初创公司尽调
尽调报告 Robotics / Hardware Series A / Pre-A completed 2026-08-30

Xingyuanzhi Robotics

面向人形机器人的北京 AI 大脑平台

Xingyuanzhi 的具身大脑逻辑可信,也有真实推进,但公开披露仍然太薄,支撑不了在独角兽区间不看价格的投资结论。

封面要素

成立时间 01
2025-08-01 [CO001]
总部 02
Beijing, China [CO003]
具名客户证明 05
AgiBot / Beijing Yizhuang / Zhongli-EP [CU002, CO028]
公开团队估计 06
50 employees+ [CO042]

公司概况

Xingyuanzhi Robotics 是一家位于北京的具身智能初创公司,由 Beijing Academy of Artificial Intelligence 孵化,重点卖机器人“大脑”层,而不是成品机器人本体。它对外讲的产品线围绕 T5 系列和更新的边缘计算平台、多模态空间智能,以及一套希望跨人形和工业机器人形态复用的通用具身大脑栈展开。AgiBot、Beijing Yizhuang Robot 以及 Zhongli/EP 工作流等公开可见交易对手,让公司具备战略意义;但和叙事强度相比,它的商业质量仍更难验证。

官网
xyz-eai.com
成立时间
2025-08-01
创立地点
Beijing, China
总部
Beijing, China
产品
T5 及后续具身大脑计算平台把边缘计算、具身模型和多模态空间智能组合起来,用于支撑人形与工业机器人场景。
客户
人形机器人 OEM、工业自动化伙伴,以及集中在中国、尤其是北京机器人集群内的公共部门或生态运营方。
商业模式
B2B 大脑层平台,靠控制器硬件、具身模型 / 软件附加、集成工作和部署支持变现,而不是销售成品机器人本体。
阶段
Series A / Pre-A completed
融资情况
到 2026 年 6 月,公司和媒体来源称其在天使轮、天使+ 轮和 Pre-A 轮累计融资约 RMB1 billion,使 Xingyuanzhi 进入独角兽区间,但完整投资条款清单仍未披露。
[CO001, CO002, CO003, CO015, CO019, CO022, CI001, CI002]

执行摘要

主要优势

  • 明确押注大脑层,让 Xingyuanzhi 有机会成为横向平台;只要嵌入多个机器人 OEM,上行空间就能打开。
  • BAAI 孵化加上早期快速融资,给公司带来技术背书、人才信号和战略势能。
  • 公开点名的 AgiBot、Beijing Yizhuang Robot 以及 Zhongli-EP 相关工作流,说明公司接入了真实机器人生态,而不只是实验室 demo。
  • 中国 2026 年具身 AI 政策和融资环境,继续托住需求培育和商业化试验。

主要风险

  • 收入、毛利率、烧钱速度和股权结构条款均未披露,投资判断信心受限。
  • 依赖 NVIDIA 级算力和合作伙伴机器人本体,带来供应链和伙伴内化风险。
  • 客户证明有战略价值,但仍然集中、间接,缺少留存和合同经济性证据。
  • 如果部署证明滞后,或具身 AI 情绪降温,独角兽区间定价可能很快被压缩。
  • 安全、隐私和信任文档明显比公司叙事薄。

未决问题

  • 最新一轮准确投后估值、清算优先权、反稀释条款和投资人权利堆叠。
  • 当前收入运行率、毛利率、现金消耗、积压订单质量和现金跑道。
  • AgiBot、Yizhuang、Zhongli-EP 及其他主要客户的集中度、合同条款和续约行为。
  • 装机量、在线率、附着率、切换成本等部署 KPI 证据。
  • 受控组件的算力 BOM、ECCN 分类和替代方案。

目录

Chapter 01

01公司概况

1.1 身份、定位与落点

作为一家如此年轻的机器人公司,Xingyuanzhi 的公开身份少见地清晰。它在中文官网、英文官网和更长的公司简介中,都把自己描述为一家位于北京、成立于 2025-08-01、由 Beijing Academy of Artificial Intelligence 孵化的具身智能公司。核心卖点不是造完整机器人,而是提供“大脑”层:多模态空间智能、具身基础模型栈,以及可嵌入不同机器人本体的边缘计算。这个定位很关键,因为 Xingyuanzhi 更像大脑优先的平台供应商,而不是资本开支很重的全栈硬件 OEM。公开地址证据指向海淀联络办公室和亦庄注册主体,也符合一家横跨北京科研与产业集群的公司形态。官方页面还显示产品迭代很快:最初的 T5 控制器撑起 2025 年发布故事,而当前产品页面已突出更新的 N5 和 BotPack 式边缘平台。因此,尽管许多经营指标披露仍薄,Xingyuanzhi 看起来已经更像使能型中间件和计算供应商,而不是单一产品初创公司。[CO001, CO002, CO003, CO004, CO005, CO019]

快照 KPI 表 — Xingyuanzhi Robotics(运行日期 2026-08-30)
指标数值 / 状态日期 / 期间置信度缺口 / 尽调备注
成立2025-08-01历史官方中文和英文资料以及 Baike 彼此印证。
孵化BAAI / Zhiyuan 孵化历史至今机构纽带明确,但具体 IP 和治理关系未披露。
所在地北京;海淀联系办公室和亦庄注册主体当前公开材料显示,公司更像是研发和产业足迹分开布局,而不是披露了单一清晰的总部园区。
核心定位只做机器人大脑的具身 AI 与边缘计算供应商当前多个来源一致显示,公司并未把自己定位成完整机器人本体 OEM。
最新轮次Pre-A2026-06-03日期和轮次标签得到官方资料及 36Kr 印证。
累计融资~RMB1 billion截至 2026-06多处反复出现,但没有公开股权结构表或融资工具组合。
已披露客户 / 伙伴AgiBot、Beijing Yizhuang Robot、Zhongli / EP Equipment 关系2025-2026公开材料没有完全量化每段关系的深度。
公开收入指标>RMB10 million(低置信度第三方说法)2025未找到官方或一线来源的收入披露。
公开员工数指标~50 名员工,90%+ 研发(低置信度第三方说法)2026 年中未找到官方员工数披露。
轮次估值未公开披露当前公开来源没有给出定价轮投后估值。

官方公司页面锚定了身份、产品定位和融资情况。收入和员工数行依赖声誉较低的第三方摘要,只应视为方向性信息。

[CO001, CO002, CO003, CO004, CO005, CO014]
FO003: 快照 KPI 与状态标记

截至 2026-08-30,Xingyuanzhi 最清晰的公开指标和状态标记。

融资和订单数字来自公开报道,但确认收入和员工数来源仍弱,不应当作经审计经营指标。

[CO001, CO002, CO015, CO028, CO041, CO042]

1.2 创始团队、治理与机构网络

领导层质量是公司公开可见资产中最强的一项。官方和第三方来源一致称,创始人兼 CEO Liu Dong 曾负责 JD.com 智能驾驶业务,这让 Xingyuanzhi 拥有一位已经管理过自动驾驶产品、推动过商业部署,并处理过硬件与软件系统级取舍的负责人。BAAI 自己的会议议程更进一步,把 Liu Dong 同时列为公司 CEO 和具身大脑研究中心 PI,说明这家公司仍与孵化它的机构紧密绑定。其他具名技术负责人包括联合创始人 Mu Yadong,公开资料称其为 Peking University 研究员和 Zhiyuan 学者;以及联合创始人 Sun Zhenguo,官方简介称其曾公开发布 ω-EVA。更不清楚的是治理结构。已审阅来源没有披露董事会、委员会设置或战略投资人的控制权。这留下一个常见的风险投资阶段图景:创始人市场匹配度和机构背书很强,但治理透明度落后于融资速度,需要直接公司尽调才能评估到位。[CO006, CO007, CO008, CO009, CO010, CO011]

领导层与创始人表
人物 / 领域公开角色背景或能力信号重要性披露缺口 / 依赖
Liu Dong创始人兼 CEO前 JD.com 智能驾驶总经理;BAAI 具身大脑项目负责人带来自动驾驶系统经验、商业化语境,以及与 BAAI 的直接连接关键人物集中度高;更广管理梯队只披露了一部分
Mu Yadong联合创始人 / 研究员Peking University 研究员,Zhiyuan 多模态和具身 AI 学者体现研究深度和学术招聘能力公开来源未披露运营职责或持股
Sun Zhenguo联合创始人 / 技术负责人官方资料称其在 BAAI 2026 公开发布 ω-EVA显示 CEO 之外还有可见技术领导力已审阅来源中的公开履历有限
董事会 / 正式治理未公开披露已审阅来源没有列出完整董事会、委员会或观察员结构公司同时混合风险资本、国资和战略资本,治理因此很关键需要直接尽调控制权和董事会构成

本表只覆盖已具名领导者,以及公开材料中可见的一项明确治理缺口;并不是完整组织架构图。

[CO006, CO007, CO008, CO009, CO010, CO011]

1.3 资金底座与投资人结构

融资叙事是 Xingyuanzhi 成为中国 2026 年具身 AI 浪潮关注对象的最直接原因。公开材料显示,公司在 2025 年 9 月完成 RMB200 million 天使轮,2025 年 12 月完成超过 RMB100 million 的天使+ 轮,并在 2026-06-03 宣布 Pre-A 轮。公司页面、36Kr、Pandaily 和其他摘要反复出现同一个标题:成立后头十个月累计融资约 RMB1 billion。投资人名单几乎和融资总额一样重要。天使轮混合了风险资本和 AgiBot、Zhongli 等战略支持方;Pre-A 则集合了财务投资人、国资背景基金,以及 Beijing Industrial Investment、CRRC Capital 等产业资本。36Kr 和官方公司简介还称 Yuansheng 连续跟投三轮,暗示至少有一位高确信度投资人愿意持续押注执行。需要警惕的是,融资速度跑在披露质量前面。公开来源给出投资人名称和高层次资金用途,但没有披露逐轮投后估值、股权结构表、清算优先权、债务或老股交易。投资人显然在为战略可选性付费,但精确的定价架构仍是私有信息。[CO012, CO013, CO014, CO015, CO016, CO017]

利益相关方或投资人图谱
利益相关方角色经济重要性仍需尽调的问题
BAAI / Zhiyuan Research Institute孵化方和持续支持方提供研究可信度、人才漏斗和公信力厘清 IP 归属、许可权,以及任何研究院治理权
CAS Star 与 Hillhouse Ventures领投的天使投资人在产品规模化验证前就展示了早期机构信念确认当前持股比例和按比例跟投权
AgiBot 与 Zhongli(天使轮战略投资人)天使轮产业资本参与方创造早期客户或渠道路径,并传递生态认可将纯信号价值与带收入的商业承诺分开
SAIF Fund 与 Kailian CapitalAngel+ 共同领投方帮公司从原型叙事过渡到规模化商业化叙事厘清 Angel+ 是否把估值大幅推高到天使轮之上
Yuansheng Venture Capital三轮连续支持方显示信念,也可能锚定后续内部治理决策确认董事席位、观察员身份和后续跟投经济性
Beijing Industrial Investment 与 CRRC CapitalPre-A 轮国资关联领投方可能加速产业试点、制造合作和政策准入厘清里程碑约束、采购预期和政治依赖
Songhe、Creation Capital、Huakong、Guojun Innovation、Jiangxi Financial、Aiteke、Hengxing、Qi’an 等投资人Pre-A 财团成员将资本基础扩展到财务、国资和产业资金池需要股权结构表细节、清算优先权,以及任何战略商业挂钩安排

公开来源披露了投资人名单,但没有披露持股、各轮估值或治理权,因此这里是利益相关方图谱,而不是真正的股权结构表摘要。

[CO002, CO012, CO013, CO014, CO015, CO016]

1.4 里程碑、客户与披露风险

对一家成立还不到一年的公司来说,Xingyuanzhi 的里程碑曲线异常密集。官方页面把 2025 年与 T5 发布及其早期接入 AgiBot 的 Genie G2 联系起来,又把 2026 年与 ω-EVA 发布、BAAI 牵头的世界模型实验室、Hannover Messe、Fortune China Tech 50 入选,以及面向 Zhongli 或 EP Equipment 的装卸解决方案全球商业化推出联系起来。客户图景也强于许多早期具身 AI 初创公司:公开材料直接点名 AgiBot 和 Beijing Yizhuang Robot,把至少 RMB500 million 的三年订单目标附在 Yizhuang 关系上,并描述覆盖政务服务、巡检、导航和物流的部署。不过,同一批来源也暴露出本报告最大的尽调约束。公开牵引力数据很薄,质量也参差。低信誉摘要称 T5 出货数百台、收入超过 RMB10 million、团队约 50 人,但主流和官方材料没有进一步给出确认收入、在手订单转化、毛利率或核验员工数。在 2026 年这个资本充裕但商业证明稀缺的赛道里,这个披露缺口是实质问题,不是表面瑕疵。[CO019, CO020, CO021, CO026, CO027, CO028]

里程碑表
日期事件类型金额 / 状态参与方含义
2025-08-01公司成立成立公司设立Xingyuanzhi / BAAI开启了一条非常快的融资时间线
2025-09-01与 Beijing Yizhuang Robot 达成战略合作合作三年订单目标 ≥ RMB500MXingyuanzhi、Beijing Yizhuang Robot锚定最早具名商业关系之一
2025-09-10披露天使轮融资RMB200MCAS Star、Hillhouse、Yuanhe、Yuansheng、战略投资人显示投资人立即认可只做大脑层的投资逻辑
2025-10 至 2025-11T5 与 AgiBot Genie G2 关联,随后在 Baidu World 2025 展示产品T5 / Jetson Thor / 2070 TFLOPSXingyuanzhi、AgiBot为产品拼出第一条可见部署叙事
2025-12-11披露 Angel+ 轮融资>RMB100MSAIF Fund、Kailian、跟投方在 2026 年商业化推进前延长现金跑道
2026-04-22Hannover Messe 亮相并发布 BotPack B规模化国际展会首秀Xingyuanzhi传递海外野心和更快硬件迭代
2026-06-03宣布 Pre-A 轮融资Pre-A;累计 ~RMB1BBeijing Industrial Investment、CRRC、Songhe 等补强资产负债表,并放大招聘 / 研发说法
2026-06-05 至 2026-06-17BAAI 大会、ω-EVA 发布,以及世界模型实验室获批产品世界模型和实验室里程碑Xingyuanzhi、BAAI把叙事从控制器供应商推向具身世界模型平台
2026-06-25 至 2026-06-30入选 Fortune China Tech 50,并全球销售 Zhongli 装卸解决方案规模化认可叠加商业推出Xingyuanzhi、Zhongli / EP增加外部可见度和一个更具体的商业化证明点
2026-08-11 起官方新闻页列出 World Robot Conference 2026 展示规模化展示 RoboBrain Pro、ω-EVA、XrossXingyuanzhi显示公司在报告日期附近仍持续投入公开生态站位

日期来自官方公司页面和 Baike。T5 发布窗口在英文时间线与后续中文摘要之间存在部分冲突,因此本时间线采用更宽的 10-11 月商业化窗口,而不是过度声称某一天。

[CO001, CO012, CO013, CO014, CO019, CO020]
FO001: 公司里程碑时间线——从创立到世界模型商业化推进

这条精选时间线呈现几项关键里程碑,解释 Xingyuanzhi 如何在一年内从孵化项目变成融资密集的具身大脑平台叙事。

[CO001, CO012, CO013, CO014, CO019, CO028]
FO002: 公司快照逻辑——BAAI 研究、边缘计算与多本体客户

Xingyuanzhi 的孵化背景、负责人履历、边缘硬件、世界模型栈和具名客户,如何串成当前业务逻辑。

[CO002, CO005, CO007, CO022, CO025, CO028]

