CU017, CU018]
FU001: 客户旅程图

公开资料勾勒出一条客户旅程:先选场景,再部署、扩张,并借伙伴放大规模。

该旅程由公开部署报道和伙伴大会材料综合而来。

[CU002, CU006, CU007, CU015, CU027]

6.2 采用轨迹

采用故事仍早,但已经不再只是概念。最清楚的公开弧线,是从首次现场部署,到更大的追加承诺,再到更广的商业化叙事。Fulin Precision 是最佳单一案例:报道先描述四台机器人正在主动分拣配送,后来又把整体交易描述为接近 100 台机器人。China Mobile 展示了另一种采用模式:大型机构愿意以显著规模运行正式招标。Longcheer 把图景拓宽到消费电子精密制造;APC 2026 则把公司自己的「大规模部署已经开始」宣言正式化。重要解释不是 Zhiyuan 已经解决了采用问题,而是公司已进入一个阶段:公开证据能看到买方、站点和铺开路径。这距离广泛市场渗透仍有距离,但实质上已超越纯原型剧场。[CU004, CU005, CU006, CU007, CU009, CU010]

客户增长 / 采用轨迹表
阶段公开证据状态重要性限制
初始现场工厂部署四台 A2-W 机器人在 Fulin Precision 运行已观察到证明真实工作流使用,而不只是演示阶段。单个现场不能证明广泛装机量。
工厂扩张信号Fulin 更大交易中接近 100 台机器人已宣布迄今最好的公开扩张代理指标。未披露交付情况或长期利用率曲线。
大型机构采购China Mobile 招标价值 CNY124m,其中 CNY78m 用于 AgiBot 全尺寸人形机器人已观察到证明大型买方愿意分配预算。招标金额不等于经常性收入或续约。
新增制造业客户标识Longcheer 产线部署文章已宣布 / 已报道将证明扩展到电子精密制造。需要独立、由客户发布的佐证和数量细节。
合作伙伴驱动的商业化APC 2026「部署年」信息,加 UK / ANZ 合作伙伴大会已观察到显示分销策略正从国内直销向外扩展。合作伙伴不等于付费终端客户。

轨迹条目混合了现场部署、订单扩张、采购和合作伙伴商业化信号;每类信号的权重不同。

[CU004, CU005, CU006, CU008, CU012, CU014]
FU002: 采用 / 部署漏斗

公开证据从广泛伙伴兴趣迅速收窄到少数具名且有规模的客户案例。

数值混合了公开资料中的伙伴 / 活动数据和客户证明数量;它反映证据形态,不是字面 CRM 漏斗。

[CU014, CU015, CU016, CU034]

6.3 点名客户证据

点名客户证据是本章最强的一部分,也是最需要谨慎的地方之一。Fulin Precision 和 China Mobile 是高价值客户标识,因为公开来源不只是泛泛点名,还描述了现场任务、单位数量或合同价值。Tencent News 把 Chery 加入证据集,Longcheer 文章则把故事延伸到消费电子制造。但这个簇仍然很窄。大部分公开案例依赖四个点名客户标识加上一条大型生态伙伴叙事。以年轻人形机器人公司标准看,这足以支撑真实客户采用论点,尤其是在早期;但还不足以宣布多元化或可重复性已经解决。同行案例在这里有启发:Figure 受益于 BMW 自己的新闻稿,Apptronik 和 UBTECH 也会公开最适合自己工作流的窄切片。衡量 Zhiyuan 的正确标准不是完美,而是在多个细分中出现更频繁、由客户或伙伴自己写出的佐证。[CU003, CU004, CU008, CU010, CU011, CU012]

具名客户证明表
客户 / 标识公开证据工作流佐证质量未解缺口
Fulin PrecisionYicai、Assembly 和上海政府报道工厂分拣和搬运;接近 100 台机器人交易信号需要交付完成情况、利用率和续约数据。
China MobileYicai 加 Tencent News 背景大规模采购人形机器人需要落地节奏和中标后的部署结果。
CheryTencent News 提及汽车相邻企业订单需要客户或合作伙伴直接发布工作流细节。
Longcheer TechnologyAI Journal 部署文章消费电子精密制造产线需要客户亲自确认和部署规模。
海外合作伙伴(UK / ANZ APC)官方合作伙伴大会发布区域商业化和渠道建设中低需要确切付费客户数量和具名终端客户标识。

对一家年轻人形机器人创业公司来说,具名证明方向上很强,但仍窄于成熟企业机器人供应商的客户台账。

[CU004, CU008, CU011, CU012, CU015, CU016]
FU003: 客户证明矩阵

按细分场景和证明深度看,Zhiyuan 当前客户证明最强的位置。

仅为基于公开证据的定性解读。

[CU003, CU004, CU008, CU012, CU017, CU019]

6.4 留存与耐久性

公开留存证据很薄,投资者需要谨慎使用代理指标。最强代理指标是 Fulin Precision 的后续扩张:从初始活跃机队走向更大的已宣布订单,这不等于 NRR,但好过一次性展示。China Mobile 招标也是有用代理指标,因为正式采购意味着门槛高于随意试用。即便如此,公开记录没有给出流失、续约、SLA 或正常运行时间分母。因此,耐久性仍是推断,而不是已披露指标。投资者应区分三件很不同的事:机器人出现在现场、客户为更多设备承诺预算、客户实现足够价值并反复续约或扩张。Zhiyuan 对前两类有公开证据,但第三类还没有达到可严谨建模的程度。[CU007, CU010, CU021, CU022, CU023, CU025]

留存 / 重复使用 / 满意度表
代理指标公开来源显示的内容解读缺口
Fulin 追加订单初始现场使用,加接近 100 台机器人的更大交易扩张和潜在满意度的最佳代理指标需要随时间变化的活跃机器人数量和实际续约行为。
China Mobile 招标流程正式采购规模和中标份额比非正式展示兴趣更能说明问题需要实际部署、使用量和二阶需求证据。
已点名披露的新增客户标识Chery 和 Longcheer 拓宽客户面说明采用并非只靠单一客户需要收入拆分和多站点转化数据。
公开留存指标未披露没有直接的 NRR、流失率或客户留存证据关键尽调阻碍。
公开服务 / SLA 指标未披露缺少硬性的耐用性或支持分母关键尽调阻碍。

本表有意把真正的留存指标与较弱但仍有参考价值的公开代理信号分开。

[CU007, CU021, CU022, CU023, CU025, CU026]
FU004: 留存 / 重复采购队列

公开留存证据主要靠代理信号,而不是指标本身。

各行代表客户原型,不是实际内部队列。

[CU007, CU021, CU022, CU023, CU025, CU026]

6.5 扩张与风险

扩张论点可信,但风险画像仍高。从地理看,公司在中国的强度远高于海外;伦敦和墨尔本伙伴会议是有用信号,但不是披露的海外客户数量。从细分看,结构化工业和采购工作流显然领先于更软的服务和交互用例。集中度是最大的投资判断风险,因为少数客户标识就能驱动早期硬件公司的大部分叙事动能。另一个风险是分析过度延展:出货领先和强客户标识容易让投资者假设留存、多元化和客户经济性已经得到证明。The Wire China 更怀疑的框架在这里是有用制衡。Zhiyuan 已跨过可信早期企业采用的门槛,但在决定成长轮投资前,投资者仍需要活跃客户数、扩张率、集中度数据,以及按客户类型拆分的服务负担视图。[CU016, CU017, CU018, CU019, CU020, CU024]

扩张与集中度风险表
风险为何重要当前公开信号尽调要求
客户集中度叙事权重可能过度压在少数早期客户上要求提供前 10 大客户收入占比和活跃机器人占比。
订单到使用转化风险公告可能跑在实际部署深度前面中高要求提供中标到安装、安装到扩张的漏斗数据。
服务负担风险客户扩张会挤压支持资源中高要求按客户群提供现场服务人员配置和问题积压。
海外变现风险伙伴动作可能超过付费客户现实要求提供海外付费客户数量和服务覆盖图。
细分市场平衡风险软收益场景的变现可能慢于工厂要求按工作流提供收入结构,并按细分市场提供毛利率。

风险从客户组合角度框定,而不是泛泛谈公司运营。

[CU017, CU018, CU019, CU024, CU033, CU035]
Chapter 07

07风险

7.1 风险概览

Zhiyuan 的风险图景不是一个单独的「技术风险」框,而是横跨监管、运营、伙伴和资本的分层投资判断问题。中国政策整体支持机器人和具身 AI,但同一股政策动能也在提高安全、测试、网络安全和治理标准。公开现场证据显示,公司已经进入真实工厂环境,这在战略上是积极的,但也意味着错误现在会带来物理、商业和法律后果。与此同时,商业化模式依赖一小簇点名客户和不断扩宽的伙伴网络,资产负债表仍不透明。因此,正确的风险框架是双面的:相比许多早期人形机器人初创公司,Zhiyuan 有更多真实世界牵引力;但牵引力提高了合规、服务纪律和现金管理的重要性,而不是降低这些要求。[CR001, CR002, CR003, CR004, CR012, CR024]

FR001: 风险热力图

基于公开证据,按影响和发生可能性梳理最重要的风险簇。

单元格使用从已审阅公开证据综合出的定性评分,不是精算损失历史。

[CR001, CR007, CR010, CR014, CR024, CR027]

7.2 监管与法律风险

监管和法律风险栈已经具有实质性。人形机器人部署位于产品安全、工作场所互动、数据采集和日益政治化的技术贸易交汇处。中国机器人和人形机器人指引文件,以及正在形成的标准体系,为部署给出更清晰路径,但也抬高了安全和文档门槛。产品责任规则重要,因为机器人现在会在物理工作空间中行动;法律评论强调,制造商、运营方和软件层都可能分担责任。数据安全和隐私风险同样重要:具身系统会在客户设施内采集环境数据,因此《数据安全法》、2026 年《网络安全法》修订以及更广的隐私框架,都不是抽象的合规背景。地缘政治供应风险又加上一层法律约束,因为出口管制和产业安全规则可能在预警有限的情况下扰乱组件、客户或海外交付义务。[CR002, CR003, CR004, CR005, CR006, CR007]

法规 / 法律风险登记表
风险为何重要严重性公开信号尽调要求
产品责任与问责机器人在物理空间中行动,可能造成人身伤害或运营损失法律指引强调制造商、运营方和软件层共同承担责任要求提供责任划分、保险安排和事故升级政策。
数据安全与隐私合规具身系统会采集环境和运营数据DSL、CSL 修订和隐私指南抬高合规义务要求提供数据地图、留存政策、访问控制和跨境规则。
人形机器人标准符合性新兴标准会影响市场准入和客户信任中高MIIT / SESEC 标准化推进势头可见要求提供内部标准路线图和测试证据。
出口管制 / 产业安全扰动规则可能限制组件、客户或跨境交付MOFCOM 和法律提示显示贸易管制变化很快要求提供出口管制筛查和替代采购计划。
未披露诉讼 / 处罚历史公开静默不等于不存在中高未找到完整公开清单要求提供诉讼、处罚和召回清单。

本登记表强调外部合规和法律暴露,而不是单纯产品缺陷。

[CR002, CR005, CR007, CR008, CR009, CR010]
FR002: 风险传导图

上游合规和执行风险如何传导到客户、利润率和融资结果。

传导链条基于公开证据和法律框架推断。

[CR007, CR010, CR016, CR024, CR028, CR032]

7.3 运营风险

机器人离开实验室的那一刻,运营风险就变成现实;Zhiyuan 已经越过这个点。Fulin 现场报道很有说明性,因为它同时展示了进展和暴露:机器人能搬箱子、在人接近时停止,并在真实工业环境中工作;但这需要维护专长、按环境调参,以及持续算法改进。公司显然大量依赖模拟数据,再补充较少真实世界数据,这对规模化是理性的,但在陌生环境中的泛化、边缘案例和安全上仍留下残余风险。更批判的公开报道在这里有价值,因为它能压住把每次部署都读成「已经完全解决」的冲动。实际风险不是机器人毫无用处,而是服务负担、可靠性波动和特定场景工程,可能比利润率或客户耐心扩张得更快。[CR012, CR013, CR014, CR015, CR016, CR017]

运营 / 质量 / 安全风险登记表
风险信号严重性为何重要尽调要求
近人作业安全风险工厂报道显示机器人会主动停止,也处在繁忙环境中安全设计必须在真实工业场景里站住,不只是演示要求提供危险分析、事故和险兆日志。
模型泛化风险训练偏重仿真,现场变量仍然存在意外环境会制造故障或服务负担要求提供故障分类和真实世界再训练闭环指标。
服务和维护负担运维需要熟练人员和部署适配现场服务负荷会侵蚀客户经济性和可扩展性要求按客群提供 MTTR、支持人员配置和问题积压。
网络安全与遥测风险机器人和云工具会形成数据攻击面中高安全薄弱点可能影响客户和监管方要求提供安全架构、渗透测试历史和响应手册。
工作流可靠性风险工业客户要求任务执行和正常运行时间保持稳定表现不稳会很快打掉标杆客户要求提供正常运行时间、任务成功率和升级流程。

多数关键运营指标未公开,因此本表区分已观察到的风险和缺失的控制证据。

[CR012, CR013, CR014, CR015, CR016, CR017]

7.4 伙伴与依赖风险

Zhiyuan 的规模化路径本来就富含伙伴,因此形成了独特依赖画像。一方面,集成商和渠道伙伴有用,因为它们可以加快本地化、站点适配和国际触达。另一方面,它们也引入质量控制、责任归属和品牌一致性风险。客户集中度也是如此:少数高曝光客户标识可以加速转化,同时如果任何旗舰客户不及预期,也会放大暴露。地缘政治和组件依赖同样重要,因为人形机器人栈不可避免地依赖高规格硬件和跨境生态,而这些会受到出口规则或产业安全监管的压力。这些不是假设性的二阶问题;它们直接约束交付可靠性、毛利率和扩张计划可信度。[CR019, CR020, CR021, CR022, CR023, CR024]

合作伙伴 / 依赖风险登记表
依赖项风险严重性为何重要尽调要求
集成商和部署专家质量不一或铺开变慢部署不完全由内部掌控,依赖伙伴执行要求提供伙伴质量审核流程和升级责任归属。
渠道 / 区域伙伴海外品牌一致性不足或服务覆盖弱中高伙伴主导的扩张可能跑在运营控制前面要求提供伙伴认证和区域支持地图。
已披露旗舰客户集中度和标杆客户风险少数客户可能承担不成比例的叙事和收入权重要求提供头部客户集中度和续约状态。
关键组件 / 供应商交期或采购冲击出口管制和单一来源部件可能拖累交付要求提供 BOM 集中度和双供应源准备度。
跨境政策环境贸易或采购限制中高政策变化可能意外堵住市场或投入品要求提供分司法辖区合规手册和应急情景。

依赖风险既是商业问题,也是技术问题:最弱节点会拖慢整套铺开引擎。

[CR019, CR020, CR021, CR022, CR023, CR024]
FR003: 依赖图

Zhiyuan 放大规模,要靠客户、伙伴、部件和政策条件织成的紧耦合网络。

依赖节点由公开部署、伙伴和法律证据综合而来。

[CR019, CR020, CR021, CR022, CR023, CR024]

7.5 财务风险

财务风险仍然偏高,主要因为公开记录不对称:野心和部署信号多,现金充足性和单位经济性信号少。UBTECH 年报在这里有用,因为它证明,即便一家规模更大、已经上市的人形机器人同行收入增长、毛利率改善,也仍会消耗大量现金并持续亏损。这说明品类资本密集度是结构性的,而不是暂时现象。Zhiyuan 的具体问题是缺少直接披露。投资者没有公开现金、债务、财务约束条款或应收账款数据,无法压力测试下行场景。因此,TrendForce 和 IDC 的出货领先叙事应读作市场地位信号,而不是资产负债表安慰剂。一家公司可以在单位数上领先,却在流动性韧性上落后。[CR027, CR028, CR029, CR030, CR031]

人员 / 执行风险登记表
风险公开证据严重性为何重要尽调要求
产品线范围蔓延可见产品组合很宽,模块也在拆分孵化在服务纪律最关键时,过宽布局会削弱焦点要求提供组织架构、P&L 归属和路线图优先级。
部署团队扩张明确需要现场专家客户成功靠执行产能,不只靠硬件质量要求按部署和支持岗位提供人员数。
商业化纪律订单和试点可能跑在可复制流程前面中高流程控制弱会伤害利润率和标杆客户要求提供从试点到规模化铺开的阶段门标准。
管理层对旗舰叙事的依赖品类热度会推高激进扩张压力叙事驱动扩张会扭曲运营决策要求提供董事会层面的风险治理和 KPI 看板。
财务运营成熟度虽有规模化野心,但未公开资金管理细节流动性失误会很快变成生死问题要求提供预算节奏、契约监控和资金管理控制。

人员和执行风险叠加了组织设计、商业化流程和财务纪律。

[CR018, CR027, CR028, CR031, CR039]

7.6 缓释措施

公开缓释图景令人鼓舞但不完整。正向信号包括:现场基本安全行为证据、愿意参与正在形成的标准环境、公开承认部署学习很重要,以及能拓宽市场触达的生态策略。但最强的缓释仍然留在私域:经审计控制包、事故日志、支持 KPI、供应商应急方案和流动性纪律。正确的尽调回应不是假设公司暴露过度,而是在投入新资本前设下硬证据门槛。如果 Zhiyuan 不能证明旗舰订单可以转成稳定现场部署,不能产出可辩护的治理和数据控制,或不能显示服务负担可持续扩张,那么即便头条市场叙事仍然火热,这些都应成为否决标准。[CR032, CR033, CR034, CR035, CR036, CR037]

缓释措施与否决条件表
主题可见缓释剩余风险否决条件所需证据
安全与运营已观察到停止行为,加上部署学习没有事故和正常运行时间数据时仍然偏高任何未解决的严重安全事故,或旗舰站点反复任务失败事故日志、危险评审、正常运行时间历史。
数据与法律合规中国政策和法律框架可梳理没有公司控制包时仍为中高更大范围铺开前,拿不出可审计的隐私 / 安全 / 责任控制安全架构、隐私地图、法律问责矩阵。
客户集中度已披露客户和招标证明需求客户组合仍窄时风险仍高一两个旗舰客户流失或不扩张,且没有替代管线头部客户集中度和客群扩张数据。
供应商与政策依赖管理层可以制定应急方案如果单一来源或出口暴露组件占主导,风险仍高关键组件没有双供应源或应急路径BOM 集中度和替代采购证明。
流动性纪律同行证据提供警示性参照基准Zhiyuan 现金和现金消耗未公开时,风险仍高无法在下行情景下证明 18+ 个月资金续航期现金、债务、资金续航期和下行情景融资计划。

否决条件刻意以证据为准,而不是叙事为准。

[CR032, CR033, CR034, CR035, CR036, CR037]
Chapter 08

08估值

8.1 投资论点

Zhiyuan Robot 值得认真进入投资讨论:公开证据已经越过许多人形机器人创业公司从未达到的门槛——可识别的出货、已点名客户、多条公开产品线,以及围绕具身 AI 搭生态。最强的正面逻辑有三点。第一,外部分析师和公司材料都支撑它在中国出货叙事中的领先位置。第二,具名客户验证已经出现在工厂和采购场景里,不只是舞台演示。第三,公司显然不想停在单场演示,而是在用产品、模型、数据和伙伴拼平台深度。上述优势并不轻。与此同时,估值讨论不能止步于「品类龙头」。投资论点必须同时承受公开经济数据薄、集中度风险和硬件资本强度高这几件事。因此,本章最后给出建设性但克制的判断:公司值得密切跟踪,但按市场喊出的估值去下高确定性判断,还不容易。[CV001, CV002, CV003, CV008, CV013, CV015]

投资论点 / 反方论点表
立场要点证据质量为何重要
投资论点在这个品类里,出货和商业化势头异常强中高支撑真正的平台潜力,而不是纯概念价值。
投资论点工厂和采购环节已有具名客户证据显示真实买方参与,并已嵌入工作流。
投资论点产品、模型和生态的宽栈可能配得上平台溢价可支撑超越单一机器人本体的上行空间。
反方论点收入和利润率分母仍披露不足在严肃测算下,估值会变脆弱。
反方论点客户证明仍集中,资本强度仍是结构性问题中高如果旗舰客户停滞或支持成本膨胀,下行风险会上升。

反方论点不是理论推演;它直接来自公开来源仍未披露的信息。

[CV001, CV002, CV003, CV013, CV015, CV034]
FV001: 建议逻辑

建议来自两股力量的抵消:品类和证明信号很强,但公开经济性弱、风险高。

该流程压缩了本章推理链,并非完整投资备忘录。

[CV001, CV002, CV003, CV015, CV016, CV040]

8.2 投资建议

当前建议是「跟踪」。这不是模糊避险,而是承认真实潜力和证据不足之间存在明确错位。Zhiyuan 的可信度已经高到不能忽视,尤其是出货位置和客户动能都在加强。但公开记录在收入分母、毛利率、服务负担和资产负债表强度上仍然太薄,还不足以支撑「几乎不问价格、现在就投」的高确信度姿态。投资人还要记住,私有人形机器人估值正处在叙事浓度很高的市场里,全球同行可以靠前沿模型兴奋感和战略可选项交易,而不是靠当期经济性。在这种环境下,纪律比速度重要。跟踪 Zhiyuan,同时强力推动披露,可能比急着为经济性仍不透明的估值背书,创造更好的入场点。[CV004, CV005, CV006, CV007, CV014, CV015]

建议摘要表
字段当前判断原因含义
建议跟踪可信的品类领导者,但公开经济性不完整保持跟进,但推迟激进定价假设。
置信度证据不是空的,但分母仍不足如果收入和利润率披露改善,快速更新。
估值立场偏贵公开估值作为叙事锚点可以支撑,但相对已披露基本面偏贵要求按里程碑测算,而不是只按叙事定价。
风险评级资本强度、集中度和披露缺口仍然重要提高信念前,要求硬尽调。
下一步深入尽调管理口径账目和客群披露可能显著改变建议直接访问数据室后重新测算。

本摘要刻意区分公司吸引力与当前公开估值语境的吸引力。

[CV001, CV015, CV016, CV040]

8.3 估值背景

Zhiyuan 当前公开估值只能靠不完美但仍有用的锚点交叉校准。最可辩护的当前估值,是 2025 年媒体报道中的约 RMB 150 亿,约 US$2.07B。绝对值看,价格不便宜;但相较 Figure 等美国最大私营同行叙事,以及 Apptronik 周边的溢价融资背景,它又明显更低。真正的问题不是 Zhiyuan 是否比 Figure 便宜,而是在经济披露薄得多的前提下,相对溢价同行的折价是否足够大。UBTECH 提供了关键公开基准,因为它把规模化品类经济性摊开了:有真实收入、毛利率改善,但仍然大幅亏损、现金消耗重。UBTECH 的公开基准说明,在 Zhiyuan 披露更多收入和毛利引擎之前,应先把它的私募估值视为偏贵。[CV004, CV005, CV006, CV007, CV017, CV018]