1.5 要点

Chapter 02

02市场分析

2.1 市场边界、纳入支出与替代方案

Xingyuanzhi 的相关市场比“人形机器人”这个热词口径更窄,也比单一控制器 SKU 更宽。公司和媒体公开来源持续把 Xingyuanzhi 描述为机器人“大脑”供应商:具身基础模型、世界模型软件、边缘推理,以及在不同机器人本体上运行这些系统所需的控制器硬件。这意味着公司参与的是硅片与整机 OEM 收入之间的一层。纳入支出因此包括智能栈、边缘计算、集成和部署支持,它们让机器人能在真实工作流中可用。排除支出包括机器人本体制造、通用执行器,以及大规模场地改造项目;大脑层供应商无法直接捕获这些支出。最接近的替代品不只是其他大脑供应商,还包括固定式工业机器人、AMR、四足机器人、自建自动化团队和人力。Xingyuanzhi 要赢,买方必须相信,外包具身智能能达到可接受的延迟、安全性和成本,同时不牺牲对成品机器人体验的控制。[CM001, CM002, CM003, CM004, CM005, CM036]

市场定义表
细分 / 类别纳入支出排除支出主要替代方案买方 / 付款方对 Xingyuanzhi 的意义
具身大脑平台控制器硬件、具身模型、边缘推理、部署集成机器人本体制造和工厂改造自建自动化团队机器人 OEM 产品 / 研发预算直接命中的核心市场
工业具身部署集成、维护、工作流调校、边缘计算与机器人智能无关的通用工厂自动化固定式工业机器人工厂自动化 / 运营预算近期商业化路径
政务服务 / 城市试点试点部署、维护、场景适配无关智慧城市基础设施人工服务人员、简单服务机器人地方政府采购有助于演示和早期客户背书,但不一定是最大的收入池
巡检和能源运营机器人大脑技术栈、感知、导航、决策机器人以外的传统 SCADA 或车辆资本开支四足机器人、无人机、人工巡检员公用事业运维预算高价值边缘 AI 的有希望细分
消费级家用机器人潜在后期大脑授权家电制造和零售渠道低成本消费机器人家庭或渠道伙伴预算目前不在证据支撑最强的市场窗口内

本定义把 Xingyuanzhi 视为智能和控制器价值供应商,而不是完整人形机器人本体卖方。

[CM001, CM002, CM003, CM004, CM005, CM015]
FM001: 市场边界金字塔

金字塔显示,庞大的人形机器人市场口径如何压缩成 Xingyuanzhi 窄得多的大脑层机会。

上层是来源自身的 TAM 视角;下层解释,为什么 Xingyuanzhi 真正可销售的市场只是更大机器人市场叙事里的智能切片。

[CM001, CM005, CM015, CM016, CM017, CM036]

2.2 市场规模口径与分析矛盾

具身智能的市场规模记录很拥挤,也不一致,因为不同发布方衡量的东西不同。Goldman Sachs 给出长期全球人形机器人 TAM 下限 US$6 billion,以及 2035 年蓝天上限 US$154 billion;CNBC 则报道 Barclays 的观点,称市场今天约为 US$2-3 billion,到 2035 年可能达到 US$200 billion。IDTechEx 给出更温和的 2036 年结果,接近 US$29.5 billion;Deloitte 则把 2026 年工业用途人形机器人规模放在仅 US$210-270 million,之后到 2032 年可能升至 US$600 million-US$1 billion。这些差异并非错误,更像口径错位:有的模型衡量硬件收入,有的衡量未来经济价值,有的衡量当前工业出货。IFR 的工业机器人基线和 EmbodiedGlobal 的资本流数据,有助于圈定更近端的机会。中国已经拥有全球最大的工业机器人装机基础,也拥有具身 AI 中最激进的资本集中度,这意味着面向中国、服务大脑层供应商的可服务市场确实存在。仍缺少的是,针对机器人“大脑”层本身的独立公开 TAM 或 SAM。[CM006, CM007, CM008, CM009, CM010, CM011]

TAM/SAM/SOM 或规模测算视角表
发布方 / 视角年份 / 展望期地域数值方法 / 单位置信度局限
Goldman Sachs 基准情景10-15 年 / 2035全球US$6B 底线;US$154B 乐观上限长期硬件市场情景情景区间很宽;并非针对 Xingyuanzhi
CNBC / Barclays2026 至 2035全球当前 US$2-3B;2035 年 US$200B主题市场预测未与 Goldman 或 IDTechEx 直接对齐
IDTechEx2036全球~US$29.5B10 年市场预测衡量的是人形机器人市场,而不是大脑层
Deloitte2026 和 2032全球工业用途人形机器人2026 年 US$210-270M;2032 年 US$600M-US$1B出货量和 ASP 情景只覆盖更窄的工业用途切片
IFR 工业基线2024全球 / 中国全球装机 542k;中国装机 295k已安装工业机器人是装机基数代理指标,不是具身大脑需求
EmbodiedGlobal 资本视角H1 2026中国322 笔交易共 RMB93.5B融资流 / 估值活动资本流入不等于终端市场收入
本章限定的 SAM2026-2028中国来自 OEM、巡检和工业试点的大脑层需求证据边界内的定性 SAM公开资料没有单独测算该切片的收入池

本表保留相互矛盾的市场视角,而不是强行给出虚假的单点估计。最后一行是定性的,因为没有公开来源干净地测算具身大脑品类。

[CM006, CM007, CM008, CM009, CM010, CM011]
FM002: 市场估算区间

区间图保留乐观主题预测、工业用途切片和中国融资强度之间的差距。

前四行以 USD billions 计,最后一行以 RMB billions 计;纳入后者是为了保留资本强度背景,而不是为了严格单位可比。

[CM006, CM007, CM008, CM009, CM012, CM037]

2.3 买方分层与采用路径

相比精确 SAM,Xingyuanzhi 的实际买方地图更清楚。公开来源至少指向四类买方。第一类是 AgiBot 等人形或具身机器人 OEM,它们想要高性能大脑层,但不想每个组件都自己造。第二类是 Zhongli 或 EP 等工业机器或车辆制造商,它们可以把大脑栈与叉车、移动搬运设备或仓库工作流结合。第三类是 Beijing Yizhuang Robot 等地方政府或开发区运营方,它们关心服务场景、示范部署和生态搭建。第四类是运行巡检和维护任务的公用事业或工业运营商。各细分市场的预算归属不同,但模式一致:付款方通常不是终端消费者,而是希望快速获得可部署能力的 OEM 产品团队、运营高管、自动化经理或公共部门采购部门。劳动力短缺、多 SKU 工作流和低延迟边缘控制需求拉动采用。最先形成有意义部署的垂直场景仍是制造、物流和巡检;家庭或广泛消费者服务场景还更远。[CM017, CM018, CM019, CM020, CM021, CM031]

细分 / 买方图谱
细分主要买方主要用户付款方 / 预算所有者工作流采用触发因素
人形机器人 OEM机器人产品和研发团队机器人开发者和运营方产品 / 平台预算把控制器和具身模型集成进机器人本体不自建全栈,也要快速上市
工业移动 / 搬运 OEM自动化或车辆制造商仓库或工厂运营方工业设备或自动化资本开支叉车、物料搬运、装卸在人造设施里跑通多步骤自主能力
政府服务运营方地方政府 / 开发区运营方公共服务团队公共采购 / 试点预算导览、巡检、清洁、展示型部署需要可见的创新和生态建设
能源 / 公用事业巡检公用事业运营管理层巡检人员和远程操作员运维预算巡检、维护、安全监测降低人员暴露风险,提高可用时长
科研 / 标准生态研究院、实验室、培训中心研究人员、开发者、数据团队研发补助或创新预算基准测试、数据生成、训练、验证需要模型、算力和具身数据基础设施

买方、用户和付款方常常分属不同组织;通常由 OEM 或运营负责人签单,而不是机器人产出的最终消费者。

[CM017, CM018, CM019, CM020, CM027, CM032]
FM003: 买方 / 细分市场图

矩阵连接与机器人“大脑层”供应商最相关的买方细分:用户画像、付款方和采用触发点。

[CM017, CM018, CM019, CM026, CM027, CM032]
FM004: 采用漏斗或价值链图

从广义机器人装机基数一路收窄,指向最可能率先采用外部具身大脑栈的买方。

[CM010, CM011, CM012, CM015, CM017, CM019]

2.4 政策驱动、集群效应与采用约束

中国政策环境对具身智能支持力度异常强,但这不是商业化免死金牌。RobotToday 对第十五个五年规划的梳理显示,具身智能被抬升到顶层产业政策位置,围绕模型开发、训练场、零部件和部署给出指令性语言。HEIS 2026 则把术语、接口、智能计算要求、部件、系统和安全规则标准化,补足这一推力。北京 E-Town 则拿出办公空间、算力、试点活动、融资撮合和产业订单激励,提供具体的本地市场支持。这些力量让北京成为 Xingyuanzhi 这类公司的尤其强势起步市场。与此同时,最难的壁垒仍是运营问题,而非口号问题。Deloitte 反复强调数据质量、互操作性、网络风险和工人安全;其汽车分析还称,当下部署仍集中在重复、低变化任务上。Chinabizinsider 也提出资本市场警告:融资和独角兽诞生速度已经跑在披露经济性前面。因此,市场机会真实存在,也由中国引领,但仍受 ROI 证明、集成复杂度和定价可见度不确定约束。[CM022, CM023, CM024, CM025, CM026, CM027]

增长驱动与约束表
驱动 / 约束方向时间影响尽调追问
“十五五”规划支持正向2026-2030政策支持应会扩大试点、采购和训练基础设施政策支持中,哪些部分会真正转成面向大脑层供应商的采购订单?
HEIS 2026 标准正向2026 年起协调成本降低、互操作性更清晰,可以加快部署哪些接口和认证对 Xingyuanzhi 产品最关键?
北京 / 亦庄集群正向当前本地生态支持让人才更容易获取,试点密度也更高当前需求有多少来自特定集群,而不是全国性需求?
中国成本优势正向当前至中期国内厂商可能更早凭性价比和部署速度胜出换到中国以外、面对不同合规规则时,这一优势还能守住吗?
数据与互操作性缺口负向当前数据质量和集成偏弱,会拖慢泛化,也抬高部署成本哪些数据集或中间件证据能证明 Xingyuanzhi 可跨不同本体泛化?
网络安全与安全风险负向当前关键环境里的信任要求会拖慢采购已有怎样的安全边缘架构和安全案例?
任务变化度与 ROI 不确定性负向当前人形机器人目前仍最适合重复、低变化任务,短期 TAM 释放受限当前哪些场景的稼动率和回本指标最强?
资本市场亢奋负向当前至 2028 年独角兽数量和融资热度可能跑在真实收入前面,估值下调融资风险上升哪些公司正把试点转成经常性收入,而不是只会融资?

市场靠政策和资本向前推,但工程现实和采购信任仍在卡关。

[CM013, CM022, CM023, CM024, CM025, CM026]

2.5 要点

Chapter 03

03竞争格局

3.1 竞争版图与边界逻辑

不应把 Xingyuanzhi 放在中国所有人形机器人公司旁边,假设它们都在同一个平面市场竞争。它最核心的战略选择,是销售外部具身大脑层,而不是成品机器人本体。这会把直接同业缩小到那些试图掌握机器人认知、世界模型、边缘控制器,或软硬件一体智能栈的公司。因此,Spirit AI、X Square Robot 和 GigaAI 比主要靠成品硬件变现的机器人制造商更适合作为直接参照。与此同时,更宽的竞争场也很重要,因为买方可以用 AMR、四足机器人和固定自动化等更简单的替代方案解决同样任务。AgiBot 和 Astribot 这类全栈机器人 OEM 尤其重要,因为它们既可能是伙伴,也可能在未来替代 Xingyuanzhi。这个混合场意味着,Xingyuanzhi 同时在和直接大脑供应商、垂直整合 OEM、NVIDIA 等既有厂商,以及不那么炫目但更容易采购的现状自动化竞争。[CP001, CP002, CP003, CP004, CP013, CP037]

竞争对手画像表
竞争对手类别规模 / 融资目标细分市场差异化局限
Spirit AI直接同业 / 全栈具身 AI据报道累计融资近 RMB2B;覆盖北京、杭州、深圳工业人形机器人和具身 AI公开工业部署证据强,还拥有模型和机器人全栈垂直一体化模式与中立供应商逻辑冲突;定价不透明
X Square Robot直接同业 / 具身模型平台留存来源未清晰披露融资具身基础模型和通用机器人研究节奏可见,开源姿态明确,世界模型深度较强商业部署证据不如 Spirit 的留存证据明确
GigaAI直接同业 / Physical AGI风险资本支持较强;RobotToday 报道 Series B2 轮约 RMB1B家庭和工业机器人世界模型定位和早期家庭试验偏家庭的方向与 Xingyuanzhi 近期工业切入点契合度较低
Astribot相邻全栈 OEM有投资人支持的深圳创业公司家庭和轻商业操作DFAI 架构,双臂操作速度很快更偏硬件,不如纯大脑供应商模式可比
AgiBot客户 / 相邻竞争者高知名度人形机器人 OEM人形机器人平台验证需求,也可能成为分销伙伴未来可能把核心智能内化
NVIDIA Isaac / Jetson现有平台全球开发者生态和芯片规模AMR、机械臂、人形机器人、机器人开发者可复用仿真、AI、ROS 和边缘计算基础组件支持内部自建,并把通用栈层商品化
EP Automation / XP15现状替代品有公开定价的商业产品仓储运输和装卸流程部署简单、定价公开、ROI 叙事清晰在更广任务上的灵活性低于具身大脑栈

本画像表混合了直接同业、相邻公司、现有平台和替代方案,因为买方不会只在同一类创业公司里比较。

[CP001, CP002, CP003, CP005, CP007, CP008]
FP001: 竞争定位图

按横向供应商纯度(x 轴)和公开部署证据(y 轴)给同行做序位定位。

坐标轴是有证据支撑的序位判断,不是基准测试分数。x 越高,供应商姿态越中立;y 越高, 公开工作流证据或企业就绪度越强。

[CP001, CP005, CP007, CP008, CP009, CP010]

3.2 同业画像、产品范围与能力压力

在直接同业中,Spirit AI 的公开工业证明最强。它把 VLA 基础模型叙事与 Moz 人形机器人平台结合,并声称在 CATL 电池产线上完成全球首个大规模部署;这种具体生产证据,Xingyuanzhi 尚未发布。X Square Robot 在研究速度和公开技术输出上更强,它用官方研究页面展示世界模型、具身基础模型,以及数百次匹配的仿真到真实落地。GigaAI 看起来更偏向 Physical AGI,并覆盖家庭和工业场景;Astribot 则更明确地围绕轮式双臂家庭助手形态推进硬件。这些差异很重要,因为它们说明买方真正比较的未必只是“大脑质量”,而是可部署具身深度、训练数据获取能力,以及智能层能在物理工作流中跑通的证据。Xingyuanzhi 的差异化仍是中立性,但只有当客户相信它在时间、成本或可靠性上胜过自建或垂直整合时,这种中立性才守得住。[CP005, CP006, CP007, CP008, CP009, CP014]

功能 / 能力矩阵
采购标准XingyuanzhiSpirit AIX Square RobotGigaAIAstribotNVIDIA / Isaac
外部纯大脑供应模式混合混合混合n/a
自有机器人本体 / 数据闭环未知 / 公开证据有限
公开工业部署证据部分部分部分部分平台,不是部署供应商
开放研究 / 开发者可见度公开证据有限部分部分部分
定价透明度unknownunknownunknownunknownunknown部分
本土市场政策与生态契合度中性
与 OEM 客户发生渠道冲突的风险逻辑上低 / 现实风险高

留存来源不足以支持对整个竞品集做标准化数字打分,因此单元格有意使用定性标签。

[CP005, CP007, CP008, CP009, CP010, CP014]
FP002: 功能广度 / 能力图

定性梳理保留来源中,哪些竞争者最清楚地把模型、本体、验证和开放性拼在一起。

单元格反映保留公开来源能看见什么,不代表隐藏的内部能力。未知项是刻意保留,不能解读为弱项。

[CP005, CP007, CP008, CP009, CP017, CP025]
FP003: 护城河 / 就绪度 KPI

选取竞争耐久度指标,说明赛道为什么有吸引力却已很拥挤。

KPI 组合混合产品、融资和部署指标,突出相对就绪度和资本强度,而不是给出单一赢家评分。

[CP006, CP007, CP008, CP009, CP011, CP018]