FV003: 估值 / 回报区间

当前视角下乐观、基准和悲观结果的示意估值区间。

区间是基于公开市场位置、风险和同业背景的分析估算,不是市场价格。

[CV004, CV006, CV023, CV024, CV025, CV026]

8.4 情景

情景框架应围绕里程碑,而不是围绕可比公司倍数打转。牛市情景里,出货领先变成可持续的商业领先:客户案例持续叠加,海外伙伴活动转成付费账户,经济性披露足以支撑平台溢价。基准情景里,商业化继续推进,但公司仍只做到部分透明,投资人只能把它当作有前景但风险未消的成长故事估值。熊市情景里,订单转化慢于预期,支持和部署成本居高不下,未来融资发生在证据足以支撑当前叙事之前。相比点估值,更应强调情景,原因很简单:公开数据包还不足以支撑虚假的精确,但已经足以界定从今天往后,好的路径依赖和坏的路径依赖会长什么样。[CV023, CV024, CV025, CV026, CV027, CV028]

乐观 / 基准 / 悲观场景表
场景核心假设示例估值区间(USD bn)建议含义关键里程碑
乐观出货领先转化为企业客户重复扩张、更清晰经济性和海外伙伴变现2.8-4.0披露显著改善时选择性投资多个旗舰客户扩张,收入 / 利润率桥接变得可信。
基准商业势头延续,但经济性仍只部分披露1.6-2.2跟踪,等待证据当前估值仍可辩护,但并不明显便宜。
悲观订单转化慢、服务负担居高不下,在强经济性出现前又启动一轮融资0.8-1.3避免追高旗舰订单没能转为持久标杆部署。

区间是分析估算,不是市场事实;用途是在公开证据下框定投资测算结果。

[CV023, CV024, CV025, CV026, CV027, CV028]
FV002: 估值敏感性

当前估值信心对经济性披露和标杆部署转化最敏感。

条形使用 1-5 的相对敏感性评分,来自公开证据包。

[CV010, CV018, CV023, CV029, CV034]

8.5 可比公司

可比公司只有在克制使用时才有价值。Figure 应被看作溢价叙事和前沿模型基准;Apptronik 是资本强度和企业伙伴基准;UBTECH 是最清晰的公开运营基准。Unitree 提供硬件定价背景,但不是披露完整的公开估值案例。因此,直接算「同行平均倍数」弊大于利:可比组混合了上市公司财务披露、私募巨额融资叙事,以及部署透明度差异很大的公司。更好的用法,是比较每条叙事当前背后有多少证据、披露和资本。按这个标准,Zhiyuan 强于单纯概念项目,但还弱于那种没有毛利率或收入分母也配得上干净溢价的公司。[CV017, CV018, CV019, CV020, CV021, CV022]

可比估值表
可比对象为何重要公开锚点估值用途局限
UBTECH最接近的中国人形机器人公开运营参照基准2025 年年报及 HKEX 披露的公司更新最适合判断已披露经济性和资本强度公开市场倍数不能直接套用到非上市 Zhiyuan。
Figure AI全球前沿模型 / 旗舰客户溢价参照基准$39B Series C 估值可用于上限叙事语境其溢价叙事远厚于 Zhiyuan 当前披露集。
Apptronik资本强度和企业伙伴参照基准$5B 估值 / >$935M Series A 语境可用于判断融资胃口和战略客户信号仍为非上市公司,不是干净的经营倍数可比。
Unitree硬件透明度和定价语境公开产品和定价线索可用于硬件市场语境留存来源中没有完整披露估值,不能作为完整可比对象。
Zhiyuan 当前估值2025 私人融资公开媒体锚点~$2.07B / RMB15B当前投资测算讨论的参考点轮次条款和分母仍披露不足。

可比对象有意选得不完全同质;重点是做情境三角校验,而不是算一个假精确的平均倍数。

[CV004, CV005, CV006, CV017, CV018, CV019]

8.6 退出与尽调

退出可选项存在,但今天不应高估。公开报道讨论过最终赴港的路径,UBTECH 等同行也说明,中国人形机器人公司走向公开市场并非不可想象。战略并购或后续交叉轮也有可能。但任何路径都不能用来替代缺失的基本面。在提高确信度之前,投资人应要求一组最能改变投资判断的事实:实际收入、扣除服务负担后的毛利率、头部客户集中度、流动性续航,以及旗舰订单正在变成可复制部署而非孤立新闻的证据。尽调要求不是流程动作。它们决定投资人买到的是强前沿资产,还是包装精美的不确定性。[CV030, CV031, CV032, CV034, CV035, CV036]

投资逻辑破裂与止损触发因素表
触发因素为何打破投资逻辑当前可见度立即动作
旗舰订单未能放大为真实且持久的部署会直接削弱商业化叙事暂停投资,要求提供部署队列证据。
下一轮融资窗口已推进,收入 / 毛利率披露仍缺位估值几乎只能靠叙事支撑没有分母证据,就拒绝溢价定价。
支持负担或事故显著上升会同时伤到毛利率和标杆客户质量低-中要求审查服务 KPI 和事故日志。
下一轮融资在证据薄弱或重度稀释下完成说明经济性跟不上故事重算下行情景下的持股比例和资金需求。
数据 / 法律治理失灵拖慢企业采用会同时打击客户信任和退出选择立即升级法律与安全尽调。

这些触发因素划出边界:故事再有潜力,到了这些点就不该再享受估值上的宽容。

[CV029, CV033, CV034, CV035]
最终尽调索取清单表
尽调索取项为何重要会改变什么?
月度收入与毛利率桥接表补上所有严肃估值方法缺失的分母如果数据强,立场可能从偏高转为合理。
客户队列 / 扩张表区分一次性客户标识和持续采用可能显著提高推荐置信度。
支持与服务负担指标检验部署能否以合算成本放大可能收窄或拉宽合理估值区间。
现金、烧钱速度、债务和现金跑道资料包厘清稀释和融资风险可能实质改变悲观情景的严重程度。
2025 轮融资条款和股权结构表决定下行情景经济性和持股路径可能重塑对公开估值锚点的解读。

这些是最可能改变投资决定的最低尽调索取项,不是穷尽清单。

[CV034, CV035, CV036, CV037, CV040]
FV004: 投资 KPI

投委会视角下的浓缩判断:Zhiyuan 当前哪里得分高,哪里仍缺证明。

定性 KPI 面板,不是量化计分卡。

[CV002, CV007, CV011, CV013, CV015, CV016]

免责声明

本报告基于截至 2026-08-01 的公开信息生成,仅用于尽调研究目的,不构成投资建议。关于私营公司估值、融资、客户和合同的结论,应以一手尽调材料核验。

证据索引

结论
编号陈述可信度来源
CO001 Company registration appears on the official timeline in February 2023, making early 2023 the most defensible founding anchor. SO006, SO020
CO002 Zhiyuan describes itself as a general-purpose AI robotics company centered on embodied intelligence rather than a single vertical application robot. SO006, SO001
CO003 Official materials say the company builds its system around three pillars: robot body, intelligent algorithms, and open platform. SO006, SO004
CO004 The published mission is “create unlimited productivity via intelligent machines.” SO006, SO001
CO005 Official leadership materials name Deng Taihua as founder, chairman, and CEO. SO007, SO017
CO006 Official leadership materials name Peng Zhihui as co-founder, president, and CTO. SO007, SO014
CO007 The disclosed executive bench also includes Jiang Qingsong, Yao Maoqing, Wang Chuang, Luo Jianlan, and Niu Jia. SO007
CO008 Independent coverage says Peng Zhihui was born in 1993, studied at UESTC, and previously worked at OPPO and Huawei's Genius Youth program. SO014, SO020
CO009 Peng Zhihui's pre-company public persona as “Zhi Hui Jun” materially increases brand reach and recruiting visibility. SO014, SO017
CO010 The official company profile says Expedition A1 was released in August 2023, roughly six months after registration. SO006
CO011 The same official timeline says a manufacturing factory landed in Shanghai in January 2024. SO006
CO012 Official materials say Expedition A2 and Lingxi X1 were released in August 2024. SO006
CO013 The official timeline says cumulative general-purpose robot output reached 1,000 units in 2025. SO006, SO005
CO014 Official materials say Zhiyuan released its first general embodied foundation model and Lingxi X2 in March 2025. SO006, SO011
CO015 Official materials say Genie Studio was released in April 2025 as a one-stop embodied-AI development platform. SO006, SO011
CO016 36Kr reported that Tencent led a March 2025 financing round with multiple industrial investors and existing shareholders following. SO014
CO017 The same 36Kr report said JD and the Shanghai Embodied Intelligence Fund joined a later 2025 financing round. SO014
CO018 Zhidx reported that Zhiyuan's valuation was about US$2.07 billion, or roughly RMB 14.7-15.0 billion, as of March 2025. SO015, SO014
CO019 Zhidx characterized a Hong Kong IPO as a 2026 plan and target scenario, not as a closed financing event. SO015
CO020 Zhidx listed strategic and financial investors including Tencent, JD, SAIC, LG Electronics, Mirae Asset, C Capital, and Hillhouse-linked capital. SO015, SO014
CO021 The A3 product page describes a 173 cm, 55 kg humanoid with 10-hour endurance, dual-battery architecture, and 10-second hot-swap capability. SO009, SO010
CO022 The July 2026 WAIC article claims Expedition A3 is the first full-size humanoid robot to play ping-pong fully autonomously in real time. SO010
CO023 The official July 2026 production release says Zhiyuan passed 15,000 embodied robots less than three months after the prior milestone. SO012
CO024 Forbes reported that AgiBot shipped 5,000 humanoid robots in three months to reach 10,000 total robots shipped by March 2026. SO018
CO025 TrendForce said AgiBot and Unitree could capture nearly 80% of China's humanoid robot market share in 2026. SO025
CO026 The Tencent/China Times profile said Zhiyuan won China Mobile and Chery orders and also referenced additional multi-million-yuan contracts. SO016
CO027 That same profile said Zhiyuan had already listed robots priced as low as RMB 98,000 in August 2025. SO017
CO028 Management publicly said most of the robot supply chain is localized, while high-end compute chips still come from overseas. SO016, SO017
CO029 The official careers page and QCC job listing together show the company remained in aggressive hiring mode during 2025-2026. SO008, SO021
CO030 The QCC recruitment page displayed more than 500 postings, which is evidence of active recruiting but not an auditable employee-count disclosure. SO021
CO031 Gasgoo described Zhiyuan's spin-off of its dexterous-hand unit as a strategic response to scale pressures on the full-stack model. SO019
CO032 Gasgoo also said dexterous hands represent roughly 17%-18% of total robot cost, underscoring why component strategy matters commercially. SO019
CO033 A Tencent News legal brief reported a 2026 copyright-infringement hearing involving Zhiyuan in Pudong. SO020
CO034 Official English and Chinese materials consistently use AGIBOT internationally and Zhiyuan Innovation domestically for the same company. SO004, SO006
CO035 Official materials tie the company to Shanghai and specifically identify a Lingang manufacturing footprint. SO006, SO009
CO036 The official values page says Zhiyuan prioritizes large, high-growth leading customers rather than undifferentiated mass-market demand. SO006
CO037 Public sources do not disclose audited revenue, debt, or a clean current headcount, so those metrics remain diligence gaps rather than cover facts. SO014, SO015, SO021
CO038 The public product footprint now spans Expedition, Lingxi, Genie, C5, data services, and embodied-AI tooling rather than a single flagship robot. SO001, SO002, SO006
CO039 The July 2026 forum article says Zhiyuan is building a pretraining, post-training, and continual-learning stack around GO-2, Act2Goal, GE-Sim 2.0, SOP, and Genie Evolver. SO011
CM001 Zhiyuan's public market should be defined as general-purpose humanoid and embodied-AI systems, not as the entire robotics market. SM001, SM004
CM002 That boundary includes robot hardware, embodied-model software, data collection, developer tooling, and related operating services where the company is public today. SM001, SM002, SM003
CM003 It excludes conventional fixed industrial arms, pure software copilots, and broader automation categories where Zhiyuan has no direct product proof in reviewed sources. SM001, SM002
CM004 Official product pages show Zhiyuan already addresses industrial, commercial, entertainment, education, and data-service scenarios rather than one narrow niche. SM002, SM005, SM006, SM007
CM005 IDC defined humanoid robots as human-form systems with torso, head, and dual arms plus perception, learning, decision, and execution capabilities. SM017
CM006 IDC said global humanoid shipments reached roughly 18,000 units and revenue around US$440 million in 2025. SM017
CM007 IDC said China vendors dominated 2025 shipments and identified AgiBot and Unitree at roughly 5,000 units each. SM017, SM011
CM008 Mordor estimated the humanoids market at about US$3.93 billion in 2026 and US$17.8 billion by 2031. SM018
CM009 Global Market Insights estimated the global humanoid robot market at about US$10.9 billion in 2026 and US$192.7 billion by 2035. SM019
CM010 The spread between IDC, Mordor, and GMI estimates shows that market sizing varies sharply with scope, geography, and forecast horizon. SM017, SM018, SM019
CM011 TrendForce expects China's humanoid output to rise 94% in 2026 as the market enters a more concrete commercialization phase. SM011
CM012 TrendForce says the industry's competitive focus is shifting from foundational capability demos toward tangible user value in real scenarios. SM011
CM013 China's 14th Five-Year robot plan is a formal policy tailwind for the robot industry. SM013, SM012
CM014 China's 2023 humanoid guidance explicitly set commercialization and technological-development targets for the humanoid segment. SM014, SM015
CM015 The 2026 standards-system release shows China is building a lifecycle framework for safety, applications, and embodied-intelligence standards. SM016
CM016 Manufacturing and logistics are among the clearest near-term buyers because they already own labor budgets and operate structured environments. SM005, SM011, SM023
CM017 Reception, guidance, and branded interaction appear commercially useful earlier than home-robotics deployment because the proof burden is lower. SM006, SM010
CM018 Education, research, and data-collection customers matter as monetizable stepping stones even if they are not the largest long-run market. SM003, SM004, SM011
CM019 Home robotics remains strategically large but still has the highest safety, cost, and reliability burden in public commentary. SM010, SM015
CM020 Buyer, user, and payer are not the same across segments: factories buy for productivity, service venues buy for customer interaction, and researchers buy for experimentation. SM005, SM006, SM010
CM021 Key adoption drivers include labor pressure, aging populations, falling component costs, maturing embodied models, and human-scale infrastructure reuse. SM018, SM019, SM011
CM022 Zhiyuan management publicly argues that some industrial robots are entering an investable ROI zone with about two-year payback potential. SM009
CM023 Management also argues that higher shipments should further lower costs and raise price-performance over time. SM009
CM024 The market is still constrained by reliability, model generalization, safety, supply-chain depth, and scenario-specific integration work. SM009, SM015, SM016
CM025 There is still no trustworthy public SAM or SOM estimate for Zhiyuan specifically, because open sources rarely disclose usable segment denominators. SM017, SM018, SM019
CM026 Zhiyuan's public evidence is strongest in China-centric industrial and commercial scenarios rather than in a balanced global segment mix. SM010, SM011
CM027 Competing official product pages from Unitree, UBTECH, Figure, Apptronik, and Agility show that the addressable market is converging around factories, logistics, and structured enterprise tasks. SM020, SM021, SM022, SM024, SM025
CM028 Figure's BMW deployment and BMW's own release support the view that automotive manufacturing is a flagship enterprise use case for advanced humanoids. SM022, SM023
CM029 UBTECH's Walker S positioning supports the same conclusion from a China-based competitor: industrial multi-task scenarios are a core budget pool. SM021
CM030 Agility and Apptronik public materials suggest North American peers are still emphasizing commercialization infrastructure and partnerships rather than broad unit volumes. SM024, SM025
CM031 The business model is already shifting from one-time hardware sales toward products plus services plus ecosystem, according to IDC and company materials. SM017, SM001, SM003
CM032 Zhiyuan's public data-service and developer-tooling surfaces imply it wants share of training, integration, and ecosystem value in addition to robot ASPs. SM002, SM003, SM008
CM033 Standards and ethics work matters commercially because enterprise buyers need proof of safe deployment and maintainable lifecycle controls. SM016, SM023
CM034 The near-term market should be read as a constrained adoption funnel, not a fully formed mass market, because proof narrows sharply from demos to repeat deployments. SM011, SM017, SM010
CM035 Public sources still over-index on excitement and shipment headlines compared with rigorous ROI, churn, or segment-penetration data. SM010, SM017, SM018
CM036 The best market reading for investors is therefore evidence-constrained: large long-run TAM, much smaller near-term SAM, and an even smaller currently proven SOM. SM017, SM018, SM019, SM010
CM037 Because Zhiyuan already spans industrial, service, research, and tooling surfaces, its go-to-market addressability is broad, but its proof burden is equally broad. SM002, SM004, SM009
CP001 Zhiyuan's closest direct peers are Unitree, Figure AI, UBTECH, Agility Robotics, and Apptronik rather than every broad robotics company. SP007, SP008, SP012, SP017, SP018
CP002 Status-quo substitutes for many industrial tasks remain fixed automation, wheeled mobile robots, and conventional logistics systems. SP005, SP011, SP012
CP003 Zhiyuan competes globally for the same buyer attention as companies pitching factories, logistics halls, or structured service venues as the first beachhead. SP005, SP009, SP011, SP012, SP017, SP018
CP004 TrendForce positioned AgiBot and Unitree as the likely shipment-share leaders in China for 2026. SP007
CP005 Figure announced more than US$1 billion in Series C financing at a US$39 billion post-money valuation in 2025. SP008
CP006 Figure publicly claims deep automotive deployment proof through BMW. SP009, SP010, SP011
CP007 BMW's own 2026 press release independently confirms ongoing Figure deployment work in Spartanburg. SP011, SP009
CP008 Figure said its robots contributed to production of 30,000 cars at BMW, which is unusually concrete enterprise proof for the category. SP010
CP009 Unitree's public H1 materials emphasize general-purpose humanoid capability and its store page indicates pricing is negotiated rather than openly standardized. SP014, SP015