3.3 定价、分发能力与切换摩擦

整个具身 AI 领域的公开定价透明度都很弱,而这本身就有战略意义。保留下来的竞争对手来源很少披露标价、合同结构或维护条款,说明市场仍主要靠战略合作、试点和协商式企业销售运转,而不是 SKU 级比价。最好的公开对照来自相邻的仓库自动化供应商,例如 EP:它们公布租赁和购买价格,承诺一周交付或一天实施,甚至声称一年内回本。这种清晰度会压迫任何服务相似物流任务的具身智能供应商。因此,Xingyuanzhi 的切换摩擦不能只来自价格不透明。它必须来自已经安装的工作流、边缘集成、模型调优、安全验证,以及客户在嵌入大脑栈后更换供应商所需付出的努力。多供应商并用同样是严肃风险,因为买方可以组合外部供应商、NVIDIA 基础组件和内部工程,而不必永远标准化到单一供应商。[CP010, CP011, CP012, CP017, CP018, CP019]

定价 / 包装对比
供应商价格 / 单位 / 合同模式包含能力折扣 / 未知项影响
Xingyuanzhi未披露控制器加具身模型栈和集成标价、软件附加率和服务费未知很难和更简单的替代品对标
Spirit AI未披露的战略企业销售基础模型加 Moz 机器人和部署支持留存来源未见公开定价竞争点可能是证据和全栈结果,而不是标价
X Square Robot未披露具身模型、研究产出、机器人系统留存来源未见公开定价技术可信度可能比公布价格更重要
GigaAI未披露世界模型栈和自有机器人留存来源未见公开定价家庭与工业混合,增加可比难度
Astribot未披露完整机器人加集成 AI 系统留存来源未见公开定价大概率按高端硬件 + 软件系统销售
EP XP15 标准包995€/mo 租赁,或 €25,000 购买 + €1,500 设置费AMR / 托盘车、App、路线设置高级包定制另计为仓储任务设定透明 ROI 基准
EP XP15 高级包1500€/mo+,或 €25,000 + 定制费定制模块和支持定制费和模块各不相同相邻买方可先从更便宜、更聚焦的自动化切入
NVIDIA 平台伙伴定价的组件和开发者栈芯片、库、ROS 包、仿真总系统成本取决于集成商选择降低内部自建或伙伴共建替代方案的成本

具身 AI 创业公司普遍定价不透明,这本身就是尽调发现。留存竞品集中,只有相邻替代品披露了产品级公开定价。

[CP017, CP018, CP019, CP020, CP030, CP036]

3.4 护城河耐久性与反向竞争证据

反对 Xingyuanzhi 拥有耐久护城河的最强证据,不是市场没有需求,而是太多资本支持的公司都在攻击同一栈的相邻层。EmbodiedGlobal 和 China Biz Insider 都描述了一个 2026 年拥挤赛场:独角兽密集、大额融资频繁,但可重复商业化证明很薄。在这种环境下,竞争对手可以补贴招聘、试点和客户支持,同时掩盖疲弱的单位经济性。它也提高了机器人 OEM 决定把最具战略意义的大脑层留在内部的概率。NVIDIA 把更多机器人栈做成可复用基础组件,带来另一种护城河压力,降低了内部自建和新进入的成本。如果 Xingyuanzhi 成为嵌入多个 OEM 的中立标准,它仍然可以赢;但公开证据尚未显示排他性客户锁定、基准测试领先或定价权。按当前记录看,公司有差异化定位,还没有被证明拥有护城河。[CP016, CP022, CP023, CP024, CP025, CP029]

护城河耐久性 / 竞争风险登记表
护城河主张威胁严重性当前证据缓释措施 / 尽调追问
面向多家 OEM 的中立供应商OEM 把大脑层内化AgiBot 式伙伴可能变成竞争对手,公开资料也缺少排他条款索取伙伴合同,以及续约 / 排他条款
大脑层技术优势开放研究同业压缩差异化X Square 和 NVIDIA 公开可复用模型、工具或研究路径要求基准同台测试和路线图证据
中国生态优势国内每个对手都享受同样政策顺风政策支持抬升整个行业,不是只抬升 Xingyuanzhi识别自有生态资产或锁定渠道
靠伙伴快速商业化更简单替代品靠 ROI 和落地速度胜出EP 针对相关仓储任务发布清晰定价和快速落地说法按用例梳理工作流层面的输赢数据
资本背书作为战略护城河融资军备竞赛同样补贴竞争对手2026 年上半年 22 家独角兽,加上反向商业化评论,削弱“融资即护城河”逻辑跟踪经常性收入和重复部署证据,而不是轮次规模
供应商层品牌杠杆成品机器人 OEM 掌握终端客户心智Xingyuanzhi 是隐藏层,硬件品牌仍然可见索取安装基数、附加率和替换成本数据

本登记表把公司的战略定位视为潜在有力但脆弱;在合同、基准测试和重复部署支撑前,仍不能坐实。

[CP016, CP022, CP023, CP024, CP025, CP029]

3.5 要点

Chapter 04

04财务情况

4.1 收入模式与商业化逻辑

Xingyuanzhi 的公开财务故事,首先来自一个结构上有用的区分:它并不试图靠完整机器人本体变现。相反,保留下来的公司和媒体来源都把它描述成一家 B2B 机器人“大脑”业务,围绕控制器硬件、具身模型软件,以及接入客户机器人本体的集成来搭建。财务上,公司处在企业硬件基础设施和机器人中间件之间,而不是纯软件或重资本 OEM 制造。这个区分重要,因为它改变了“好”的财务表现应当长什么样。投资人应该预期看到硬件带动的部署、按场景定制的集成,以及战略客户关系,而不是自助式订阅或大众消费设备收入。公开证据仍太薄,无法精确拆分每份合同中硬件、软件和服务各占多少;但现有记录强烈暗示,变现是混合型的。如果公司能在控制器销售之上捕获持续软件附加,这种组合可能成为优势;但今天的公开披露还不足以证明这一点。[CI001, CI002, CI003, CI004, CI005, CI006]

收入来源表
收入流机制计费单位当前价值 / 状态质量尽调追问
控制器硬件销售销售 T5 / N5 级具身大脑计算平台按台商业上真实存在;公开定价未披露提供出货量和按 SKU 实现的 ASP
具身模型软件附加销售软件随控制器内嵌,或在部署中授权按部署 / 按台 / 未知大概率存在,但公开资料未单列披露软件附加率和续约条款
集成 / 定制按场景部署、调优和工作流适配项目费早期部署中大概率有意义展示服务在合同价值中的占比和毛利率
维护 / 支持持续技术支持和迭代服务合同 / 未知未公开披露提供标准支持包和 SLA 定价
战略联合开发 / 合作经济收益与 OEM 或运营商联合开发,或承诺规模化里程碑 / 合同公开资料有提及,但合同透明度不足核对预计订单价值与确认收入

公开资料能证明公司存在多层变现方式,但无法支持精确收入结构或收入确认政策。

[CI001, CI002, CI003, CI004, CI005, CI017]
定价 / 变现表
价格 / 单位 / 合同标价与实际成交价折扣 / 未知项来源含义
Xingyuanzhi T5 / N5 价格未披露未公开标价ASP、软件绑定率和服务费未知官网 + 媒体无法直接对标价值获取能力
三年战略订单预估价值 > RMB500M前瞻口径,非实际成交价交易对手口径冲突,且无合同条款Lavx + News Globe Now只能作为方向性需求信号
EP XP15 标准版:995€/mo 或 €25,000 + 设置费公开标价折扣未说明EP 官方产品页相邻替代品给出透明 ROI 基准
EP XP15 高级版:1500€/mo+ 或 €25,000 + 定制含定制的公开标价定制费浮动EP 官方产品页说明部分买家可能偏好模块化、风险更低的支出
基于 NVIDIA 的平台经济性组件和合作伙伴定价系统成本取决于集成商选择NVIDIA 资料客户可内部自建,而非购买 Xingyuanzhi 整套方案

留存证据中,只有相邻替代品披露产品级公开定价;Xingyuanzhi 自身没有披露。

[CI006, CI020, CI021, CI027, CI028]
FI001: 收入模式桥接

产品交付如何可能从 OEM 需求转化为混合收入流。

桥接关系依据保留业务模式证据推断,因为公开资料没有收入确认政策或分部披露。

[CI001, CI002, CI003, CI004, CI017]

4.2 公开牵引力与销售效率代理指标

公司不是零收入,这一点很重要。Lavx 和 News Globe Now 均报道,Xingyuanzhi 在 2025 年出货数百台 T5,并产生超过 RMB10 million 的收入。这足以证明产品已经商业化,而不只是实验室样品。相对于已融资金额,它仍然很小。News Globe Now 又给出接近 10,000 台的 2026 年出货目标,但这个数字应被视为雄心,而不是可承销预测。公开订单说法也需要同样谨慎。战略合作公告和预期三年订单额,不等于确认收入、复购或现金回款。销售效率可见度更弱。保留下来的公开来源没有披露 CAC、转化率、平均销售周期或回本周期。最好的可用代理指标来自 EP 等相邻仓库自动化供应商,它们公布实施周期和 ROI 主张。这个对照凸显出,Xingyuanzhi 把技术承诺转化为透明商业机器的进程仍很早。[CI007, CI008, CI009, CI010, CI018, CI019]

单位经济性表
指标数值 / 状态置信度重要性尽调要求
2025 年出货量数百 / 几百台证明产品不只是停留在试点叙事按客户和季度核对精确出货量
2025 年收入> RMB10M,公开媒体报道证明已有商业收入,但规模仍小提供经审计或管理层认证的收入
毛利率未公开决定规模扩大是缓解还是放大融资风险按硬件、软件和服务拆分毛利率
单次部署实施成本未公开是贡献利润率和回本周期的关键变量提供平均部署人力和支持成本
获客成本 / 回本周期未公开评估 GTM 效率所必需按渠道 / 合作伙伴类型提供 CAC
盈亏平衡出货量未公开情景投资测算需要该指标提供固定成本基数和单台贡献毛利

本章把公开可见事实与没有私有数据就无法计算的内容分开。

[CI007, CI008, CI009, CI019, CI032, CI037]
FI002: 单位经济模型桥接

公开出货和收入证据仍无法回答贡献经济性的问题。

公开资料只证明存在硬件收入;其余关键节点都披露不足。

[CI012, CI014, CI019, CI022, CI032, CI037]
FI003: 财务估算区间

区间图保留已观察历史证据与前瞻商业目标之间的差异。

历史出货和收入值是媒体报道的大致下限。2026 年出货量和 3 年订单数字来自管理层或媒体目标, 不是经审计的已实现值。

[CI007, CI008, CI009, CI010, CI025, CI027]

4.3 成本结构、营运资本与资金需求

Xingyuanzhi 的成本结构可能比完整人形机器人 OEM 更轻,但绝不是轻软件模式。Lavx 报道,一个 50 人团队中约 90% 在研发,说明组织主要为技术迭代而搭建,而非为规模化一线销售而搭建。公开产品页面显示控制器平台从 T5 演进到突出 N5 和 Jetson Thor 的界面,这意味着持续硬件刷新和平台移植开支。业务跑在 NVIDIA 级边缘计算和软件栈上,因此很可能承担供应商依赖、库存规划、集成成本和支持义务;这些不会出现在纯基础模型公司身上。服务交付成本也比一般 AI 叙事暗示的更重要。真实具身部署往往需要现场集成、调优、验证和操作员支持。所有这些意味着,毛利率最终可能好于完整机器人 OEM,但仍明显低于干净的软件模式。当前披露记录无法说明 Xingyuanzhi 到底落在这个光谱的哪个位置。[CI011, CI012, CI013, CI014, CI015, CI016]

资本充足性表
账上现金月度烧钱现金跑道(月)资金计划用途下一轮触发因素 / 义务证据状态
未披露未披露无法计算最新一轮资金用于研发、规模化生产和招聘下一轮触发点大概率是更大规模商业转化和出货公开资料不完整
累计融资约 RMB1B烧钱速度未披露现金跑道无法计算打造下一代具身大脑和世界模型需要证明规模转化为高质量收入,而不只是试点融资可见,现金不可见
债务 / 项目融资未知债务对烧钱速度的影响未知Unknown留存资料未披露债务融资安排不能排除潜在隐性义务需要一手尽调
供应商营运资金未知库存需求未知Unknown硬件迭代和 NVIDIA 依赖意味着需要营运资金需要 BOM、库存周转和采购承诺仅为估计
客户集中度未知回款情况未知Unknown合作伙伴驱动的增长可能抬高集中度风险需要头部客户收入占比和 DSO仅为估计

这张表刻意保留可见融资与不可见现金充足性之间的缺口。

[CI011, CI016, CI022, CI023, CI024, CI025]
FI004: 资本强度 / 现金流图

控制器加模型的业务可以避开整机资本开支,但仍可能依赖融资。

图中区分看得见的融资和看不见的经营现金动态。

[CI011, CI015, CI016, CI023, CI025, CI029]

4.4 财务判断与投资判断缺口

因此,财务判断是混合的。Xingyuanzhi 拥有真实商业信号:已出货产品、披露收入、具名交易对手,以及足以支持激进研发和扩张的融资历史。但这些信号背后,每一个投资判断关键指标都仍被隐藏。公开来源没有披露账面现金、烧钱速度、现金跑道、毛利率、支持负担、应收账款、库存,或按收入口径的客户集中度。即便来源集合中最大的前瞻性商业数字——据报道 RMB500 million 以上的三年订单机会——也存在来源之间的交易对手冲突,而且没有合同细节。这并不说明公司弱;它说明证据不完整。基于当前公开信息,Xingyuanzhi 方向上有吸引力,但财务披露仍不足。严肃投资人可以继续跟踪或推进尽调,却不能对收入质量、毛利路径或资本充足性给出高确信度投资结论。[CI023, CI024, CI025, CI026, CI027, CI028]

公开财务缺口表
缺失的私有指标影响具体尽调路径
现金、烧钱速度和现金跑道无法有把握评估融资依赖索取当前资产负债表和 12 个月现金预测
实际成交价和毛利率无法验证单位经济性或定价权按 SKU 索取发票级 ASP 和 COGS
预估 RMB500M 机会的合同质量无法判断 backlog 具约束力还是偏宣传审阅已签协议和取消条款
客户集中度和回款无法验证收入持久性或营运资金压力索取收入集中度和应收账款账龄
实施与支持成本无法判断部署放量后是否能盈利按客户索取部署人力、支持工单和质保成本

这些是建立价格敏感型投资判断前所需的最低财务尽调要求。

[CI023, CI027, CI028, CI030, CI031, CI037]

4.5 要点

Chapter 05

05产品与技术

5.1 产品定义与模块地图

Xingyuanzhi 的产品故事在定位层面异常清楚,在模块文档层面则异常稀薄。官方中文和英文触点持续把公司描述为面向物理世界打造通用具身大脑,重点是多模态空间智能和跨本体泛化。这个框架意味着,客户购买的是决策和控制能力,而不是机器人本体。公开模块可见度至少显示出两代与硬件绑定的平台。公司概览指向早期 T5 具身大脑计算平台;当前产品页面则突出 N5,并把它描述为用于端侧部署的紧凑型 Jetson Thor 平台。媒体描述进一步强化了这一点:这些产品是与具身 AI 模型搭配的控制器级系统,而不是成品机器人。仍缺少的是成熟平台通常会公开的材料:更丰富的 SKU 拆分、性能表,或面向开发者的模块文档。因此,战略产品定义很强,但产品线证据仍不完整。[CE001, CE002, CE005, CE006, CE007, CE008]