CP010 Unitree's Dex5 page underscores how dexterous hands are becoming a visible competitive battleground, not a hidden subsystem. SP016, SP021
CP011 UBTECH positions Walker S explicitly around industrial multi-task scenarios, reinforcing factory work as a core peer battleground. SP012
CP012 Agility's 2026 press chronology emphasizes public-market readiness and new facilities, suggesting commercialization infrastructure still matters as much as shipped units. SP017
CP013 Apptronik's 2026 press flow emphasizes capital raised, customer production partnerships, and deployment infrastructure more than shipment volume. SP018, SP019, SP020
CP014 CNBC reported Apptronik raised US$520 million at a US$5 billion valuation in 2026, placing it well above Zhiyuan's last public 2025 mark. SP020, SP019
CP015 Zhiyuan's public moat argument is full-stack integration across robot body, model stack, developer tools, and supply chain rather than dominance in one single component. SP001, SP022, SP023
CP016 Zhiyuan's China-based scale narrative is stronger on shipments and investor ecosystem than most U.S. peers, but weaker on globally trusted revenue disclosure. SP007, SP020, SP025
CP017 Zhiyuan's customer references to China Mobile and Chery give it more public China enterprise proof than many peers have in named logos. SP005
CP018 Figure still dominates the headline global valuation narrative by a wide margin over Zhiyuan. SP008, SP020
CP019 Public pricing transparency is poor across the set: Unitree negotiates, Zhiyuan has one low-end price signal, and most peers disclose no standardized ASP. SP015, SP024, SP018
CP020 This weak pricing transparency means feature matrices are easier to build than true value-for-money comparisons. SP015, SP018, SP024
CP021 Zhiyuan's portfolio breadth spans factory, service, entertainment, tooling, and data surfaces rather than a single industrial humanoid SKU. SP001, SP002, SP003, SP023
CP022 Figure's moat rests more heavily on frontier-model and flagship-enterprise narrative, while Zhiyuan's rests more on China ecosystem and scaling speed. SP008, SP009, SP023, SP025
CP023 UBTECH and Unitree show that China-specific manufacturing depth can compress hardware and component iteration cycles. SP007, SP012, SP014
CP024 Gasgoo's dexterous-hand spin-off analysis is adverse evidence that the full-stack model can strain focus and efficiency even for category leaders. SP021
CP025 Apptronik, Agility, and Figure all highlight facilities, production systems, or partner sites, which suggests manufacturing execution itself is now a competitive layer. SP009, SP011, SP017, SP018
CP026 Distribution leverage remains underdisclosed across nearly all peers because integrator, service, and deployment economics are only partially public. SP005, SP017, SP018
CP027 Zhiyuan's public shipment narrative is stronger than many U.S. peers, but its global brand trust and financial disclosure are still weaker than best-in-class enterprise vendors. SP025, SP020, SP011
CP028 Figure has the strongest visible valuation advantage, but that does not automatically prove a stronger China deployment moat than Zhiyuan or Unitree. SP008, SP007
CP029 Unitree's pricing and component disclosures make it a more transparent hardware comparator than most peers. SP014, SP015, SP016
CP030 UBTECH offers a clearer industrial-workflow story than entertainment-first robots, which is why it belongs in Zhiyuan's serious enterprise peer set. SP012, SP013
CP031 Apptronik and Agility appear earlier in large-scale public deployment than Figure on unit volume, but not necessarily behind on commercialization intent. SP017, SP018, SP019
CP032 The strongest status-quo substitute in factory work is still fixed or simpler automation when tasks do not require human-form flexibility. SP011, SP012
CP033 Moat durability in humanoids likely depends on the loop among data, deployment, serviceability, supply-chain depth, and financing—not on one hero demo. SP009, SP017, SP018, SP021
CP034 Public proof suggests Zhiyuan is ahead on China-centric scale narrative, industrial ecosystem density, and breadth of public product surfaces. SP006, SP007, SP023, SP025
CP035 Public proof suggests Zhiyuan is behind Figure on valuation, behind the best U.S. peers on investor familiarity, and behind mature public companies on financial transparency. SP008, SP011, SP020
CP036 The largest unresolved competitive denominator is repeat paying deployment count by vendor, because valuation and demo quality are poor substitutes for it. SP007, SP017, SP018, SP025
CI001 Zhiyuan’s public monetization surface is best read as a mixed hardware, deployment-service, and ecosystem model rather than as a pure robot-box sale. SI001, SI002, SI003, SI011
CI002 Official pages show Zhiyuan now sells or supports full-size humanoids, wheeled platforms, data-related infrastructure, and solutions work across scenarios. SI001, SI002, SI003, SI005
CI003 The China Mobile tender and Fulin deployment both imply project-style enterprise revenue recognition rather than simple off-the-shelf consumer checkout. SI006, SI009
CI004 Public order headlines should not be treated as equivalent to realized revenue because delivery, acceptance, service, and deployment milestones still matter. SI007, SI009, SI026
CI005 The strongest public revenue proof so far is enterprise deployment activity in factories and telecom procurement rather than disclosed annual sales figures. SI006, SI008, SI009, SI011
CI006 The Fulin Precision case shows industrial humanoid deployment can begin with a small on-site fleet before scaling toward larger contracted unit counts. SI006, SI007, SI008
CI007 Yicai reported four A2-W robots handling more than 800 boxes in three hours at Fulin Precision’s plant. SI006
CI008 Assembly and Shanghai municipal reporting said the broader Fulin deal covered nearly 100 A2-W robots, indicating a meaningful follow-on scale step beyond the first four units. SI007, SI008
CI009 Fulin’s engineering team estimated one A2-W delivers about 0.7 of a human worker under normal operations and as much as 1.4 to 2 workers when running 24/7. SI006
CI010 That same article said local labor cost was about RMB80,000 per worker annually, giving investors a rough ROI benchmark rather than a guaranteed payback model. SI006
CI011 Yicai reported similar humanoids currently price in a roughly RMB400,000 to RMB800,000 range and suggested A2-W sits inside that corridor. SI006
CI012 Zhiyuan management separately discussed public pricing as low as roughly RMB98,000 for one model, which is useful as an entry signal but not a reliable enterprise ASP benchmark. SI010
CI013 Taken together, the public price record implies a wide monetization ladder across products and use cases rather than one standard catalog price. SI001, SI006, SI010
CI014 The China Mobile tender was worth about RMB124 million in total, with AgiBot’s portion reported at about RMB78 million for full-sized humanoids. SI009
CI015 That tender supports the view that Zhiyuan can monetize through large project packages with strategic customers, but it does not prove repeat annual renewal economics. SI009, SI011
CI016 Management’s commercialization commentary suggests deployment specialists and integration partners are part of the delivery model, meaning gross margin depends on more than factory BOM. SI006, SI010
CI017 The Fulin article explicitly states operation-and-maintenance roles add cost, which means deployment support is a real unit-economics line item, not a footnote. SI006
CI018 Initial deployments can take three to four months before later rollouts compress to weeks or hours when scenarios overlap, implying integration learning effects but also front-loaded labor. SI006
CI019 Zhiyuan’s public product and solution pages imply revenue can expand beyond robot sales into data services, deployment tooling, and customer-specific solutions. SI002, SI003, SI004
CI020 The business model therefore has more resemblance to capital-intensive systems integration than to a pure software subscription business. SI002, SI003, SI006, SI011
CI021 TrendForce’s shipment forecast and management commentary both suggest higher volumes are central to lowering unit costs and improving price-performance. SI010, SI012
CI022 Public peer evidence supports this logic: Figure’s BotQ and Apptronik’s manufacturing announcements frame factory throughput as a financial lever, not just a technical brag. SI016, SI019, SI021
CI023 Figure’s BMW proof, Apptronik’s Mercedes agreement, and GXO initiative all indicate that enterprise humanoid vendors are monetizing through deployment programs before any mass consumer market exists. SI017, SI018, SI021, SI022
CI024 UBTECH’s 2025 annual report provides the clearest listed-peer benchmark for category financials. SI013, SI014
CI025 UBTECH reported 2025 revenue of RMB2,001.0 million, up 53.3% year over year. SI014
CI026 UBTECH said full-size embodied humanoid products and services generated roughly RMB820.6 million in 2025 and became its largest revenue source. SI014
CI027 UBTECH reported 2025 gross profit of RMB753.8 million and a gross margin of 37.7%, up from 28.7% in 2024. SI014
CI028 UBTECH still posted a 2025 net loss of roughly RMB789.8 million despite that revenue growth, underscoring how capital-intensive the category remains even at larger scale. SI014
CI029 UBTECH reported operating cash outflow of about RMB784.1 million in 2025, which is a useful public burn benchmark for the sector. SI014
CI030 UBTECH ended 2025 with about RMB4,887.9 million in cash and cash equivalents and RMB1,123.0 million of external bank borrowings. SI014
CI031 UBTECH also disclosed accounts receivable of roughly RMB1.84 billion, illustrating how working-capital intensity can rise with enterprise and state-linked customers. SI014
CI032 Zhiyuan has not provided a comparable public cash, receivable, inventory, or borrowings disclosure, leaving investors to infer capital needs from fundraising and deployment pace. SI011, SI014, SI026
CI033 The absence of a public cash balance means Zhiyuan’s runway cannot be underwritten directly from open sources. SI011, SI014
CI034 Because the company is still scaling manufacturing, deployment teams, and data infrastructure, it is reasonable to treat Zhiyuan as financing-dependent in 2026. SI001, SI002, SI006, SI011, SI026
CI035 Apptronik’s >US$935 million Series A and CNBC’s US$520 million 2026 raise report reinforce that peer humanoid vendors still require very large external capital pools. SI019, SI020
CI036 The right public reading of Zhiyuan’s financial profile is early-commercial but not yet fully underwritable: revenue is emerging, pricing is partially visible, margins are unproven, and capital adequacy is still opaque. SI006, SI009, SI010, SI014, SI026
CI037 The biggest missing metrics before underwriting are realized ASP by model, gross margin by deployment type, service cost per active robot, receivable aging, and monthly burn. SI006, SI010, SI014
CE001 Zhiyuan’s public product surface now spans industrial humanoids, interaction-oriented humanoids, cleaning systems, data infrastructure, and developer-facing assets rather than one flagship robot. SE001, SE007, SE008, SE009, SE010, SE011
CE002 The company should therefore be analyzed as a product family plus full-stack platform, not only as a single hardware SKU. SE001, SE002, SE013
CE003 G1 is publicly framed around industrial, commercial, and home scenarios, with one-handed 3kg operation, >2m work height, and compatibility with most factory aisles. SE007
CE004 G1’s page also highlights a one-stop embodied-AI development platform, multimodal data collection, and an open million-scale dataset narrative. SE007, SE006
CE005 X2 is positioned around multimodal interaction, autonomous navigation, and a high-DoF anthropomorphic body for service-like or consumer-adjacent use cases. SE008
CE006 A3 is presented as a performance- and coordination-oriented humanoid with long endurance, precision positioning, and 100+ unit group-control support. SE009
CE007 C5 extends the portfolio beyond humanoids into large-scenario intelligent cleaning with sensors, workstation automation, and digitalized maintenance workflows. SE011
CE008 The public combination of humanoids, cleaning, and data-related assets implies Zhiyuan wants share of workflow coverage rather than only one-body sales. SE010, SE011, SE013
CE009 AgiBot Research, D1/DaaS surfaces, and APC 2026 all reinforce a full-stack thesis that includes models, data, and tooling on top of robot hardware. SE002, SE010, SE013
CE010 The AGIBOT X1 open-source docs and GitHub repositories disclose a modular humanoid stack built around AimRT middleware and reinforcement-learning-centric development. SE003, SE004, SE005
CE011 The open repositories are meaningful developer signal because they move part of the architecture from marketing copy into inspectable technical artifacts. SE003, SE004, SE005
CE012 OpenDriveLab’s AgiBot-World repository strengthens that signal by pointing to a large-scale manipulation platform and benchmark-oriented data ecosystem around AgiBot. SE006
CE013 The public stack emphasizes data and learning loops as much as mechanical design. SE002, SE006, SE007, SE022
CE014 Yicai reported that one Zhiyuan deployment relied on roughly 95% simulated data and 5% real-world data, highlighting a simulation-heavy deployment method. SE016
CE015 The same source says real-world variability still challenges safety and reliability, which means simulation leverage does not eliminate field adaptation risk. SE016
CE016 Fulin reporting also showed the robot stopped immediately when a person approached, providing concrete evidence of at least one operational safety behavior in the field. SE016
CE017 The company’s public architecture story appears to have at least five layers: robot bodies, perception/control, embodied models, data and training, and deployment operations. SE002, SE003, SE007, SE014, SE016
CE018 WITA is positioned as a multimodal foundation-model layer for embodied interaction rather than as a simple app feature. SE014
CE019 WITA’s benchmark result is a useful model-ambition signal, but it does not by itself prove production reliability, customer ROI, or safety in deployment. SE014, SE016
CE020 WAIC 2026 forum materials show Zhiyuan placing itself inside the “physical AI” and world-model conversation rather than only the hardware conversation. SE015, SE014
CE021 APC 2026 said more than 2,500 partners from over 30 countries attended, which supports ecosystem ambition and partner-led distribution claims. SE013
CE022 That partner conference signal is strategically meaningful because deployment of humanoids likely depends on integrators, distributors, and local solution partners. SE013, SE016
CE023 Gasgoo’s spin-off coverage is adverse evidence that full-stack ambition can create modularization or focus pressure around dexterous-manipulation capabilities. SE012, SE026
CE024 OmniHand’s separate product presence supports the idea that dexterous manipulation is being modularized as a component-level capability. SE012, SE026
CE025 China’s humanoid guidance and emerging standards framework show that trust, safety, and interoperability are becoming design constraints, not afterthoughts. SE018, SE019, SE020
CE026 SESEC’s standards summary suggests China is building a lifecycle system for humanoid safety, applications, and embodied-intelligence standards. SE020, SE019
CE027 Despite that policy backdrop, Zhiyuan’s reviewed public pages do not provide a full enterprise-grade disclosure set for uptime, incidents, cybersecurity, or formal certifications. SE001, SE016, SE019
CE028 No reviewed public source supplied a clean uptime or MTBF metric for Zhiyuan’s deployed robots. SE016, SE017
CE029 No reviewed public source supplied a detailed privacy or cybersecurity control sheet for telemetry, customer data, or model-data governance. SE001, SE017, SE019
CE030 No reviewed public source supplied a customer-facing SLA or formal support commitment schedule. SE016, SE017
CE031 Compared with Figure, Zhiyuan discloses more about open technical artifacts than some peers do, but much less about measured production performance than a buyer might want. SE004, SE005, SE021, SE022
CE032 Figure’s logistics and pretraining posts show a similar emphasis on data scale and workflow learning, which suggests the competitive frontier is moving toward data-and-model loops. SE021, SE022
CE033 UBTECH Walker S1, Unitree H2, and Apptronik Apollo 2 all show that peers also pitch integrated planning, mobility, and manipulation stacks, so Zhiyuan’s differentiation must come from execution, openness, and scenario coverage rather than architecture labels alone. SE023, SE024, SE025
CE034 Zhiyuan’s clearest public differentiation signals are breadth of form factors, open-source/developer surfaces, and aggressive ecosystem-building around embodied AI. SE003, SE007, SE013, SE015
CE035 The current roadmap signal is expansion, not simplification: new model releases, open-source assets, global partner conferences, and benchmark announcements all point to stack broadening. SE013, SE014, SE015
CE036 That roadmap broadening creates both upside and risk: it can deepen the moat if modules reinforce each other, but it can also outpace public proof on reliability, supportability, and governance. SE014, SE016, SE026
CU001 Zhiyuan’s public customer proof set is centered on industrial manufacturing, telecom procurement, and partner-led commercialization rather than on consumer adoption. SU001, SU002, SU005, SU007, SU009
CU002 The solutions page and deployment reporting imply several customer buckets: factories, logistics or warehousing sites, service / interaction venues, and ecosystem partners who localize deployments. SU001, SU002, SU009
CU003 Manufacturing is the strongest public segment today because it has the clearest named customer evidence and workflow detail. SU002, SU003, SU004, SU008
CU004 Fulin Precision is the most concrete named-customer proof in the reviewed set because public reporting covers the site, tasks, labor logic, and scale-up path. SU002, SU003, SU004
CU005 At Fulin Precision, four A2-W robots were initially installed to identify and sort deliveries, giving unusually specific workflow evidence for the category. SU002
CU006 Assembly and Shanghai government reporting said the broader Fulin deal covered nearly 100 A2-W robots, which is the clearest public signal of expansion beyond a tiny pilot. SU003, SU004
CU007 The Fulin case therefore provides the best public proxy for customer durability: expansion from first on-site deployment into a much larger announced order. SU002, SU003, SU004
CU008 China Mobile is the clearest telecom and strategic-procurement proof point in public sources. SU005, SU006
CU009 Yicai reported a CNY124 million China Mobile tender, with AgiBot’s portion at CNY78 million for full-sized humanoids. SU005
CU010 That China Mobile deal matters because it shows a large budget owner is willing to procure humanoid systems at meaningful scale, even if recurring economics remain undisclosed. SU005, SU006
CU011 Tencent News reporting adds Chery to the named-customer set, broadening Zhiyuan’s public proof into automotive-adjacent accounts. SU006
CU012 Longcheer adds a consumer-electronics precision-manufacturing proof point that extends the customer set beyond automotive-parts and telecom examples. SU008
CU013 The Longcheer announcement is important because it positions embodied AI on a mass-production line rather than only in a lab or marketing venue. SU008
CU014 PR Newswire’s “Deployment Year One” framing suggests the company is intentionally shifting the story from technology milestones toward customer deployment milestones. SU007, SU009
CU015 APC 2026 and the UK / ANZ follow-on events show customer expansion is expected to run partly through partner and channel relationships, not only direct selling from Shanghai. SU009, SU010, SU011
CU016 APC 2026 said more than 2,500 partners from over 30 countries attended, which is strong ecosystem evidence but not the same as disclosed paying-customer count. SU009
CU017 Public customer proof still skews heavily toward China because the strongest named deployments and procurement examples are Chinese. SU002, SU005, SU006, SU008
CU018 The UK and Australia / New Zealand APC events show overseas channel-building momentum, but they do not yet disclose a large base of overseas paying customers. SU010, SU011
CU019 Service, tourism, and interaction scenarios are visible in company messaging but less concretely proven than industrial manufacturing in reviewed sources. SU001, SU009, SU024
CU020 The public customer narrative should therefore be read as strongest in structured enterprise workflows and weakest in consumer-adjacent or soft-benefit scenarios. SU001, SU002, SU024
CU021 No reviewed public source provides NRR, logo retention, churn, or annual renewal rates for Zhiyuan. SU002, SU006, SU024
CU022 No reviewed public source provides an exact customer concentration metric beyond a small handful of named logos and partner events. SU005, SU006, SU009
CU023 No reviewed public source provides a formal enterprise SLA or support KPI for named customers. SU002, SU006
CU024 Because the named proof set is still small, concentration risk remains meaningful even if the logos themselves are encouraging. SU002, SU005, SU006, SU024
CU025 The best public sign of customer satisfaction is follow-on scale behavior such as Fulin’s move toward nearly 100 units, not survey or NPS disclosure. SU003, SU004
CU026 Another useful durability proxy is that China Mobile’s tender was large enough to suggest a formal procurement process rather than an informal showcase relationship. SU005
CU027 Investors should still distinguish orders, tenders, and signed projects from live, fully scaled production usage. SU003, SU005, SU007, SU024
CU028 Peer customer-proof patterns reinforce this caution: Figure uses BMW, Apptronik cites Mercedes and GXO, and UBTECH markets industrial solutions, but all vendors still curate their strongest workflows publicly. SU013, SU014, SU015, SU017, SU018, SU021
CU029 BMW’s own release makes Figure’s customer proof higher-confidence than company-only claims, illustrating the standard Zhiyuan should ideally meet more often. SU013, SU014
CU030 Apptronik’s GXO and Mercedes materials show how customer proof usually matures: from pilot-style workflow fit into broader operational categories like manufacturing and 3PL. SU017, SU018, SU019, SU020
CU031 UBTECH’s industrial solution page shows another peer pattern: publicizing solution archetypes even when exact logo economics remain thin. SU015, SU016
CU032 TrendForce and IDC both support the idea that AgiBot is among the category’s shipment leaders in China, which makes the still-limited customer disclosure more notable, not less. SU012, SU025
CU033 The Wire China’s adverse framing is a reminder that category excitement can run ahead of repeat enterprise usage and that some public proof remains narrative-heavy. SU024
CU034 Zhiyuan’s public customer case is real, but it still depends on a narrow cluster of named proofs—Fulin, China Mobile, Chery, and Longcheer—plus broad partner ecosystem signals. SU002, SU005, SU006, SU008, SU009
CU035 That is enough to support a “credible early enterprise adoption” conclusion, but not enough to support a mature retention or diversification conclusion. SU007, SU021, SU022, SU024
CU036 The highest-priority missing metrics are active customer count, deployments by stage, repeat-order rate, revenue concentration, overseas paying-customer count, and post-installation support load. SU002, SU005, SU009, SU024
CR001 Zhiyuan’s risk profile is best understood as a combination of regulatory, operational, partner, and financing risk rather than one single technical risk. SR003, SR014, SR017, SR023
CR002 China’s 14th Five-Year robot plan and the humanoid-guidance documents are policy tailwinds, but they also create a rising compliance bar for safety, standards, and industrial applicability. SR001, SR002, SR003
CR003 SESEC’s standards summary suggests China is moving toward a more formal lifecycle framework for humanoid safety, applications, and embodied-intelligence standards. SR013, SR003
CR004 That trend lowers long-run ambiguity but raises near-term execution pressure because companies must align product behavior, testing, and documentation to emerging standards. SR003, SR013
CR005 Product-liability law matters directly because embodied robots act in physical environments and can cause property damage, bodily harm, or workflow disruption. SR008, SR009
CR006 Hill Dickinson’s humanoid-law note says autonomy complicates accountability by spreading responsibility across manufacturer, operator, and software layers. SR009
CR007 Data-security and privacy obligations are not peripheral here, because robots collect operational and environmental data inside customer facilities. SR005, SR006, SR007, SR012
CR008 The amended Cybersecurity Law effective in 2026 and the broader PIPL / DSL framework increase compliance burden for AI-driven deployments in China. SR005, SR006, SR007, SR012
CR009 Zhiyuan’s public disclosures do not yet provide a detailed external control pack for telemetry, customer data, or cross-border data handling. SR006, SR007, SR012