产品模块 / 资产矩阵
模块 / 资产 / 产品线用户状态 / 成熟度差异化尽调缺口
T5 具身大脑计算平台机器人 OEM / 集成商已有商业引用,公开代际较旧为具身大脑逻辑提供已披露参照点未公开性能表或标价
N5 紧凑型计算平台机器人 OEM / 集成商当前对外主推产品Jetson Thor 端侧部署叙事未公开基准测试、ASP 或兼容性清单
具身 AI / 世界模型软件OEM 工程和运营团队持续迭代跨具身形态的“大脑”定位未公开模型卡或详细评测包
RoboBrain Pro 工作流层工业物流 / 装载场景媒体已有商业化引用将大脑栈接到具体工业工作流未公开产品页或工作流 KPI 表
嵌入合作伙伴的集成层人形和工业机器人合作伙伴必要但资料不足跨具身形态的供应商中立架构合作伙伴合同和兼容深度未知

产品线真实存在,但公开文档比战略定位浅得多。

[CE001, CE005, CE006, CE007, CE008, CE010]
FE001: 产品架构图

技术栈展示 Xingyuanzhi 的公开产品逻辑如何把硬件、模型和合作伙伴本体叠成机器人“大脑”系统。

[CE001, CE004, CE005, CE006, CE007, CE008]

5.2 架构与客户工作流

从保留下来的来源推断,Xingyuanzhi 的架构符合真实机器人需求。Lavx 描述的是一个域控制器,搭配在边缘硬件上实时运行的通用具身 AI 模型,并明确避开云端往返。这符合外界对抓取、导航和物理交互工作流的预期:这些场景不能容忍延迟和连接失败。公开用例覆盖货架拣选、巡检、引导、餐饮服务、养老,以及智能装卸;Lavx 还特别把 RoboBrain Pro 与 EP Equipment 装卸工作联系起来。伙伴触点和官方主张显示,这套架构意在跨人形和工业本体泛化,而不是只服务一台定制机器人。与此同时,公开记录强烈暗示部署高度依赖人工介入。大脑层供应商仍依赖机器人本体、传感器、工作流和现场验证。NVIDIA 与 Isaac 的品类级文档显示,在“机器人‘大脑’”这个营销词下面,埋着大量管线工作:传感器、ROS 2 打包、仿真、建图、姿态估计和控制环验证。[CE003, CE004, CE009, CE010, CE011, CE012]

工作流 / 用例表
用户任务现有工作流公司方案可衡量收益限制
货架拣选人工或半自动取货具身大脑加机器人本体集成在人类环境中,规划和灵巧性可能更好未公开任务速率指标
巡检和引导人工巡查或更简单的服务机器人具身 AI 负责导航和场景理解可能减少重复性人工监控未公开运行时间或安全数据
养老照护 / 服务支持人工人员和任务专用服务设备通用具身辅助层跑通后任务灵活性更广泛化风险高,公开证据薄
餐饮服务任务人工加设备具身规划和操作工作流可能替代或增强劳动力未公开生产率基准
装卸人工或更简单的仓储自动化RoboBrain Pro 搭配合作伙伴机器人硬件任务处理可能比固定自动化更丰富相邻替代品披露的 ROI 比 Xingyuanzhi 更清晰

官方用例覆盖面宽;公开工作流指标窄。

[CE003, CE004, CE010, CE022, CE032]
技术 / 运营架构表
层 / 流程 / 组件角色依赖风险
端侧计算平台在本地跑具身栈NVIDIA Jetson Orin / Thor 级硬件供应商依赖和硬件迭代风险
具身模型 / 世界模型规划和决策引擎训练数据、模型迭代、算力泛化和评估风险
感知、SLAM 和姿态估计定位和环境理解传感器加可复用机器人软件外部参考栈压低护城河
仿真 / 合成数据训练、验证、硬件前测试Isaac Sim 式工作流和场景资产如果公开验证弱,仿真到现实存在落差
合作伙伴机器人本体物理执行和本体约束OEM 硬件和控制接口产品质量部分取决于第三方本体
客户现场集成工作流调试、验证和支持运营方流程和部署团队放量可能变成服务重交付

这张架构表把看似自研的部分,与明显依赖外部机器人基础设施的部分拆开。

[CE009, CE012, CE013, CE014, CE015, CE017]
FE002: 客户工作流 / 运营流程

从客户工作流需求到具身部署,可从公开资料推断的运营流程。

合作伙伴集成手册未公开,流程只能从官方用例和媒体描述推断。

[CE003, CE004, CE010, CE022, CE023, CE032]
FE003: 关键依赖图

可能强化或削弱产品逻辑的关键依赖。

[CE014, CE015, CE016, CE017, CE024, CE033]

5.3 依赖、路线图与成熟度

Xingyuanzhi 的技术机会真实存在,但它处在一个快速成熟的外部栈之中。NVIDIA 的硬件和软件文档,加上 Isaac ROS 周边可见的 GitHub 生态,显示机器人开发者已经可以拿到定位、建图、姿态估计、仿真和人形模型开发的通用包。这并不让 Xingyuanzhi 变得无关紧要;它划定了公司必须贡献自研价值的位置。优势必须来自跨本体集成、数据、工作流经验或部署速度,而不是重新发明每个机器人 基础组件。公开路线图支持“积极开发阶段”的解读。官方页面显示 T5 到 N5 的可见切换;36Kr 则称新融资将用于下一代具身大脑和世界模型研发。不过,公开成熟度证明仍落后于愿景。商业部署已经存在,但 Xingyuanzhi 尚未发布足以让外部测试其平台是否如宣传般泛化的基准、正常运行时间或 SDK 深度。[CE014, CE015, CE016, CE018, CE019, CE024]

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
2025T5 公开里程碑已建立参照显示首个披露的具身大脑计算代际官方 About 页面
2025-2026商业出货和合作伙伴用例早期商业化产品并非纯概念Lavx / Pandaily
2026 年当前官网N5 对外主推产品活跃产品入口表明硬件和包装已刷新官方产品页
2026 融资用途下一代具身大脑和世界模型开发中路线图仍在实质推进36Kr 快讯
2026 生态定位跨具身形态通用大脑持续主线架构仍是横向供应商模式英文官网首页

阶段反映外部可见内容,而非完整内部路线图。

[CE006, CE019, CE020, CE025]
FE004: 产品成熟度 / 能力图

定性成熟度视角,把可见战略优势和缺失的验证材料分开。

矩阵比较可见证据类别,不评判隐藏的内部质量。有限项反映公开披露不足,不代表假定技术薄弱。

[CE021, CE026, CE027, CE030, CE037, CE038]

5.4 差异化、信任与未解技术缺口

当前产品判断最适合概括为:架构可信、商业化尚早、文档不足。Xingyuanzhi 最强的差异化主张,是愿意做中立的跨本体大脑层,而不是又一家机器人本体公司。这是一个有意义的设计选择,也可能具备可扩展性。但保留下来的公开记录还没有把这个选择转化为可核验级证据。没有清晰的公开信任或安全页面,没有可见安全论证,没有产品认证组合,也没有公司特定基准包来展示系统跨任务或本体的表现。公开专利检索触点也没有给出清晰的自研 IP 图景。与此同时,Isaac ROS、Isaac Sim 和 GR00T 等开放与既有平台持续降低拼出一套合格机器人参考栈的成本。这个组合意味着产品逻辑可信,但公开尽调负担仍很高。投资人和客户还需要更深的技术文档,才能有把握地判断其稳健性、安全性或可防守护城河。[CE026, CE027, CE029, CE030, CE031, CE033]

信任 / 质量 / 合规表
控制 / 认证 / 质量指标状态范围缺口
公开安全论证未见公开产品级安全就绪度无公开材料
公开网络安全架构未见公开安全部署和软件更新能力无公开材料
公开认证未见公开产品或公司认证无公开材料
公开基准测试包未见公开延迟、成功率、正常运行时间、故障模式无公开材料
公开事故或召回历史留存资料中未见运营风险透明度无公开材料

最强的公开证据集中在定位和用例;正式信任材料基本缺位。

[CE021, CE026, CE027, CE030, CE036]

5.5 要点

Chapter 06

06客户情况

6.1 客户基础与分层

Xingyuanzhi 的具名客户数量并不多,但战略意义不低。保留下来的公开记录指向三大细分。第一类是 AgiBot 等机器人 OEM,它们可能把 Xingyuanzhi 的大脑层嵌入或评估进自己的本体。第二类是与仓储或装卸工作流绑定的工业自动化和搬运伙伴,最明显的是 Zhongli/EP 线索。第三类是 Beijing Yizhuang Robot 等公共部门或生态运营方,其商业逻辑看起来与示范部署、试点密度和机器人集群活动绑定。这是一张中国优先的客户地图,扎根北京和国内机器人枢纽,而不是国际多元化。公司的公开触达可能宽于具名客户名单:36Kr 报道其覆盖超过 70% 的头部具身智能公司。但这种宽口径生态主张,不能替代已披露账户名单、细分组合、收入集中度表或持久续约历史。[CU001, CU002, CU003, CU004, CU030, CU035]

客户分层表
客群买方 / 用户 / 付款方用例规模收入 / 战略价值缺口
人形机器人 OEMOEM 产品团队 / 机器人运营方 / 平台预算人形平台内嵌具身大脑战略空间大,logo 数量少若广泛嵌入,可能形成深度持续的绑定收入未公开按 OEM 拆分的收入占比
工业自动化 / 搬运合作伙伴运营或自动化负责人 / 现场运营方 / 资本开支预算装载、卸载、重复物流、物料搬运运营场景具体工作流关联强,可能跨站点扩张未单独拆出 Xingyuanzhi 特定价值获取
公共部门机器人生态园区运营方 / 访客或服务团队 / 项目预算试点项目、展示、采购探索、服务工作流密集但可能波动大可加速获客并提升生态影响力量产与展示场景的经济性不清晰
研究与生态参与方开发者 / 实验室 / 创新团队 / 研发预算测试、演示、数据生成、场景验证可能很广,但披露不足支撑品类触达和数据收集具名账户基本缺失
更广泛的具身 AI OEM 生态多个国内玩家 / 工程团队 / 预算来源混合跨机器人形态的大脑集成36Kr 称覆盖广泛若属实,是重要战略切口未披露活跃账户分母

客群结构比实际客户数更清楚。

[CU001, CU002, CU003, CU004, CU030]
FU001: 客户旅程图

从生态认知到规模化部署的典型旅程,看起来更靠关系推动,而不是自助式转化。

[CU017, CU018, CU024, CU026, CU034]

6.2 具名客户证明与采用轨迹

客户故事中最有说服力的一点,是 Xingyuanzhi 看起来接入了真实运营环境,而不只是概念演示。Lavx 和 News Globe Now 都报道,公司在 2025 年出货数百台并实现超过 RMB10 million 收入,这足以证明真实采用。AgiBot 是价值最高的具名证明,因为它是规模化人形机器人 OEM,并在 2026 年有可见部署。AGIBOT 官方材料描述,超过 60 台机器人在 WAIC 场馆运行,并披露具备现场产线指标的工业部署,以及快速扩张的产品组合。如果 Xingyuanzhi 继续嵌在这个生态里,这段关系可能非常重要。EP 证据讲的是另一条同样重要的故事:智能装车、卸车和仓库移动任务,是具身智能可以变现的具体购买需求。但这些证明仍是间接的。公开记录更清楚地显示客户工作流,却没有同样清楚地显示 Xingyuanzhi 在其中捕获了多少价值。[CU005, CU006, CU007, CU008, CU009, CU010]

客户增长 / 采用轨迹表
指标数值日期来源置信度含义缺失分母
出货量数百 / 几百台2025Lavx + News Globe Now商业采用已经发生各客户具体台数未知
收入> RMB10M2025Lavx + News Globe Now商业化并非空白各客户收入占比未知
预计战略机会> RMB500M,3 年内2026 年披露窗口Lavx / News Globe Now中低锚定账户扩张可能带来实质增量交易对手和合同质量不清楚
AGIBOT 部署规模代理指标截至 2026 年 3 月宣布第 10,000 台机器人2026PR Newswire合作伙伴规模可能放大附加价值Xingyuanzhi 附加率未知
AGIBOT 公共场馆运营WAIC 现场 60+ 台机器人2026AGIBOT 官方文章显示合作伙伴已在真实环境运营未披露 Xingyuanzhi 模块占比

本表把真实历史证明与预测值、合作伙伴规模代理指标分开。

[CU005, CU006, CU009, CU011, CU028]
具名客户证明表
客户客群部署 / 用例生产 / 试点结果局限
AGIBOT人形机器人 OEM人形机器人集成场景;AGIBOT 生态内的场馆和工业部署AGIBOT 生态已有生产级部署;Xingyuanzhi 附加状态部分来自间接证据WAIC 现场 60+ 台机器人;工业案例披露了具体吞吐量和正常运行时间指标公开资料没有单独拆出 Xingyuanzhi 的性能或收入
Beijing Yizhuang Robot / E-Town 生态公共部门机器人运营方试点、展示、采购摸底、机器人集群活动可能偏战略 / 试点导向密集活动和政策环境支撑需求发现和早期订单收入、合同阶段、重复使用不清楚
Zhongli / EP 关联工作流工业搬运 / 自动化装载、卸载、出库流转、重复运输工作流证明真实存在;Xingyuanzhi 角色部分来自间接证据相邻案例研究声称一天安装,并提升产能 / 效率案例指标属于自动化工作流,尚未清楚归因到 Xingyuanzhi 技术栈

公开资料清楚展示底层工作流时,具名客户证据最强,即便 Xingyuanzhi 的具体经济性仍不透明。

[CU008, CU009, CU010, CU013, CU014, CU015]
FU002: 采用 / 部署漏斗

客户证据从广泛生态触达收窄到具名部署,再到留存可见度时急剧变薄。

数值是证据密度的序位评分,不是客户数。分数显示证据质量在漏斗各阶段如何塌缩。

[CU003, CU007, CU020, CU021, CU034, CU036]
FU003: 客户证据矩阵

证据质量矩阵,对比三个最可见的客户或合作伙伴证据点。

[CU008, CU009, CU014, CU016, CU028, CU029]

6.3 耐久性、扩张与留存缺口

Xingyuanzhi 客户记录的主要弱点不是相关性,而是耐久性。保留下来的公开来源没有披露活跃账户数、站点数、续约行为、GRR、NRR,甚至基础队列指标。最好的可用耐久性代理,是持续的伙伴节奏——AgiBot 快速公开产品排期、多年战略订单语言,以及反复出现的生态存在感。这有用,但仍远弱于正式续约数据。先落地、再扩张的逻辑是合理的:一个 OEM 可以增加更多本体,一个工业伙伴可以增加更多站点,一个场馆运营方可以从试点工作流扩展到更广采购。但这条路径仍是假设。有些关系也可能主要为数据收集和品类背书创造战略价值,而不是立即贡献收入,这意味着仅看 客户名容易高估客户质量。按当前记录看,公司仍早到足以把客户证明解读为留存前阶段。[CU016, CU017, CU018, CU019, CU020, CU021]

留存 / 重复使用 / 满意度表
指标数值 / 空值客群置信度尽调要求
活跃客户数全部按客群提供当前活跃账户数和付费账户数
站点数 / 部署数全部按主要客户提供已安装站点和机器人数量
GRR / NRR全部按队列提供续约和扩张指标
续约率战略合作伙伴提供已签续约和合同延期
客户满意度 / NPS全部提供客户推荐引述或满意度调查结果

留存数据缺失本身就是本章主要发现。

[CU020, CU021, CU022, CU027, CU029, CU034]
公开客户证据缺口表
缺口重要性具体尽调路径
按具名客户标注生产 / 试点避免高估商业成熟度索取包含阶段和付款状态的账户清单
留存和续约历史需要用它把客户标识转化为可持续收入质量索取队列续约和合同延期
头部账户收入集中度需要用它衡量单一客户冲击风险索取前 10 大客户集中度明细
成果归因到 Xingyuanzhi 层需要区分合作伙伴成功和 Xingyuanzhi 成功按部署索取工作流级前后对比指标
独立客户证言需要验证满意度和实施质量开展客户访谈或获取具名案例研究

这些是在把当前客户记录视作稳定基础之前的最低要求。

[CU007, CU015, CU020, CU021, CU027, CU029]

6.4 集中度风险、采购摩擦与客户结论

这么年轻的公司出现客户基础集中并不意外,但这仍是实质风险。AgiBot 是有价值的具名客户,也可能贡献深度规模,但它同时在打造自己的具身栈。Yizhuang 可以加速发现和试点,但公开证据没有显示它已经是一个持久高收入客户。EP 搬运工作流显示真实运营需求,但也显示许多买方任务可以先用更简单的自动化解决,买方不一定必须立刻承诺采用宽口径具身大脑平台。这让 Xingyuanzhi 形成了一个有前景但风险不低的客户形态:战略重要具名客户、具体工作流相关性,以及很少的公开留存证据。更宽的赛道背景也提醒谨慎。中国具身 AI 资本涌入,意味着具名客户和试点积累速度可能快于多元、可重复的客户经济性。正确结论既不是怀疑一切,也不是过度自信。公司有真实客户证明,但仍集中、间接,还没有形成持久队列故事或可重复收入底座。[CU012, CU023, CU026, CU031, CU032, CU033]

扩张与集中度风险表
扩张驱动因素集中度风险影响尽调路径
单一 OEM 内更多机器人形态单一战略 OEM 变得过于重要按 OEM 系列索取附加率和收入占比
单一工业合作伙伴内更多站点工作流扩张取决于一个合作伙伴的路线图索取站点级铺开时间表和流失风险
Yizhuang / Beijing 生态密度北京地域集中会扭曲需求质量中高按城市和省份拆分管线
公共部门展示活动试点曝光跑在付费生产转化前面为每个具名账户标注试点或付费生产
数据收集和战略信号价值客户标识对叙事或训练的作用大于收入中高索取收入型 / 非收入型合作分类

本表关注“先落地再扩张”在哪些地方能帮忙、在哪些地方会误导。

[CU012, CU019, CU023, CU025, CU026, CU032]

6.5 要点

Chapter 07

07风险

7.1 风险排序与法律 / 监管敞口

Xingyuanzhi 的风险画像,核心不是具身 AI 是否有市场关注度,而是一家非常年轻的大脑层供应商,能否在自己并不完全控制的栈里完成商业化。今天最高的剩余风险集中在先进算力依赖、法律责任分配不清,以及缺少公开治理材料;这些材料本应让外部投资人核验安全、隐私和网络安全如何处理。中国政策条件支持,但政策支持不等于运营已经去风险。事实上,快速政策支持可能会在文档、测试和跨境合规成熟之前,加大商业化压力。这形成了典型前沿技术格局:可触达机会真实存在,但风险承担机制仍在搭建。尽调读者应把法律和监管问题排在清单前列,而不是放到最后,因为它们决定增长能否被融资支持。[CR001, CR002, CR003, CR004, CR005, CR007]

监管 / 法律风险登记表
规则 / 问题 / 案例管辖区状态可能性严重性缓释措施剩余敞口尽调路径
面向中国关联实体的先进计算出口管制美国 / 域外2026 年仍有效且持续变化中高极高替代 BOM 规划;供应商尽调;分类审查获取 ECCN 备忘录、供应商图谱和替代计划
大脑供应商、OEM、集成商和运营方之间的产品责任分配多管辖区公开记录中法律上尚未厘清合同赔偿安排和保险审阅客户 / OEM 合同和责任上限
未来海外扩张中的 AI 透明度、人工监督和网络安全义务欧盟及其他受监管市场规则已生效 / 分阶段落地文档、日志、人类在环控制中高将产品架构映射到 AI Act 义务
多模态机器人工作流中的隐私和生物识别数据处理中国 / 欧盟 / 全球影响重大,但公开文档不足数据最小化、同意、存储和治理控制索取数据治理政策和类 DPIA 材料
部署或维护期间的工作场所安全和现场故障报告客户现场所在地管辖区属于运营义务,未发现公开事故日志调试作业手册和事故响应流程中高索取事故登记册和纠正行动档案

各行按公开证据下的剩余严重性排序,而非公司内部风险评分。

[CR003, CR004, CR005, CR007, CR008, CR009]
FR001: 风险热力图

公司当前主要风险的可能性—影响矩阵。

各位置为作者基于截至 2026-08-30 的公开证据作出的判断,并非概率预测。

[CR001, CR013, CR021, CR025, CR027, CR033]

7.2 运营安全、商业化与数据治理风险

运营上,Xingyuanzhi 处在 AI 最难的角落之一:系统必须在物理世界中感知、决策并安全行动。公开记录支持一个可信产品愿景,但没有消除演示与稳健站点级表现之间的现实落差。OSHA 的通用机器人指南和 Deloitte 的物理 AI 分析都强化了一点:故障常常出现在调试、维护、调优或意外环境变化中。IFR 又给出一个重要的市场特定警示:即便宣传正在扩张,人形机器人在真实生产环境中的能力仍有限。这很关键,因为出售具身大脑的公司,价值取决于整条部署闭环的一致性。数据治理再加一层风险。如果中国具身 AI 头部企业正激进扩大数据采集运营,任何在多模态机器人中工作的供应商,都应当解释数据来源、授权、日志和下游控制;即使当前尚未出现公开丑闻,也不能跳过这些问题。[CR014, CR015, CR016, CR017, CR018, CR027]

运营 / 质量 / 安全风险登记表
故障模式可能性严重性缓释成熟度剩余敞口未解缺口
新环境中的仿真到现实性能差距部分未见跨机器人形态或站点的公开基准包
调试或维护阶段安全事故Unknown未披露公开事故日志
边缘 AI 或联网机器人工作流中的网络安全弱点Unknown未见公开安全架构或信任中心
视频 / 传感器数据治理失效Unknown未公开详细数据血缘或留存披露
硬件刷新或集成错配造成部署停机中高部分中高未见公开兼容性矩阵或现场故障统计

机器人故障具有物理性、站点特异性,仅靠公开营销材料难以诊断,因此运营风险偏高。

[CR012, CR013, CR014, CR015, CR016, CR017]
FR002: 风险传导图

有向图展示技术和法律失误如何传导为融资和估值损伤。

[CR006, CR008, CR018, CR021, CR026, CR033]

7.3 合作伙伴、地域与团队 / 执行风险

Xingyuanzhi 的伙伴网络既是扩张路径,也是脆弱性来源。当它接入 AgiBot 和北京机器人生态等严肃机器人交易对手时,公司最强,因为这些关系会加快进入用例、活动和买方注意力。同样的关系也会带来内部化、集中度和地域集群风险。大型 OEM 可以先向伙伴学习,之后把更多栈拉回内部。北京中心生态可以迅速制造可见度,同时也让管线质量过度依赖会议活动、本地政策和单一区域集群。执行风险也会上升,因为相比媒体报道暗示的覆盖广度,公司仍极其年轻。公开来源对管理层深度、治理或部署团队规模披露还不够,无法判断组织是否已经准备好在没有关键人压力的情况下扩展到多个账户。[CR021, CR022, CR023, CR024, CR031, CR038]

合作伙伴 / 依赖风险登记表
依赖交易对手角色集中度失效情景严重性缓释措施剩余敞口
先进边缘计算NVIDIA 及上游芯片生态运行大脑层工作负载供应受限、许可摩擦或路线图错配极高合格替代方案和库存规划
人形 / 具身 OEM 合作伙伴AgiBot 及类似 OEM机体平台和部署路径OEM 内化大脑栈,或重新定价集成多 OEM 策略和工作流深度
北京生态和活动Yizhuang / WRC 集群获客线索和政策可见度中高区域集中削弱管线质量中高向北京以外和更多垂直领域扩张
系统集成和现场交付合作伙伴集成商 / 客户运营团队调试和运营交接现场故障或部署周期慢会伤害信任标准化作业手册和支持层中高
公开叙事和投资者信号行业媒体 / 政策动能支撑招聘和融资经济性成熟前叙事先降温更快发布运营证明中高

公司卖的是赋能层,不是完整机器人,因此依赖既是技术问题,也是商业问题。

[CR001, CR003, CR006, CR021, CR022, CR023]
人员 / 执行风险登记表
角色 / 职能依赖或缺口可能性严重性缓释措施尽调路径
创始人 / 核心架构领导年轻平台公司可能仍高度依赖创始人扩充技术领导梯队索取组织架构图和授权分工
部署工程与现场支持账户扩张可能变成重服务交付固化集成与支持手册要求披露部署团队人数与利用率
安全 / 合规责任归属公开资料未显示信任、安全或隐私的明确负责人指定负责人并发布治理材料要求提供政策负责人名单和治理会议机制
商业客户管理少数战略客户可能吃掉大量管理注意力中高建立体系化客户成功职能要求披露头部客户复盘节奏
董事会与投资人监督公开资料未清楚披露董事会构成建立正式监督与接班安排要求提供董事名单、委员会和否决权安排

这些是基于公开证据的风险判断;内部团队结构可能明显强于外部可见信息。

[CR031, CR038, CR039, CR040]
FR003: 依赖关系图

有向图展示公司最重要的生态依赖。

[CR003, CR006, CR021, CR023, CR024, CR032]

7.4 财务 / 模型风险、缓释措施与终止条件

从财务看,核心问题不是 Xingyuanzhi 表面上缺钱;而是公开证据还看不出,现有资本能否转化成持久、可防守的经济性。不到一年融到约 ¥1 billion 很亮眼,但它放大预期的速度,可能快过证明可重复收入的速度。在一个拥挤、补贴充足的赛道里,专做大脑层的公司既要守住技术价值,也要守住定价权,而更简单的自动化替代方案还在持续进步。因此,收入集中度、烧钱速度、毛利率、续约等数据缺失,投资人应视为一阶风险,而不是小缺口。好消息是,缓释清单很清楚:在可行处降低算力暴露,发布更强的信任与验证材料,证明增长不只来自少数战略客户,并设置与芯片获取、客户自研替代、部署证明和安全事故挂钩的明确叫停标准。投资人还应要求证据,证明当曝光度、工程野心和融资压力同时上升时,管理层仍能把安全优先级排对。上述证据出现前,合适的剩余风险评级仍是高。[CR025, CR026, CR032, CR033, CR034, CR035]

风险缓释与止损标准表
风险可监测触发项阈值 / 事件行动含义
先进算力获取中断供应商或政策更新失去所需芯片供应,或一个季度内没有获批替代方案暂停投资判断;重新评估交付路线图
核心 OEM 内部化合作方产品发布或合同变更AgiBot 或同类合作方在关键工作流中替换 Xingyuanzhi 层合作方带动规模化的投资逻辑被打破
部署证据滞后新一轮融资临近下一次融资事件前,仍没有可信的公开部署 KPI、留存指标或收入证据立场转向回避,或只在大幅折价下考虑
安全或合规事件事故、监管问询或法律通知重大现场伤害、重大数据事件或正式执法行动升级为红旗尽调和法律审查
关键人员 / 治理冲击高管或董事会变动且未见补位核心技术负责人离职,或隐藏治理争议暴露重新评估执行折价与条款
集中度持续不透明尽调数据请求回复公司在尽调中拒绝分享客户集中度和烧钱数据视为估值支撑不足

这些阈值是基于公开证据推导的投资人监控规则,不是公司发布的内部限制。

[CR001, CR020, CR025, CR026, CR033, CR040]

7.5 展示材料

Chapter 08

08估值

8.1 融资背景、证据质量与当前建议

Xingyuanzhi 很容易让人觉得投资逻辑有吸引力,却很难精确定价。保留下来的证据支撑了一条强战略叙事:公司立足中国,由 BAAI 孵化,聚焦机器人大脑层而非硬件本体,并在不到一年里融到约 ¥1 billion。这样的组合足以让严肃投资人重视,也足以把公司视为到 2026 年中处在独角兽区间。但这还不足以支持在某个明确价格上给出买入结论。公开证据仍没有收入、毛利率、烧钱速度、积压订单或股权结构表细节,尽调团队也就无法把叙事强度转换成常规投资测算模型。因此,正确建议必须随证据而动。按今天的材料,公司应归入继续研究:值得密切跟踪,也可能很有吸引力,但现有支撑还不足以做出肯定且不敏感于价格的投资决定。[CV001, CV002, CV003, CV004, CV005, CV006]

建议摘要表
建议置信度风险评级估值立场决策含义
继续研究偏高密切跟踪,但没有内部财务数据和投资条款证据前,不支持买入

建议对价格和证据敏感,并非通用的公司质量评分。

[CV006, CV007, CV008, CV041, CV042]
投资逻辑 / 反向逻辑表
论点哪些证据会改变判断
大脑层平台可能成为多类机器人本体的横向标准需要证明挂载率、留存,以及客户对大脑层的依赖
中国具身 AI 的政策和资本顺风支撑快速商业化需要确认政策动能会转化为可持续经济性,而不是试点膨胀
BAAI 孵化和早期强融资传递人才与技术可信度需要证明这种可信度能转化为收入效率,而不只是融资成功
具名客户和合作伙伴带来战略相关性需要客户集中度和合同经济性,证明这些具名客户的变现质量
如果后续证据支撑独角兽区间定价,估值可能仍然合理需要精确的条款清单定价、优先权和更新版 KPI 包

反向逻辑集中在经济性缺失,以及平台承诺可能已被估值计入的风险。

[CV003, CV004, CV019, CV020, CV021, CV024]
FV001: 推荐逻辑

从市场、产品、验证和估值缺口推导最终“继续研究”建议的流程图。

[CV005, CV006, CV011, CV019, CV024, CV042]

8.2 市场增长与可比估值锚点

市场背景有利,但还不足以弥补公司层面指标缺失。多家第三方预测都指向未来十年人形机器人和物理 AI 的显著增长,这支撑了一个判断:如果机器人大脑平台拿到横向角色,价值可能很高。但这些预测对市场规模和增长斜率分歧很大,更适合验证战略方向,而不是锁定狭窄的当下估值。公开可比公司也讲了类似故事。UBTECH、Serve Robotics 和 Symbotic 的估值跨度很大,因为市场奖励的是不同东西:清晰收入规模、公开证明、商业模式清晰度和部署深度。Xingyuanzhi 的平台论点有吸引力,但还没有 Symbotic 级别的证明,也没有 UBTECH 级别的公开市场可比性。Spirit AI 等私有同业进一步压住了溢价情景,因为同一个中国生态里已经有更强的公开部署证据。更宽泛的边界可比公司也提醒投资人,品类热度本身并不能生成可用的定价公式。[CV009, CV010, CV011, CV012, CV013, CV014]

可比估值表
可比公司 / 标的指标倍数 / 估值 / 状态相关性局限
Xingyuanzhi私募融资背景到 mid-2026 已处于独角兽区间;确切当前投后估值未公开披露直接标的公司没有确切条款或收入披露
Spirit AI未上市竞品融资与验证融资接近 ¥2B;留存来源显示其工业验证更强中国最接近的具身 AI 同行,部署公开程度更高留存材料中没有清晰披露的公开市场价值
UBTECH Robotics公开市值~$5.36B 市值(Aug 2026)最接近的上市人形机器人参照公开市场交易逻辑和硬件结构不同
Serve Robotics公开市值~$0.43B 市值(Aug 2026)可作为早期机器人验证的下行边界商业模式和配送场景不同
Symbotic公开市值~$24.13B 市值(Aug 2026)显示企业级验证足够深后可能达到的上限价值收入规模大得多,且聚焦仓储自动化
NVIDIA平台市值~$5.253T 市值(Aug 2026)展示不可替代平台层的潜在价值捕获业务太宽、太成熟,不能直接用作定价可比

该表有意混用不同参照,因为披露充分的纯机器人“大脑”估值很稀缺。

[CV003, CV013, CV014, CV015, CV016, CV017]
FV004: 投资 KPI

面向投资委员会的计分卡,以 1-5 分刻画当前最重要的估值维度。

分数为作者仅基于截至 2026-08-30 留存公开证据作出的判断。

[CV007, CV011, CV021, CV023, CV024, CV041]