CR010 MOFCOM’s 2026 export-control announcement and the related legal commentary show that technology and supply-chain restrictions can shift quickly in this environment. SR004, SR010, SR011
CR011 For a humanoid startup that depends on advanced components and international expansion, such export-control volatility is a real supply and customer-delivery risk. SR004, SR010, SR011
CR012 Public field reporting from Fulin demonstrates that busy factory environments create real safety and reliability requirements, not just demo-stage concerns. SR014
CR013 The same report showed the robot stopped when a human approached, which is a positive operational control signal but also evidence that human-proximity risk is central to deployment design. SR014
CR014 Fulin reporting also said operation-and-maintenance roles require higher technical expertise, highlighting field-service and staffing risk. SR014
CR015 Zhiyuan’s deployment model still depends on significant scenario adaptation, because early deployments can take months before later rollouts compress. SR014
CR016 Yicai also reported a roughly 95% simulated / 5% real-world data mix in one deployment context, which underscores both the power and the risk of simulation-heavy training. SR014
CR017 Simulation leverage reduces deployment cost, but field variability still creates generalization and safety risk that must be managed in production sites. SR014, SR015
CR018 The Wire China’s skeptical framing is useful adverse evidence that category-level enthusiasm can outpace real deployment maturity. SR017
CR019 Gasgoo’s coverage of a dexterous-hand spin-off suggests product-scope broadening can create focus and coordination risk even when it improves specialization. SR018
CR020 Partner and integrator dependence is structural to the current commercialization model. SR014, SR020, SR021
CR021 Fulin’s site used Anu Intelligent for integration, showing that deployment execution is not purely internal to Zhiyuan. SR014
CR022 APC 2026’s 2,500+ partners from 30+ countries and the company’s deployment-year messaging imply growing channel dependence as commercialization scales. SR020, SR021
CR023 Partner dependence can accelerate reach, but it also creates execution, quality-control, and brand-consistency risk across geographies. SR020, SR021
CR024 Customer concentration risk is visible because the named public proof set clusters around a small group of logos such as Fulin, China Mobile, Chery, and Longcheer. SR016, SR019, SR022
CR025 The China Mobile tender is large enough that slippage or underperformance on a few accounts could matter disproportionately to narrative and economics. SR019
CR026 No reviewed public source gives a full customer concentration table, so actual exposure could be better or worse than the visible narrative implies. SR016, SR019, SR022
CR027 People and org-execution risk is meaningful because commercialization requires product, service, data, and partner-management functions to scale together. SR014, SR018, SR020
CR028 Financial risk remains high because Zhiyuan does not publicly disclose cash, burn, or borrowings with the specificity investors would want. SR016, SR017, SR023
CR029 UBTECH’s 2025 annual report shows that even a larger listed peer can generate RMB2.0 billion of revenue and still post a substantial net loss and operating cash outflow. SR023
CR030 That peer benchmark supports the conclusion that category capital intensity is structural, not merely a Zhiyuan-specific issue. SR023, SR027, SR028
CR031 TrendForce and IDC may support shipment leadership, but shipment leadership is not the same as de-risked unit economics, customer durability, or regulatory readiness. SR029, SR030, SR017
CR032 Figure, UBTECH, and Apptronik all rely on curated flagship workflows or heavy capital buildout, which reinforces that customer, operations, and financing risks are category-wide. SR023, SR024, SR025, SR026, SR027, SR028
CR033 Visible mitigations already exist: safety-stop behavior in the field, open policy engagement, standards participation, partner expansion, and deployment-learning loops. SR003, SR013, SR014, SR020
CR034 But these mitigations are incomplete until backed by more robust uptime, incident, service, and compliance disclosures. SR006, SR008, SR014
CR035 A rational kill criterion would be failure to convert high-profile orders into stable, referenceable live deployments within a reasonable follow-on period. SR017, SR019, SR020
CR036 Another kill criterion would be evidence that service burden, safety incidents, or support costs rise faster than deployment value. SR014, SR023
CR037 A third kill criterion would be inability to produce an auditable control pack for data governance, product safety, and liability ownership before broader enterprise rollouts. SR006, SR007, SR008, SR009
CR038 Overseas expansion increases compliance complexity because different jurisdictions may apply different safety, procurement, and data expectations on top of China-based supply constraints. SR009, SR010, SR011, SR021
CR039 Standards and policy milestones can lower long-term adoption risk by clarifying expectations, but they do not remove the need for company-level execution discipline. SR001, SR003, SR013
CR040 Spinning out modules or broadening the stack can improve specialization, but it also increases coordination and integration risk across product lines. SR018, SR020
CR041 Because the current proof set is narrow, poor performance at one or two flagship accounts could materially slow future sales conversion. SR016, SR019, SR022
CV001 Zhiyuan merits a real investment discussion because public evidence now supports non-trivial shipment, customer, and product breadth rather than a concept-stage story. SV003, SV004, SV024, SV025, SV028
CV002 The strongest pro-thesis pillars are China shipment leadership, growing named-customer proof, broad product surfaces, and an ecosystem-building posture around embodied AI. SV004, SV008, SV024, SV028
CV003 The strongest anti-thesis pillars are underdisclosed revenue quality, capital intensity, customer concentration, and the possibility that deployment excitement is outrunning durable economics. SV012, SV023, SV024, SV026
CV004 Zhiyuan’s most defensible current valuation anchor remains the roughly RMB15 billion / about US$2.07 billion 2025 mark reported in public media. SV001, SV002
CV005 That anchor is meaningful because it places Zhiyuan clearly above an ordinary early-stage hardware startup but far below the most inflated global private humanoid narratives. SV002, SV018, SV022
CV006 Figure’s US$39 billion Series C valuation and Apptronik’s US$5 billion financing context show how wide the global valuation band is for premium humanoid narratives. SV018, SV022
CV007 Zhiyuan’s public valuation is therefore not extreme by global private-peer standards, but it is stretched relative to its current public disclosure quality. SV002, SV012, SV026
CV008 TrendForce, IDC, and AGIBOT’s Omdia-backed shipment messaging all support the idea that Zhiyuan is among the category’s shipment leaders in China and possibly globally. SV004, SV005, SV008
CV009 Shipment leadership deserves a valuation premium because it can generate more deployment data, customer references, and manufacturing learning than slower peers receive. SV004, SV008, SV019
CV010 Shipment leadership still fails to prove revenue quality, gross margin durability, or customer retention. SV012, SV024, SV025, SV026
CV011 The market case is directionally strong because major analyst houses all model large long-run growth for humanoids, even if the ranges vary wildly. SV004, SV005, SV006, SV007
CV012 IDC, Mordor, and GMI differ sharply on market size, which lowers valuation confidence and argues against over-anchoring on any single TAM number. SV005, SV006, SV007
CV013 Public customer proof meaningfully supports the valuation case because China Mobile, Fulin, and Longcheer move Zhiyuan beyond pure lab narrative. SV024, SV025, SV029
CV014 The PR Newswire “Deployment Year One” framing adds momentum to the commercialization story, but it remains company-authored and should not be over-weighted alone. SV028, SV026
CV015 The right recommendation under current evidence is track rather than chase, because the company is credible but the public underwriting pack is incomplete. SV003, SV012, SV024, SV026
CV016 The right valuation stance under current evidence is stretched rather than absurd, because the company has real proof yet still lacks public financial transparency. SV002, SV012, SV026
CV017 UBTECH is the most useful public operating benchmark because it is a listed Chinese humanoid peer with actual filings. SV012, SV013, SV014, SV015
CV018 UBTECH’s 2025 annual report shows RMB2.0 billion of revenue, a 37.7% gross margin, and a large net loss, which is a powerful reminder of category capital intensity. SV012
CV019 That benchmark implies that a premium private valuation for Zhiyuan should still be discounted for missing revenue, burn, and balance-sheet detail. SV012, SV016
CV020 Figure is the best benchmark for premium narrative power, frontier-model ambition, and flagship-customer branding, not for near-term valuation discipline. SV018, SV019, SV020, SV027
CV021 Apptronik is the best benchmark for how much capital private peers still need even after achieving meaningful enterprise visibility. SV021, SV022, SV023
CV022 Public listed-peer multiples are hard to apply directly because Zhiyuan does not disclose a revenue denominator cleanly enough to support a standard comp table. SV012, SV013, SV026
CV023 That means valuation should lean more on milestone-based underwriting than on false precision from peer-multiple arithmetic. SV012, SV024, SV026
CV024 A milestone-based approach should reward proof on live deployments, repeat orders, margin path, and data-governance readiness rather than just aggregate hype. SV024, SV025, SV028
CV025 The bull case requires Zhiyuan to convert shipment leadership into repeat enterprise expansion, publish stronger economics, and broaden customer proof beyond a few flagship accounts. SV004, SV024, SV025, SV029
CV026 The bull case is also helped if overseas partnerships, open-source credibility, and model-layer advances create a platform premium rather than a hardware-only multiple. SV010, SV011, SV019, SV028
CV027 The base case assumes real commercialization momentum continues, but disclosure improves only gradually and capital intensity remains high. SV003, SV012, SV024, SV028
CV028 In that base case, a watchful “track” recommendation is superior to paying up aggressively on limited data. SV012, SV026
CV029 The bear case assumes orders convert more slowly than expected, service burden stays heavy, and the next financing occurs before public economics are compelling. SV012, SV024, SV025, SV026
CV030 The bear case becomes more plausible if hype and market-forecast dispersion lull investors into confusing TAM with evidence. SV006, SV007, SV026
CV031 IPO optionality is part of the valuation context because public reporting has discussed a Hong Kong path, but it should be treated as a possibility, not as a bankable exit. SV002, SV014, SV015
CV032 A plausible exit set includes a Hong Kong listing, a strategic transaction, or a later growth round at a higher mark if deployment proof compounds. SV002, SV014, SV017
CV033 Exact exit timing and dilution remain too underdisclosed to model with confidence from public sources alone. SV001, SV002, SV012
CV034 The recommendation logic is strengthened by the fact that named-customer proof exists, yet is still too narrow to remove concentration risk. SV024, SV025, SV029, SV026
CV035 The thesis-break triggers are straightforward: failure to turn flagship orders into stable deployments, inability to disclose credible economics, or a need for funding on weak terms. SV012, SV024, SV026
CV036 Another thesis-break trigger would be a regulatory, liability, or data-governance failure that slows enterprise adoption at the exact moment the company needs proof compounding. SV012, SV026, SV028
CV037 The most valuable final diligence asks are management accounts, customer cohorts, gross-margin bridges, support burden, and balance-sheet detail. SV012, SV024, SV025
CV038 Market forecast dispersion should reduce confidence but not erase the thesis, because multiple independent sources still agree the category could be large. SV004, SV005, SV006, SV007
CV039 Peer capital intensity raises dilution risk materially, because even scaled leaders and premium U.S. peers continue to raise or consume large capital pools. SV012, SV021, SV022
CV040 Because Zhiyuan’s public revenue run-rate is not defensible with high confidence, any fair-multiple claim today would be more narrative than finance. SV003, SV012, SV026