8.3 乐观 / 基准 / 悲观区间与进入纪律

精确财务报表尚未公开,因此最可防守的估值方法是情景法。乐观情景假设 Xingyuanzhi 成为多个机器人 OEM 事实上的中立大脑标准,把战略客户名单转化成可重复部署,并发布足够的可靠性和留存证据,从而拿到平台溢价。基准情景假设公司仍然重要、资金充足,但仍以中国为先、由伙伴牵引,且经济性只做到部分透明。悲观情景则假设,在硬证据到来前,伙伴自研替代、芯片摩擦或情绪降温先发生。这些情景并不假装给出精确目标;它们是有纪律的区间,用来表达价值对尚未验证里程碑的敏感度。因此,进入纪律很重要。一家公司可以具备战略重要性,但如果投资人为尚未发布的证据付钱,仍可能是一笔糟糕买入。[CV018, CV019, CV021, CV022, CV025, CV026]

乐观 / 基准 / 悲观情景表
情景假设估值 / 回报逻辑关键风险概率信号
乐观跨 OEM 的大脑标准、部署指标更强、客户更广、无算力冲击示意性 EV 区间 $2.0B-$2.8B;平台溢价显现执行仍然很难;溢价取决于证据下一轮前需要看到 KPI 改善
基准中国优先增长、选择性扩张、合作方留存,完整经济性仍不透明示意性 EV 区间 $1.2B-$1.7B;接近独角兽区间延续,带小幅溢价如果增长证据滞后,可能显得估值已打满最符合当前公开证据
悲观合作方内部化、出口管制摩擦、情绪降温、没有新证据示意性 EV 区间 $0.7B-$1.0B;存在跌破独角兽估值重置风险降价轮或结构化轮次条款如果下一轮融资早于更强商业证据出现,将触发该情景

区间以 USD 计,只是情景锚点,不是精确目标或正式估值。

[CV025, CV026, CV027, CV028, CV029, CV030]
FV002: 估值敏感性

按情景中点列示示意性企业价值,显示证据改善或恶化时,价值变化会有多快。

数值为示意性的百万美元口径,不是目标价。最后一根柱是来自留存独角兽区间报道的叙事锚点,不是经核实的投后融资条款清单数字。

[CV003, CV027, CV029, CV031, CV032]
FV003: 估值 / 回报区间

情景区间凸显不确定性很大,原因是缺少经营指标和融资条款清单细节。

所有数字均为示意性的百万美元口径。每个区间的宽度反映了经济性、合作伙伴依赖和融资条款的不确定性。

[CV027, CV029, CV031, CV032]

8.4 什么会改变投资结论

本章估值里令人鼓舞的一点是,上调路径很直接。Xingyuanzhi 不需要一个截然不同的市场来支撑更强估值判断;它需要更好的证据。尽调只要拿到当前收入和烧钱速度、客户集中度、与旗舰伙伴的合同结构以及最新优先股堆叠,就能立刻收窄估值区间。运营证明也一样:更明确的部署指标、少数具名客户之外更广的客户面,以及直接证据证明公司的大脑层一旦嵌入就难以替换。在此之前,投资人应设定清晰的叫停触发器。如果算力获取恶化、伙伴自研替代整套技术栈,或下一轮融资早于商业证明改善到来,下行风险会迅速上升。更强披露也可能提高价格和条款谈判力,而这与公司叙事本身同样重要。换句话说,这家公司值得贴近跟踪,但不适合粗糙定价。[CV033, CV034, CV035, CV036, CV037, CV038]

投资逻辑破裂与止损触发项表
触发项阈值对投资逻辑的传导行动含义
先进算力获取冲击一个季度内没有合格替代关键算力交付路线图和客户信任同时走弱暂停投资判断并立即重估下行
合作方内部化旗舰 OEM 在核心工作流替换 Xingyuanzhi 层平台标准投资逻辑破裂除非估值重置,否则转向回避
证据滞后拖到下一轮下一次融资事件前没有有意义的部署 KPI、收入证据或客户广度更新叙事跑在证据前面;降价轮风险上升假设会出现结构化条款,或普通股价值更低
安全或合规事件重大现场事故、执法行动或数据治理争议信任折价和客户犹豫加深施加更大估值折扣
集中度持续不透明公司在尽调中拒绝分享头部客户和烧钱数据无法准确测算下行维持继续研究 / 不买入立场

这些是基于公开证据推导的投资人定义监控规则。

[CV033, CV034, CV035, CV036, CV040]
最终尽调问题清单
主题缺失证据重要性负责人或尽调路径
最新融资条款精确的投前 / 投后估值、期权池、清算优先权、反稀释、投资人权利判断标题估值是否匹配普通股价值要求提供条款清单、股权结构表和董事会同意文件
经营指标收入运行率、毛利率、烧钱速度、现金、积压订单估值和融资风险的核心输入要求提供月度管理层 KPI 包
客户集中度头部客户收入占比、合同期限、扩张管线把具名客户证据转化为收入质量判断要求提供前 10 大客户明细表
部署证据已安装基数、留存、任务级可用率、替换成本证明平台溢价所需要求提供部署评分卡和客户访谈
供应链韧性算力 BOM、ECCN、替代方案、采购计划理解增长天花板和下行风险所需审阅供应商地图和合规备忘录

如果管理层能可信提供这五项,建议可能明显改变。

[CV034, CV036, CV041, CV042]

8.5 展示材料

免责声明

本报告基于截至 2026-08-30 可获得的公开资料。审阅材料未包含该私营公司的财务报表、客户合同、董事会材料、融资文件、安全审计和部署日志;作出任何投资决定前,应通过一手尽调验证这些内容。

证据索引

结论
编号陈述可信度来源
CO001 Beijing Xingyuanzhi Robot Technology Co., Ltd. was established on 2025-08-01. SO002, SO004, SO009, SO012
CO002 Xingyuanzhi was incubated by the Beijing Academy of Artificial Intelligence (BAAI / Zhiyuan Research Institute). SO001, SO004, SO009, SO011, SO012
CO003 Public materials show Beijing operations split between a Haidian contact office and a Yizhuang-registered legal entity, confirming a Beijing base but not a single published headquarters address. SO008, SO012, SO013
CO004 The company describes its mission as achieving multimodal spatial intelligence and building a general embodied brain for the physical world. SO001, SO002, SO009
CO005 Xingyuanzhi publicly positions itself as a provider of the robot “brain” and compute layer rather than a maker of robot bodies or actuators. SO014, SO016, SO028, SO029
CO006 The official company profile names Liu Dong as founder and CEO. SO004, SO012
CO007 Liu Dong previously served as JD.com’s intelligent-driving general manager before founding Xingyuanzhi. SO011, SO012, SO028, SO029
CO008 BAAI’s conference materials identify Liu Dong as both Xingyuanzhi CEO and PI of the institute’s embodied-brain research center. SO011
CO009 Public biographies describe co-founder Dr. Mu Yadong as a Peking University researcher and Zhiyuan scholar. SO012, SO028, SO029
CO010 The official company profile says co-founder Sun Zhenguo publicly released the ω-EVA world model at the June 2026 BAAI conference. SO002, SO004
CO011 Reviewed public sources do not disclose a full board roster, independent directors, or a formal governance chart. SO004, SO012, SO028
CO012 Xingyuanzhi disclosed a RMB200 million angel round in September 2025. SO004, SO012, SO029
CO013 Xingyuanzhi disclosed an angel+ round of more than RMB100 million in December 2025. SO002, SO004, SO012
CO014 The company announced a Pre-A round on 2026-06-03. SO004, SO012, SO015, SO017
CO015 Company and media sources consistently report cumulative financing of roughly RMB1 billion within the first ten months after incorporation. SO004, SO014, SO015, SO016, SO017, SO019
CO016 Publicly named Pre-A backers include Beijing Industrial Investment, CRRC Capital, Songhe Capital, Creation Capital, Huakong Fund, Guojun Innovation Investment, Jiangxi Financial Holding, Aiteke, Hengxing Group, and Qi’an Investment. SO004, SO015, SO016, SO017
CO017 36Kr and the official company profile both state that Yuansheng Venture Capital followed the company across three rounds and that BAAI continued supporting the business. SO004, SO015, SO017
CO018 The angel round included strategic investors such as AgiBot/Zhiyuan Robot and Zhongli alongside venture investors such as CAS Star and Hillhouse Ventures. SO004, SO012
CO019 Official English and Chinese pages agree that T5 launched in 2025 and was linked to AgiBot’s Genie G2 robot. SO002, SO003, SO004, SO012
CO020 The official English about page places the first T5 milestone on 2025-08-08. SO003
CO021 Chinese official and Baike timelines place the G2/T5 commercialization window in October-November 2025 rather than August. SO002, SO012
CO022 Xingyuanzhi’s public product stack centers on RoboBrain Pro, the ω-EVA world model, and edge-compute hardware. SO004, SO012, SO029
CO023 Baike reports that the T5 controller used NVIDIA Jetson Thor and provided 2070 TFLOPS for on-device large-model acceleration and real-time decision-making. SO003, SO012, SO013
CO024 The current English product page markets newer edge-compute hardware on Nvidia Blackwell architecture with dense I/O, implying product iteration beyond the first T5 controller. SO005, SO006
CO025 Public materials describe Xingyuanzhi’s technical path as full-model edge deployment with a soft-hardware integrated system, not cloud-dependent control. SO012, SO029
CO026 The official company profile claims the embodied-brain platform already covers more than 70% of domestic robot-body enterprises. SO004
CO027 The official company profile claims Xingyuanzhi is the world’s largest shipper of the NVIDIA Jetson Thor platform. SO004
CO028 Xingyuanzhi signed a strategic cooperation agreement with Beijing Yizhuang Robot and public materials describe a three-year order target of at least RMB500 million. SO004, SO012, SO028, SO029
CO029 Public materials link the Yizhuang relationship to government-service, inspection, tour-guide, and shopping-guide scenarios. SO004, SO012, SO016
CO030 Public sources name AgiBot as a customer or strategic partner and Zhongli/EP Equipment as a commercialization partner for embodied loading-and-unloading systems. SO004, SO012, SO028, SO029, SO030
CO031 Official company profiles say the Zhongli/EP embodied loading-and-unloading solution entered global sale in late June 2026. SO002, SO004, SO012
CO032 Official profiles say Xingyuanzhi debuted at Hannover Messe in April 2026 and launched the BotPack B series there. SO002, SO004
CO033 The company profile says Xingyuanzhi entered Fortune China’s Tech 50 in June 2026. SO002, SO004, SO012
CO034 Xingyuanzhi and BAAI announced the first embodied-interaction world-model key laboratory in June 2026. SO002, SO004, SO012
CO035 Xingyuanzhi says it released the embodied-interaction world model ω-EVA at the June 2026 BAAI conference. SO002, SO004, SO011, SO012
CO036 BAAI’s conference program put Liu Dong alongside other high-profile embodied-AI CEOs and framed 2026 as a debate over bubbles, commercialization, and scalability. SO011
CO037 China Biz Insider reports that China’s embodied-AI sector added 15 new unicorns in H1 2026 and that many startups have only 18-24 months of cash runway. SO020
CO038 EmbodiedGlobal reports H1 2026 embodied-AI funding of RMB93.5 billion across 322 deals, illustrating how aggressively capital is concentrating around the sector. SO021
CO039 EmbodiedGlobal says Morgan Stanley projects roughly 50,000 humanoid units shipped in China in 2026. SO021
CO040 Goldman Sachs, CNBC, and IDTechEx all publish multi-billion-dollar long-term market scenarios for humanoid robots, helping explain why investors pay premium multiples for enabling infrastructure. SO023, SO024, SO025
CO041 Two low-reputation third-party summaries say Xingyuanzhi shipped hundreds of T5 units in 2025 and generated more than RMB10 million of revenue. SO028, SO029
CO042 One low-reputation summary says the company had roughly 50 employees in mid-2026 and more than 90% of staff were in R&D. SO028
CO043 No reviewed official or mainstream source disclosed run-rate revenue, gross margin, backlog conversion, or cash burn. SO004, SO014, SO015, SO016, SO019
CO044 Reviewed sources do not disclose a priced-round post-money valuation for the angel, angel+, or Pre-A rounds. SO014, SO015, SO016, SO020
CO045 No reviewed public source provides an official current headcount disclosure, so employee scale remains unverified beyond low-reputation summaries. SO004, SO014, SO015, SO028
CM001 Xingyuanzhi’s relevant market is the embodied-brain and edge-compute layer for robots, not the full robot-body hardware market. SM001, SM003, SM004
CM002 The company’s relevant included spend covers controller hardware, embodied models, world-model software, edge inference, and deployment integration into multiple robot embodiments. SM001, SM002, SM023
CM003 The same market boundary excludes robot-body manufacturing, actuators, commodity service robots, and facility re-engineering spend. SM003, SM010, SM020
CM004 Status-quo substitutes include fixed industrial robots, AMRs or AGVs, quadrupeds, in-house robot-control stacks, and human labor. SM010, SM011, SM020, SM021
CM005 The independent brain-vendor model is valuable when robot makers want real-time intelligence but do not want to build a full autonomy stack in-house. SM003, SM019, SM022
CM006 Goldman Sachs models a global humanoid-robot market of at least US$6 billion over the next 10-15 years and as much as US$154 billion in a blue-sky 2035 scenario. SM008
CM007 CNBC reported Barclays analyst work that pegs the humanoid market at roughly US$2-3 billion today and US$200 billion by 2035. SM009
CM008 IDTechEx forecasts the humanoid-robot market will reach about US$29.5 billion by 2036. SM010
CM009 Deloitte estimates industrial-use AI humanoid shipments at roughly 15,000 units in 2026 and a resulting market of about US$210-270 million, potentially rising to US$600 million-US$1 billion by 2032. SM019
CM010 IFR says 542,000 industrial robots were installed worldwide in 2024, with annual installations above 500,000 for a fourth straight year. SM013, SM014
CM011 IFR says China installed 295,000 industrial robots in 2024, representing 54% of global deployments, with operational stock above 2 million and domestic suppliers taking 57% local share. SM013, SM014
CM012 EmbodiedGlobal reports that China’s embodied-AI sector raised RMB93.5 billion across 322 deals in H1 2026. SM006
CM013 China Biz Insider reports that China added 15 embodied-AI unicorns in H1 2026 and that many startups have only 18-24 months of cash runway. SM005
CM014 EmbodiedGlobal says about 80% of H1 2026 embodied-AI capital concentrated in Beijing, Guangdong, and Shanghai. SM006
CM015 The serviceable near-term market for Xingyuanzhi is China-based humanoid, industrial mobile, and inspection OEMs buying brains or controllers, not end-consumer home robots. SM002, SM010, SM019, SM020
CM016 Reviewed market reports do not publish a standalone TAM or SAM for embodied-brain vendors as a separate category from robot makers or hardware suppliers. SM008, SM009, SM010, SM019
CM017 Named buyer segments for Xingyuanzhi-style products include humanoid OEMs such as AgiBot, industrial automation players such as Zhongli or EP, local-government operators such as Yizhuang, and inspection-oriented energy or infrastructure groups. SM002, SM004, SM025
CM018 Budget authority is likely to sit with robot-OEM product or R&D teams, factory automation managers, local-government procurement units, or utility operations leaders depending on use case. SM018, SM020, SM021
CM019 Core adoption triggers are labor scarcity, multi-SKU automation demands, real-time edge control, and the desire to avoid building a full autonomy stack internally. SM019, SM020, SM021, SM022
CM020 IDTechEx and Deloitte both identify manufacturing and logistics or warehousing as the first scalable deployment verticals for humanoid or embodied systems. SM010, SM019, SM020
CM021 IDTechEx explicitly treats home-use humanoids as a longer-term opportunity rather than a core 2026-2036 scaling market. SM010
CM022 CNBC’s Barclays-cited market view says China dominates production and deployment and builds humanoids at roughly half the cost of Western competitors, typically around US$50,000. SM009
CM023 China’s 15th Five-Year Plan elevates embodied intelligence to a named future-industry priority and uses directive procurement language around training grounds, model evolution, and deployment. SM016
CM024 HEIS 2026 created a six-pillar national standard system spanning common standards, intelligent computing, components, integrated systems, applications, and safety or ethics. SM015, SM017
CM025 The HEIS standardization effort involved more than 120 institutions and is designed to reduce compatibility and coordination costs across the robotics supply chain. SM015, SM017
CM026 Beijing E-Town’s 2026 humanoid-marathon program paired competition with financing matchmaking, free workspace, affordable compute, and industrial-order rewards above RMB1 million. SM018
CM027 BAAI, Yizhuang, and Beijing policy infrastructure give Xingyuanzhi a local cluster advantage in research talent and pilot deployments. SM002, SM018, SM025
CM028 Standards-led modularization and clearer data or interface rules can accelerate commercialization by lowering integration cost and clarifying safety baselines. SM015, SM017
CM029 Deloitte identifies data quality, integration, interoperability, cybersecurity, and worker safety as core bottlenecks for broader physical-AI adoption. SM019, SM021
CM030 Deloitte’s automotive analysis says humanoids work best today on simple repetitive low-variability tasks, while warehouses remain harder because product and packaging variability is high. SM020
CM031 IFR’s baseline implies China already has an enormous industrial automation installed base, creating a large adjacent opportunity for embodied-intelligence upgrades before true humanoid mass adoption. SM013, SM014
CM032 IFR reports 102,900 transportation and logistics service robots sold in 2024 and says RaaS grew 42%, showing buyer willingness to adopt service-style robot deployment models. SM013
CM033 IFR also reported more than 42,000 hospitality robots and more than 25,000 professional cleaning robots sold in 2024, but Xingyuanzhi’s disclosed use cases skew toward higher-value industrial and inspection work. SM013, SM002
CM034 Faxiangongchang reports China shipped about 14,400 humanoid robots in 2025 and could reach roughly 62,500 in 2026, underscoring rapid unit growth in the home market. SM012
CM035 Deloitte says cumulative installed industrial-robot capacity could reach 5.5 million globally by 2026 even though annual industrial-robot sales have remained roughly flat since 2021. SM019
CM036 Xingyuanzhi’s horizontal supplier model can scale across embodiments, but it also depends on OEMs not insisting on owning the brain themselves. SM003, SM005, SM011
CM037 The current market prices optionality more than proven ROI, as funding volumes and unicorn counts outpace publicly disclosed customer economics. SM005, SM006, SM019
CM038 Public sources do not disclose Xingyuanzhi’s per-unit controller pricing, licensing model, or standard contract structure. SM001, SM002, SM003, SM023
CM039 Chinese-origin standards and dense domestic clusters can give local vendors a home-market procurement and interoperability advantage before equivalent Western frameworks mature. SM015, SM016, SM017
CM040 The near-term market case for Xingyuanzhi is therefore China-first industrial embodied intelligence rather than global consumer robotics. SM002, SM010, SM016, SM019
CP001 Xingyuanzhi’s direct competitive set is narrower than the overall humanoid field because it sells an external embodied-brain layer rather than a full robot body. SP001, SP003, SP005
CP002 The closest direct peers are other firms trying to own embodied models, world models, or robot-brain controllers, such as Spirit AI, X Square Robot, and GigaAI. SP006, SP007, SP008, SP009
CP003 Full-stack robot OEMs like AgiBot and Astribot are both adjacent competitors and potential customers because they may buy external brain components while also building internal intelligence. SP010, SP011, SP012, SP005
CP004 Status-quo substitutes include fixed industrial automation, AMRs, warehouse robots, quadrupeds, and human operators rather than only other humanoid startups. SP016, SP017, SP018, SP020
CP005 Spirit AI competes from a more vertically integrated posture than Xingyuanzhi because it markets both the Spirit V1 foundation model family and the Moz humanoid robot. SP006, SP007
CP006 Spirit AI also appears to have stronger public industrial proof than Xingyuanzhi because RobotToday reports large-scale CATL battery-line deployment with 99%+ insertion success and human-like cycle times. SP007
CP007 X Square Robot competes on research velocity and open model depth, with a public research page showing WALL-OSS, world-model work, and 600+ matched sim-to-real rollouts. SP008
CP008 GigaAI competes on Physical AGI, world-model positioning, and early home and industrial deployments rather than a pure supplier model. SP009
CP009 Astribot is more hardware-forward than Xingyuanzhi, emphasizing a DFAI software-hardware architecture and a wheeled dual-arm S1 robot for household and light commercial tasks. SP010, SP011
CP010 NVIDIA Isaac, GR00T, and Jetson Orin form an incumbent platform layer that can either enable Xingyuanzhi or erode its differentiation if customers assemble their own stack on NVIDIA primitives. SP013, SP014, SP015
CP011 Jetson Orin-class hardware is already powerful enough to support multiple competing edge-AI stacks, which lowers hardware barriers to entry for rival embodied-intelligence vendors. SP014, SP015
CP012 EP Equipment shows that some buyer jobs Xingyuanzhi might chase can also be solved by lower-complexity warehouse automation sold as products with public pricing, rapid installation, and sub-one-year ROI narratives. SP016, SP018, SP019
CP013 Unitree-style quadruped inspection solutions are substitutes in dangerous and repetitive industrial tasks, especially where humanoid dexterity is unnecessary. SP020
CP014 Xingyuanzhi’s core differentiation is supplier neutrality: it wants many robot OEMs as customers rather than competing directly for finished robot sales. SP002, SP003, SP005
CP015 That neutrality is strategically valuable only if OEMs believe external brain vendors can move faster or cheaper than internal model teams. SP005, SP013, SP023
CP016 AgiBot is therefore a structurally ambiguous counterpart: it validates demand for Xingyuanzhi’s layer but also represents the long-term internal-build threat. SP005, SP012
CP017 Public pricing transparency is weak across embodied-AI competitors, which makes the competitive battle look more like enterprise solution selling than standard product commerce. SP006, SP008, SP009, SP010, SP012
CP018 EP’s XP15 is a useful contrast because it publishes both rental and purchase pricing, showing that at least some adjacent robotics substitutes compete with explicit economics rather than opaque strategic partnerships. SP018, SP019
CP019 Switching costs in embodied intelligence are likely to come from integration work, data pipelines, safety validation, and installed customer workflows rather than sticker price alone. SP015, SP017, SP023
CP020 Multi-homing remains plausible for OEMs because they can mix NVIDIA primitives, internal software, and external brain components instead of choosing a single all-or-nothing stack. SP013, SP015, SP005
CP021 Spirit AI’s nearly RMB2 billion financing and reported three-city footprint suggest a scale advantage in recruiting, deployment support, and experimentation budget. SP006, SP007
CP022 EmbodiedGlobal’s report of 22 unicorns in H1 2026 implies that Xingyuanzhi faces a crowded capital-backed field where many rivals can subsidize commercialization. SP021
CP023 China Biz Insider’s adverse view suggests many embodied-AI startups still lack durable commercialization, so funding size alone is a poor moat signal. SP022
CP024 Policy support and China’s dense robotics ecosystem may advantage domestic vendors collectively, but it does not guarantee that a horizontal brain supplier rather than a full-stack OEM captures the margin pool. SP024, SP025, SP022
CP025 X Square’s research cadence shows that open and semi-open model ecosystems can compress Xingyuanzhi’s technology moat if its own benchmarks remain private. SP008, SP005
CP026 Spirit AI’s public manufacturing-line metrics make it a more threatening competitor for industrial accounts than peers focused mainly on household or research narratives. SP007, SP011, SP023
CP027 GigaAI’s home-trial posture suggests it competes more strongly for future household assistants than for Xingyuanzhi’s near-term industrial brain-supplier niche. SP009, SP023
CP028 Astribot’s S1 differentiates on dexterous household manipulation, but that same focus makes it less directly comparable to Xingyuanzhi’s cross-embodiment controller thesis. SP010, SP011
CP029 NVIDIA’s open robotics development stack reduces time to market for new entrants, increasing commoditization risk around generic perception, simulation, and deployment components. SP013, SP015
CP030 Warehouse operators can often defer embodied-intelligence purchases by buying simpler AGV or AMR systems first, which gives low-complexity substitutes a meaningful sales advantage. SP016, SP018, SP019
CP031 Because Xingyuanzhi does not own the finished robot, it may have weaker end-customer brand power than full-stack players whose hardware becomes the visible product. SP003, SP010, SP012
CP032 Conversely, the brain-only model can widen distribution if multiple OEMs adopt the same stack, creating an ecosystem position more like a component standard than a robot brand. SP002, SP003, SP004
CP033 Public evidence suggests competitor trust posture today comes more from deployment proof and ecosystem partnerships than from disclosed compliance or security credentials. SP006, SP007, SP008, SP009, SP010
CP034 Xingyuanzhi’s moat is therefore less about exclusive hardware and more about whether its controller-plus-model layer can become embedded in partner workflows before OEMs internalize the capability. SP005, SP015, SP023
CP035 The company’s reliance on named strategic partners makes supply and partner access part of the competitive game, not merely a distribution afterthought. SP002, SP005
CP036 Xingyuanzhi’s public materials still do not disclose list pricing, benchmark superiority, or long-term exclusivity terms with customers, which limits confidence in moat durability. SP001, SP002, SP005
CP037 Spirit AI, X Square, GigaAI, Astribot, and AgiBot together show that most serious Chinese embodied-AI rivals combine models with at least some owned robotic embodiment or data-acquisition system. SP006, SP008, SP009, SP010, SP012
CP038 That pattern makes Xingyuanzhi’s pure supplier stance distinctive, but it also means its closest analogs are scarce and investors must compare it against both component vendors and robot OEMs. SP003, SP005, SP013
CI001 Xingyuanzhi’s revenue model is B2B and centered on supplying a robot-brain stack rather than selling finished robot bodies. SI001, SI007, SI010
CI002 That stack appears to combine controller hardware, embodied models, and deployment integration, making the business financially closer to hardware-enabled enterprise infrastructure than pure software. SI002, SI004, SI005
CI003 Likely monetization components include hardware sales, software or model attach, customization, deployment support, and ongoing maintenance, although public sources do not break the mix out explicitly. SI004, SI010, SI015
CI004 The company’s GTM motion appears to be direct enterprise and strategic-partner selling rather than self-serve or channel-led volume commerce. SI002, SI006, SI024
CI005 Revenue recognition quality is difficult to judge because public materials mix shipped units, booked revenue, projected orders, and strategic-partnership announcements. SI010, SI011
CI006 Public list pricing for T5, N5, or Xingyuanzhi software attach is not disclosed in retained sources. SI001, SI004, SI005
CI007 Lavx reports that Xingyuanzhi shipped hundreds of T5 units in 2025. SI010
CI008 Lavx also reports more than RMB10 million of 2025 revenue. SI010
CI009 News Globe Now similarly says the company shipped several hundred units with revenue exceeding RMB10 million. SI011
CI010 News Globe Now says the company aims to ship nearly 10,000 units in 2026, but this is a target rather than audited revenue. SI011
CI011 36Kr says the latest financing will fund next-generation embodied-brain and world-model R&D, scale production, and talent hiring. SI009
CI012 Lavx says the company has about 50 staff and that more than 90% are in R&D, implying a cost structure dominated by engineering rather than field sales. SI010
CI013 The disclosed cost structure is likely lighter than a full robot OEM’s because Xingyuanzhi avoids manufacturing full bodies and actuators. SI007, SI014
CI014 The same model still carries meaningful COGS because controllers, edge compute, integration, and field support are not pure-software expenses. SI004, SI019, SI020