CV041 The correct investment posture is therefore selective patience: track the company closely, avoid assuming public marks are obviously cheap, and re-underwrite when economic disclosure improves. SV012, SV026, SV028
来源
编号出版方标题引文
SO001 AGIBOT AGIBOT homepage
SO002 AGIBOT AGIBOT products page
SO003 AGIBOT Research AgiBot Research
SO004 AGIBOT About Us
SO005 AGIBOT AGIBOT initiates the commercial mass production of general robots
SO006 智元创新 关于智元
SO007 智元创新 Leadership page
SO008 智元创新 招贤纳士
SO009 智元创新 远征A3 product page
SO010 智元创新 全球首个!实现自主打乒乓球的全尺寸人形机器人智元远征A3亮相WAIC
SO011 智元创新 WAIC 2026智启具身论坛成功举办,全球顶尖力量同台共探物理AI智能涌现之路
SO012 智元创新 不到3个月再破里程碑!智元第15000台具身机器人下线,创下全球量产新纪录
SO013 智元创新 APC 2026 | 智元邓泰华:万亿级产业如何从三条曲线照进现实
SO014 36Kr 「稚晖君」的机器人公司,京东投了 | 36氪独家
SO015 智东西 智元机器人拟赴港IPO!
SO016 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SO017 腾讯新闻 / 澎湃新闻 智元机器人高层集体亮相,逐条回应“技术路线、模型争议、商业如何落地”
SO018 Forbes Agibot Shipped A Staggering 5,000 Humanoid Robots In The Last 3 Months
SO019 Gasgoo Zhiyuan Robot Spins Off Dexterous Hand Business: Humanoid Robots Trigger a Division-of-Labor Revolution
SO020 腾讯新闻 / 投资时间网 智元机器人成被告,发生了什么?
SO021 企查查 智元创新(上海)科技股份有限公司_招聘信息
SO022 中国政府网 十五部门关于印发《“十四五”机器人产业发展规划》的通知
SO023 新华网客户端 / 工信部 工业和信息化部关于印发《人形机器人创新发展指导意见》的通知
SO024 中国政府网 国务院关于印发新一代人工智能发展规划的通知
SO025 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SM001 AGIBOT AGIBOT homepage
SM002 AGIBOT AGIBOT products page
SM003 AgiBot Research AgiBot Research
SM004 智元创新 关于智元
SM005 智元创新 精灵G1 product page
SM006 智元创新 灵犀X2 product page
SM007 智元创新 远征A3 product page
SM008 智元创新 WAIC 2026智启具身论坛成功举办
SM009 腾讯新闻 / 澎湃新闻 智元机器人高层集体亮相,逐条回应“技术路线、模型争议、商业如何落地”
SM010 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SM011 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SM012 中国政府网 国务院关于印发新一代人工智能发展规划的通知
SM013 中国政府网 十五部门关于印发《“十四五”机器人产业发展规划》的通知
SM014 新华网客户端 / 工信部 工业和信息化部关于印发《人形机器人创新发展指导意见》的通知
SM015 工信部解读 《人形机器人创新发展指导意见》解读
SM016 SESEC China’s First Standards System for Humanoid Robots and Embodied Intelligence
SM017 IDC Worldwide Humanoid Robotics Market Analysis 2026
SM018 Mordor Intelligence Humanoids Market Size, Forecast Report (2026-2031)
SM019 Global Market Insights Humanoid Robot Market Size, Forecasts Report 2026-2035
SM020 Unitree Unitree H1 / H1-2
SM021 UBTECH UBTECH Walker S Industrial Humanoid Robot
SM022 Figure AI F.03 Arrives at BMW
SM023 BMW Group BMW Group advances the use of Physical AI in production with Figure 03 project in Spartanburg
SM024 Apptronik Apptronik - Press Releases
SM025 Agility Robotics Press Releases | Agility
SP001 智元创新 关于智元
SP002 智元创新 精灵G1 product page
SP003 智元创新 灵犀X2 product page
SP004 智元创新 远征A3 product page
SP005 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SP006 36Kr 「稚晖君」的机器人公司,京东投了 | 36氪独家
SP007 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SP008 Figure AI Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation
SP009 Figure AI F.03 Arrives at BMW
SP010 Figure AI F.02 Contributed to the Production of 30,000 Cars at BMW
SP011 BMW Group BMW Group advances the use of Physical AI in production with Figure 03 project in Spartanburg
SP012 UBTECH UBTECH Walker S Industrial Humanoid Robot
SP013 UBTECH About-UBTECH | UBTECH Robotics
SP014 Unitree Unitree H1 / H1-2
SP015 Unitree Unitree H1 (Contact us for the real price)
SP016 Unitree Unitree Dex5-1
SP017 Agility Robotics Press Releases | Agility
SP018 Apptronik Apptronik - Press Releases
SP019 Apptronik Apptronik Closes Over $935 Million Series A
SP020 CNBC Apptronik raises $520 million to beat Chinese humanoids, Tesla Optimus to market
SP021 Gasgoo Zhiyuan Robot Spins Off Dexterous Hand Business: Humanoid Robots Trigger a Division-of-Labor Revolution
SP022 AgiBot Research AgiBot Research
SP023 AGIBOT AGIBOT Unveils Four New Products at WAIC 2026, Showcasing Embodied AI in Real
SP024 腾讯新闻 / 澎湃新闻 智元机器人高层集体亮相,逐条回应“技术路线、模型争议、商业如何落地”
SP025 Forbes Agibot Shipped A Staggering 5,000 Humanoid Robots In The Last 3 Months
SI001 AGIBOT AGIBOT A2-W product page
SI002 智元创新 智元 D1 Ultra page
SI003 智元创新 智元 solutions page
SI004 AgiBot Research AgiBot Research
SI005 智元创新 关于智元
SI006 Yicai Global AgiBot Robot Starts at Fulin Precision Plant, Could Potentially Replace Two Human Workers
SI007 Assembly Magazine Chinese Car Parts Manufacturer Orders 100 AgiBot Humanoid Robots in Landmark Deal
SI008 Shanghai Municipal Government Shanghai-made robots become full-time factory workers
SI009 Yicai Global China Mobile Awards Record USD17 Million Robot Tender to AgiBot and Unitree
SI010 腾讯新闻 / 澎湃新闻 智元机器人高层集体亮相,逐条回应“技术路线、模型争议、商业如何落地”
SI011 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SI012 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SI013 UBTECH Financial Reports | UBTECH Robotics
SI014 UBTECH UBTECH 2025 Annual Report
SI015 UBTECH UBTECH Walker S Industrial Humanoid Robot
SI016 Figure AI BotQ: A High-Volume Manufacturing Facility for Humanoid Robots
SI017 Figure AI F.03 Arrives at BMW
SI018 BMW Group BMW Group advances the use of Physical AI in production with Figure 03 project in Spartanburg
SI019 Apptronik Apptronik Closes Over $935 Million Series A
SI020 CNBC Apptronik raises $520 million to beat Chinese humanoids, Tesla Optimus to market
SI021 Apptronik Apptronik and Mercedes-Benz Enter Commercial Agreement
SI022 Apptronik GXO advances humanoid strategy, announces multi-phase R&D initiative with Apptronik
SI023 Unitree Unitree A2-W
SI024 Unitree Unitree H1 (Contact us for the real price)
SI025 IDC Worldwide Humanoid Robotics Market Analysis 2026
SI026 The Wire China The Robot Reckoning: China’s Humanoid Robots
SE001 智元创新 关于智元
SE002 AgiBot Research AgiBot Research
SE003 AGIBOT AGIBOT X1 open source docs
SE004 GitHub AgibotTech/agibot_x1_infer
SE005 GitHub AgibotTech/agibot_x1_train
SE006 GitHub OpenDriveLab/AgiBot-World
SE007 智元创新 精灵G1 product page
SE008 智元创新 灵犀X2 product page
SE009 智元创新 远征A3 product page
SE010 智元创新 D1 Ultra page
SE011 智元创新 智元绝尘C5
SE012 智元创新 OmniHand O12 page
SE013 AGIBOT APC 2026 overview
SE014 AGIBOT AGIBOT’s WITA
SE015 智元创新 WAIC 2026智启具身论坛成功举办
SE016 Yicai Global AgiBot Robot Starts at Fulin Precision Plant, Could Potentially Replace Two Human Workers
SE017 腾讯新闻 / 澎湃新闻 智元机器人高层集体亮相,逐条回应“技术路线、模型争议、商业如何落地”
SE018 中国政府网 十五部门关于印发《“十四五”机器人产业发展规划》的通知
SE019 工信部解读 《人形机器人创新发展指导意见》解读
SE020 SESEC China’s First Standards System for Humanoid Robots and Embodied Intelligence
SE021 Figure AI Helix Accelerating Real-World Logistics
SE022 Figure AI Project Go-Big: Internet-Scale Humanoid Pretraining and Direct Human-to-Robot Transfer
SE023 UBTECH UBTECH Walker S1 Humanoid Robot
SE024 Unitree Unitree H2
SE025 Apptronik Apollo 2
SE026 Gasgoo Zhiyuan Robot Spins Off Dexterous Hand Business: Humanoid Robots Trigger a Division-of-Labor Revolution
SU001 智元创新 解决方案
SU002 Yicai Global AgiBot Robot Starts at Fulin Precision Plant, Could Potentially Replace Two Human Workers
SU003 Assembly Magazine Chinese Car Parts Manufacturer Orders 100 AgiBot Humanoid Robots in Landmark Deal
SU004 Shanghai Municipal Government Shanghai-made robots become full-time factory workers
SU005 Yicai Global China Mobile Awards Record USD17 Million Robot Tender to AgiBot and Unitree
SU006 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SU007 PR Newswire AGIBOT Declares 2026 Deployment Year One at APC 2026
SU008 The AI Journal AGIBOT and Longcheer Technology Achieve World’s First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line
SU009 AGIBOT APC 2026 overview
SU010 AGIBOT AGIBOT Hosts UK APC2026 in London, Advancing Commercial Deployment of Humanoid Robotics in Europe
SU011 AGIBOT AGIBOT Brings APC 2026 to Australia and New Zealand
SU012 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SU013 Figure AI F.03 Arrives at BMW
SU014 BMW Group BMW Group advances the use of Physical AI in production with Figure 03 project in Spartanburg
SU015 UBTECH UBTECH Humanoid Robot Industrial Application Solution
SU016 UBTECH UBTECH Walker S1 Humanoid Robot
SU017 Apptronik GXO advances humanoid strategy, announces multi-phase R&D initiative with Apptronik
SU018 Apptronik Apptronik and Mercedes-Benz Enter Commercial Agreement
SU019 Apptronik Manufacturing
SU020 Apptronik 3PL
SU021 Figure AI Helix Accelerating Real-World Logistics
SU022 Unitree Unitree R1
SU023 Figure AI Introducing Figure 03
SU024 The Wire China The Robot Reckoning: China’s Humanoid Robots
SU025 IDC Worldwide Humanoid Robotics Market Analysis 2026
SR001 中国政府网 十四五机器人产业发展规划
SR002 新华网 / 工信部 人形机器人创新发展指导意见
SR003 工信部解读 《人形机器人创新发展指导意见》解读
SR004 商务部 商务部公告2026年第23号 公布将10家美国实体列入出口管制管控名单的决定
SR005 National People’s Congress Data Security Law of the People's Republic of China
SR006 A&O Shearman China Cybersecurity Law Amendments 2026: Key AI & Compliance Changes
SR007 Chambers & Partners Data Protection & Privacy 2026 - China
SR008 Chambers & Partners Product Liability & Safety 2026 - China
SR009 Hill Dickinson Humanoid robots and the law - preparing for a new era of risk
SR010 Mayer Brown China Expands Its Playbook: New Industrial Supply Chain and Counter-Extraterritoriality Regulations Create Direct Compliance Conflicts for Multinationals
SR011 Arnold & Porter China Imposes Export Control and Government Procurement Restrictions on Designated U.S. Companies
SR012 Recording Law China Data Privacy Laws: PIPL, CSL & DSL Compliance Guide (2026)
SR013 SESEC China’s First Standards System for Humanoid Robots and Embodied Intelligence
SR014 Yicai Global AgiBot Robot Starts at Fulin Precision Plant, Could Potentially Replace Two Human Workers
SR015 腾讯新闻 / 澎湃新闻 智元机器人高层集体亮相,逐条回应“技术路线、模型争议、商业如何落地”
SR016 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SR017 The Wire China The Robot Reckoning: China’s Humanoid Robots
SR018 Gasgoo Zhiyuan Robot Spins Off Dexterous Hand Business: Humanoid Robots Trigger a Division-of-Labor Revolution
SR019 Yicai Global China Mobile Awards Record USD17 Million Robot Tender to AgiBot and Unitree
SR020 PR Newswire AGIBOT Declares 2026 Deployment Year One at APC 2026
SR021 AGIBOT APC 2026 overview
SR022 The AI Journal AGIBOT and Longcheer Technology Achieve World’s First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line
SR023 UBTECH UBTECH 2025 Annual Report
SR024 UBTECH UBTECH Humanoid Robot Industrial Application Solution
SR025 Figure AI F.03 Arrives at BMW
SR026 BMW Group BMW Group advances the use of Physical AI in production with Figure 03 project in Spartanburg
SR027 Apptronik Apptronik and Mercedes-Benz Enter Commercial Agreement
SR028 Apptronik Welcome to Robot Park, where Apptronik’s Apollo goes to work
SR029 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SR030 IDC Worldwide Humanoid Robotics Market Analysis 2026
SV001 36Kr 「稚晖君」的机器人公司,京东投了 | 36氪独家
SV002 智东西 智元机器人拟赴港IPO!
SV003 腾讯新闻 / 华夏时报 智元机器人闯关实录:斩获中移动及奇瑞大单后,用开源生态撬动机器人产业
SV004 TrendForce China’s Humanoid Robot Output to Surge 94% in 2026; Unitree and AgiBot to Capture Nearly 80% Market Share, Says TrendForce
SV005 IDC Worldwide Humanoid Robotics Market Analysis 2026
SV006 Mordor Intelligence Humanoids Market Size, Forecast Report (2026-2031)
SV007 Global Market Insights Humanoid Robot Market Size, Forecasts Report 2026-2035
SV008 AGIBOT Omdia Ranks AGIBOT No.1 Worldwide in Humanoid Robot Shipments in 2025
SV009 AGIBOT About Us
SV010 AGIBOT AGIBOT X1
SV011 36Kr Europe Dissecting "ZhiYuan Robotics": The Capital Chess Game and the "Huawei Affiliated" Operators
SV012 UBTECH UBTECH 2025 Annual Report
SV013 UBTECH UBTECH 2025 Interim Report
SV014 UBTECH Listed on the Main Board of the HKEX
SV015 UBTECH Officially Included in HKEX Tech 100 Index
SV016 UBTECH UBTECH partners with collaborators to accelerate global expansion, Bringing humanoid robots to European retail logistics environments
SV017 UBTECH UBTECH and Hitachi Enter into Strategic Partnership to Jointly Explore Intelligent Solutions Across Multiple Fields
SV018 Figure AI Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation
SV019 Figure AI Introducing Helix 02: Full-Body Autonomy
SV020 Figure AI Introducing Figure 03
SV021 Apptronik Apptronik Closes Over $935 Million Series A
SV022 CNBC Apptronik raises $520 million to beat Chinese humanoids, Tesla Optimus to market
SV023 Apptronik Apptronik and Mercedes-Benz Enter Commercial Agreement
SV024 Yicai Global China Mobile Awards Record USD17 Million Robot Tender to AgiBot and Unitree
SV025 Yicai Global AgiBot Robot Starts at Fulin Precision Plant, Could Potentially Replace Two Human Workers
SV026 The Wire China The Robot Reckoning: China’s Humanoid Robots
SV027 BMW Group BMW Group advances the use of Physical AI in production with Figure 03 project in Spartanburg
SV028 PR Newswire AGIBOT Declares 2026 Deployment Year One at APC 2026
SV029 The AI Journal AGIBOT and Longcheer Technology Achieve World’s First Embodied AI Deployment in Consumer Electronics Precision Manufacturing Mass-Production Line
SV030 Unitree Unitree H1 (Contact us for the real price)