CI015 The official product surface now highlights N5 and English product copy highlights a Jetson Thor compact platform, indicating continuing hardware-refresh needs alongside the earlier T5 platform. SI003, SI004, SI005
CI016 NVIDIA silicon and software dependencies imply supplier concentration and potentially meaningful inventory or platform-transition costs. SI019, SI020
CI017 The revenue stream likely includes lumpy project economics because customers appear to buy through strategic deployments rather than standardized subscriptions. SI002, SI010, SI024
CI018 Public GTM evidence suggests a long enterprise sales cycle involving pilots, co-development, or scenario-specific integration before scaled rollout. SI015, SI021, SI025
CI019 No retained public source discloses CAC, payback, conversion rates, or direct sales-efficiency metrics. SI001, SI007, SI010, SI013
CI020 Adjacent robotics substitutes such as EP XP15 publish both rental and purchase pricing, underlining how little monetary transparency Xingyuanzhi currently gives investors. SI017, SI018
CI021 EP also markets one-week delivery, four-week rollout, and sub-one-year ROI messaging, which is the kind of economic benchmark Xingyuanzhi must eventually beat in overlapping logistics use cases. SI015, SI018
CI022 Working-capital visibility is poor because public sources do not disclose inventory, receivables, warranty reserves, or support obligations. SI001, SI012, SI013
CI023 No retained public source discloses cash on hand, monthly burn, or runway for Xingyuanzhi. SI007, SI012, SI013
CI024 China Biz Insider’s warning that many embodied-AI startups have only 18-24 months of runway is sector context rather than Xingyuanzhi-specific proof, but it raises concern about financing dependency. SI023
CI025 Against that opacity, RMB1 billion raised versus only ~RMB10 million publicly reported 2025 revenue implies Xingyuanzhi remains highly financing dependent. SI008, SI010, SI011
CI026 If the disclosed 2026 shipment target converts, scale could improve revenue and purchasing leverage materially, but the target is too preliminary to underwrite. SI011, SI022
CI027 Sources conflict on the identity of the counterparty behind the projected RMB500 million-plus three-year order opportunity, with one source tying it to Beijing Yizhuang Robot and another to EP Equipment. SI010, SI011
CI028 Because that RMB500 million figure is forward-looking order potential rather than recognized revenue, it should not be treated as proof of current financial scale. SI010, SI011
CI029 The public financial record does not disclose debt, project-finance obligations, or special financing facilities. SI012, SI013
CI030 The company registry page confirms core corporate basics such as Beijing address and contact details but does not provide operating financial statements in the retained evidence. SI012
CI031 PitchBook’s public profile preview confirms the existence of a valuation-and-investor tracking profile but does not expose enough public financial detail to replace primary diligence. SI013
CI032 The combination of high R&D concentration, ongoing hardware refresh, and opaque pricing means gross-margin path is unproven even if topline grows quickly. SI010, SI015, SI019
CI033 Revenue quality today should be treated as mixed because commercial shipments are real, but supporting metrics on repeat purchases, service attach, and collections are not public. SI010, SI011, SI024
CI034 Customer concentration risk is likely meaningful because only a small number of named strategic counterparties are publicly associated with the business. SI002, SI006, SI024
CI035 The company’s use-of-funds plan prioritizes growth and technical moat building over near-term profitability. SI009, SI022
CI036 Service-delivery costs likely include on-site integration, training, tuning, and support because embodied deployments rarely behave like zero-touch software rollouts. SI015, SI016, SI021
CI037 Public evidence does not allow a defensible estimate of gross margin, contribution margin, or breakeven shipment volume. SI001, SI012, SI013
CI038 On current evidence, Xingyuanzhi is commercially real but still financially under-disclosed, making the business directionally promising and not yet fully underwritable. SI008, SI010, SI011, SI023
CI039 Active public product and news cadence at AGIBOT reinforces that key counterparties are themselves fast-moving product companies, which raises the risk that partner-customers eventually internalize more of the stack. SI024, SI030
CI040 Public patent-search surfaces exist, but the retained public review did not surface a company-specific patent corpus detailed enough to support an IP-backed financial underwriting case. SI028, SI029
CI041 NVIDIA ecosystem pages emphasize platform breadth rather than solution-level pricing, so component-cost benchmarking for Xingyuanzhi remains incomplete even when the dependency is obvious. SI026, SI027
CE001 Xingyuanzhi defines its product mission as building a general-purpose embodied brain for the physical world. SE001, SE006
CE002 The company positions itself around multimodal spatial intelligence and cross-embodiment generalization rather than a single robot form factor. SE001, SE006
CE003 The public workflow claim is not just perception but planning and decision-making for robots operating in the physical world. SE001, SE006, SE010
CE004 Official use cases span shelf picking, inspection and guidance, eldercare, food service, and intelligent loading and unloading. SE003
CE005 The company’s technical stack combines compute hardware with embodied AI models rather than shipping software alone. SE004, SE005, SE010
CE006 Public sources point to at least two platform generations: an earlier T5 embodied-brain compute platform and a newer product surface focused on N5. SE002, SE004, SE005
CE007 The English product page describes N5 as a compact Jetson Thor platform for on-device deployment. SE005
CE008 Pandaily and Lavx both describe T5 as a robot AI-brain/domain-controller platform rather than a complete robot. SE007, SE010
CE009 Lavx says the platform is paired with general-purpose embodied AI models that run edge inference locally in real time. SE010
CE010 Lavx also identifies RoboBrain Pro as the system used for commercial loading and unloading work with EP Equipment. SE010
CE011 The partner and customer surface implies the product is intended to work across humanoid and industrial embodiments rather than one dedicated robot body. SE003, SE021
CE012 Jetson Orin-class hardware offers up to 275 TOPS, illustrating the edge-compute envelope within which Xingyuanzhi’s earlier T5 generation likely operated. SE011
CE013 NVIDIA’s Orin software documentation shows the robotics stack depends on Linux, sensor interfaces, bootloaders, and camera/IMU support, underscoring that embodied-brain delivery is a systems-engineering problem, not just model training. SE012
CE014 Isaac ROS Common shows the surrounding robotics ecosystem already provides reusable packages, scripts, and test infrastructure for robotics deployments. SE013
CE015 Isaac ROS Visual SLAM shows that GPU-accelerated localization and mapping is available as a public developer building block with benchmark detail and ROS 2 packaging. SE014
CE016 The retained GitHub pages indicate a strong external practitioner ecosystem around navigation, mapping, and pose estimation, which reduces the moat of any vendor relying on generic robotics primitives. SE013, SE014, SE015, SE016
CE017 Isaac Sim documentation shows how much of modern robot development depends on simulation, synthetic data, ROS 2 wiring, and software-in-the-loop validation. SE017
CE018 Isaac GR00T demonstrates that open humanoid-foundation-model tooling now exists from a platform incumbent, increasing commoditization pressure on generic embodied-model claims. SE018
CE019 36Kr says the latest funding is being used for next-generation embodied-brain and world-model R&D, which implies the current product is still evolving rather than frozen. SE009
CE020 The official product surface foregrounds N5 more prominently than T5, suggesting portfolio refresh or architectural iteration during 2026. SE004, SE005
CE021 The product appears commercially deployed but still early in proof of generalized reliability because public sources do not expose standardized success-rate, latency, or uptime benchmarks for Xingyuanzhi itself. SE007, SE008, SE010
CE022 Deployment likely requires partner robot bodies, customer-specific tuning, and site integration rather than zero-touch delivery. SE003, SE010, SE021
CE023 Because the company sells a brain layer, its technical quality depends on integration with third-party embodiment, sensors, and workflow constraints. SE003, SE010, SE021
CE024 Critical dependencies include NVIDIA compute, partner robot embodiments, embodied data, and the customer workflows used to collect or validate performance. SE005, SE011, SE017, SE021
CE025 BAAI incubation and the Beijing robotics ecosystem likely strengthen access to talent, research, and pilot settings. SE006, SE022, SE023
CE026 The public site does not expose a dedicated trust, safety, or security page for the product. SE001, SE004, SE005
CE027 Retained public sources do not expose certifications, safety cases, or security attestations for Xingyuanzhi deployments. SE001, SE003, SE020
CE028 Conference and ecosystem visibility can support commercial credibility, but they do not substitute for validation-grade benchmark or reliability evidence. SE022, SE023
CE029 Public patent-search surfaces exist, but the retained search did not provide a clean company-specific patent set that can be used as hard proof of proprietary technical moat. SE019, SE025
CE030 The product surface does not provide public API, SDK, or detailed performance tables comparable to open developer ecosystems in the surrounding robotics stack. SE001, SE004, SE013, SE014
CE031 Xingyuanzhi’s public technical differentiation rests more on architecture and cross-embodiment positioning than on published benchmark leadership. SE001, SE006, SE010
CE032 The customer and partner surfaces imply the architecture is intended to serve both humanoid and industrial material-handling or inspection contexts. SE003, SE021, SE022
CE033 Open and incumbent robotics platforms like Isaac ROS, Isaac Sim, and GR00T increase commoditization pressure around generic perception, simulation, and training layers. SE013, SE017, SE018
CE034 Those same open tools raise the bar for Xingyuanzhi to prove that its proprietary integration, data, or deployment know-how is better than the reference stack customers could assemble themselves. SE013, SE014, SE018
CE035 A platform that needs heavy per-customer tuning would weaken the economics of the cross-embodiment thesis even if demos are strong. SE010, SE017
CE036 The absence of public trust and reliability artifacts leaves safety, uptime, and security as unresolved product-quality risks. SE001, SE004, SE005
CE037 The surrounding practitioner stack has visible developer momentum, but Xingyuanzhi lacks a comparable public developer surface, making integration depth hard to assess from outside. SE013, SE014, SE015, SE016
CE038 Overall product maturity looks strongest at the concept and partner-demo level, weaker at public trust evidence, and only partially visible at detailed operating benchmarks. SE003, SE010, SE017, SE022
CU001 Xingyuanzhi’s customer base appears segmented across robot OEMs, industrial automation partners, and public-sector ecosystem operators. SU001, SU002, SU005
CU002 Publicly named counterparties include AgiBot, Beijing Yizhuang Robot, and Zhongli/EP-linked industrial workflows. SU001, SU005, SU008
CU003 36Kr reports that Xingyuanzhi’s solutions cover more than 70% of leading embodied-intelligence companies, implying broad ecosystem reach even if account names remain sparse. SU007
CU004 The visible customer footprint is heavily China-first rather than international. SU001, SU020, SU024
CU005 Lavx and News Globe Now both report hundreds of T5 shipments and more than RMB10 million of revenue in 2025, establishing non-zero commercial adoption. SU005, SU008
CU006 The reported RMB500 million-plus three-year strategic opportunity indicates at least one anchor account may have material expansion potential, even though contract detail is not public. SU005, SU008
CU007 Named customer evidence is stronger at the strategic-partnership level than at the end-customer production-outcome level. SU001, SU005, SU008
CU008 AgiBot is a meaningful validation counterparty because it is a major humanoid OEM rather than a small pilot customer. SU001, SU011
CU009 AGIBOT’s official 2026 WAIC article says more than 60 robots were operating across venues, showing the partner is running real deployments in public and service environments. SU009
CU010 The same AGIBOT article cites industrial deployments with partners such as Longcheer and PIA Automation, including 3,000 units per shift, 64+ hours of cumulative continuous operation, and downtime below 4% on one line. SU009
CU011 PR Newswire says AGIBOT had rolled out its 10,000th robot by March 2026, which makes any Xingyuanzhi integration into the AGIBOT ecosystem strategically relevant if it deepens. SU010
CU012 AGIBOT is simultaneously a customer-proof source and a structural customer-concentration risk because it is developing its own full embodied stack. SU009, SU010, SU011, SU025
CU013 EP-linked workflow evidence shows Xingyuanzhi is targeting real logistics and handling jobs rather than purely showcase humanoid demos. SU008, SU012, SU015
CU014 EP case studies show the kinds of downstream workflows that matter for Xingyuanzhi-linked deployments: outbound transport, repetitive warehouse loops, and long-distance production-hall movement. SU012, SU013, SU014
CU015 Those EP case studies are production-adjacent workflow proof for the category, but they do not directly disclose how much of the value stack comes from Xingyuanzhi itself. SU012, SU013, SU014, SU015
CU016 The Beijing Yizhuang Robot relationship appears more ecosystem and pilot-oriented than revenue-transparent, based on the visible public record. SU001, SU005, SU020, SU021
CU017 WRC 2026 and E-Town event infrastructure make Yizhuang an unusually dense discovery and pilot channel for embodied-AI customer acquisition. SU016, SU017, SU020
CU018 ChinaPower’s robotics analysis supports the idea that Beijing’s robotics hubs offer unusually favorable conditions for customer acquisition and deployment density. SU018, SU020
CU019 Rest of World’s reporting suggests that some ecosystem relationships in embodied AI may also be serving data-collection and training objectives, not just immediate revenue. SU019
CU020 Public evidence does not disclose active account count, deployed-site count, or installed-base denominator for Xingyuanzhi customers. SU001, SU003, SU005
CU021 Public evidence also does not disclose GRR, NRR, churn, renewal rates, or cohort data. SU001, SU002, SU005
CU022 The best durability proxy in the public record is ongoing partner cadence and multi-year order projection rather than actual retention metrics. SU005, SU006, SU025
CU023 Because named logos are few and strategic, customer concentration risk is likely high even if total ecosystem reach is broad. SU001, SU007, SU011
CU024 The likely customer journey is relationship-led: ecosystem awareness, pilot or strategic cooperation, embodied integration, then site-level scaling. SU005, SU016, SU020
CU025 Land-and-expand is plausible through more embodiments per OEM, more sites per industrial operator, and more task types per venue or customer. SU001, SU009, SU010
CU026 Public procurement and industrial-integration friction likely remain meaningful because the visible customer set skews toward enterprises, OEMs, and public-sector ecosystems rather than self-serve buyers. SU012, SU016, SU020
CU027 Official public surfaces do not disclose user satisfaction scores, testimonials tied to measured outcomes, or formal customer references beyond logos and partnership descriptions. SU001, SU002, SU003
CU028 Named customer proof is fresh in 2026 but outcome specificity is mixed, with the strongest concrete operating data coming from AGIBOT’s own deployments rather than Xingyuanzhi-specific KPI disclosure. SU009, SU010, SU025
CU029 The public record does not reveal revenue share by customer, contract length, or whether any named accounts have already renewed. SU005, SU008, SU024
CU030 AgiBot and EP together imply Xingyuanzhi’s customers are buying into serious operating workflows rather than consumer gadget channels. SU008, SU009, SU012
CU031 At the same time, simpler EP workflows show that some buyer needs can be met by focused automation without needing Xingyuanzhi’s broader embodied-brain stack. SU013, SU014, SU015
CU032 China’s embodied-AI funding surge and unicorn formation raise the chance that logos and pilots are abundant before retention and recurring economics are proven. SU022, SU023
CU033 The customer record therefore supports a real-world industrial and OEM demand thesis, but not yet a broad or diversified customer-base thesis. SU001, SU005, SU008, SU022
CU034 Customer-proof evidence is stronger for buyer relevance than for retention durability. SU009, SU012, SU021
CU035 A Beijing-first ecosystem strategy may help customer acquisition speed, but it can also increase geographic concentration in early revenue. SU017, SU018, SU020
CU036 Overall, Xingyuanzhi’s public customer proof is promising, strategically important, and still pre-retention. SU005, SU009, SU021, SU022
CR001 Xingyuanzhi’s highest residual risks appear to be compute supply dependence, partner/customer internalization, and under-disclosed economics rather than lack of market demand. SR005, SR007, SR010, SR013, SR014
CR002 The company positions itself as a robot-brain vendor rather than a robot-body manufacturer. SR005, SR030
CR003 Public company materials and third-party reporting tie Xingyuanzhi’s platforms to NVIDIA-class edge compute, making external silicon and platform continuity material to execution. SR002, SR003, SR005, SR007
CR004 Trade.gov explicitly warns that advanced computing integrated circuits shipped to China face evolving license requirements and end-user scrutiny, making high-end compute access a real geopolitical risk. SR013, SR014
CR005 BIS guidance published in 2026 says a license is required for advanced-computing items shipped to entities headquartered in Country Group D:5, reinforcing the risk that China-linked robotics companies face supply or diligence friction. SR013, SR014
CR006 Because Xingyuanzhi sells the brain layer and depends on partner embodiments, any export-control, hardware-refresh, or platform-policy disruption can transmit directly into deployment delays. SR003, SR005, SR013, SR014
CR007 Hill Dickinson’s 2026 legal analysis highlights unresolved liability allocation among software provider, manufacturer, and operator when humanoid systems cause harm. SR015
CR008 That liability ambiguity is especially relevant for Xingyuanzhi because it does not control the full stack from robot body through operating environment. SR001, SR005, SR015
CR009 The EU AI Act imposes documentation, traceability, human-oversight, robustness, and cybersecurity obligations on high-risk AI, which would raise compliance costs for any future expansion into regulated markets. SR016
CR010 NIST’s AI RMF and the 2026 critical-infrastructure concept note underline that trustworthy AI increasingly requires formal risk-management artifacts, not only demo performance. SR017
CR011 CISA’s AI guidance shows that secure deployment and cybersecurity hardening are expected parts of modern AI operations, especially where systems touch critical workflows. SR018
CR012 Xingyuanzhi’s public site does not expose a rich trust center, product-security page, incident page, or detailed privacy/safety documentation comparable to mature enterprise platforms. SR001, SR003, SR030
CR013 The absence of public trust artifacts does not prove weak controls, but it materially increases diligence burden and leaves residual legal and customer-acceptance risk. SR012, SR017, SR018
CR014 OSHA notes that robot accidents often occur during non-routine activities such as programming, maintenance, testing, setup, or adjustment. SR019
CR015 Those non-routine failure modes matter for Xingyuanzhi because a brain-layer vendor still has to survive integration, commissioning, and operator handoff at customer sites. SR003, SR014, SR019
CR016 IFR’s 2026 note says real-world production capability for humanoids remains limited and many deployments are still demonstrators or pilot projects. SR011
CR017 Deloitte’s physical-AI analysis identifies the reality gap, trust and safety, regulatory change, and data complexity as the major barriers to scaling embodied systems. SR021
CR018 Together, those sources support a view that generalization and production hardening remain significant operational risks even when pilot demos look compelling. SR011, SR021
CR019 EmbodiedGlobal reports ¥93.5 billion of China embodied-AI funding in H1 2026 and 22 unicorns, signaling unusually intense competition for talent, capital, and customer mindshare. SR010
CR020 China Biz Insider explicitly frames the sector’s 2026 surge as one where a reality check looms, strengthening the case that hype risk is material. SR009
CR021 AgiBot is simultaneously Xingyuanzhi’s best visible validation logo and a future internalization threat because large humanoid OEMs are building more of their own embodied stack. SR001, SR024, SR025
CR022 36Kr’s statement that Xingyuanzhi solutions cover more than 70% of leading embodied-intelligence companies implies ecosystem breadth, but it also implies a service and integration load that can stretch a young team. SR006, SR004
CR023 Beijing and Yizhuang conference infrastructure create strong demand-generation density for the company, but they also concentrate reputation and pipeline risk geographically. SR026, SR027, SR028
CR024 ChinaPower’s robotics analysis supports the view that robotics adoption in China is unusually cluster-driven and policy-supported, which helps sales but raises policy-dependence risk. SR026, SR029
CR025 Public sources do not disclose revenue concentration, renewal rates, GRR, NRR, or active-account counts for Xingyuanzhi. SR001, SR005, SR030
CR026 That disclosure gap means concentration risk should be assumed high until a top-customer schedule proves otherwise. SR001, SR021, SR025
CR027 Rest of World’s reporting on large-scale robotics data collection in China suggests data provenance, worker privacy, and governance can become material issues in the embodied-AI stack. SR023
CR028 Any company pursuing multimodal spatial intelligence without strong public governance artifacts faces heightened diligence questions around biometric, video, and training-data handling. SR015, SR016, SR023
CR029 Global regulatory divergence is real: China is accelerating deployment while EU-style regimes emphasize liability, transparency, and rights protection. SR012, SR015, SR016, SR020
CR030 That divergence could advantage China-first growth in the short term while making cross-border commercialization more compliance-intensive later. SR011, SR012, SR016, SR020
CR031 QCC shows the company was incorporated on 2025-08-01, so the organization is still unusually young for the scope of risk it is now carrying. SR004
CR032 Raising roughly ¥1 billion in about ten months is a strength, but it also creates pressure to scale headcount, product maturity, and commercial proof quickly enough to support the next financing step. SR005, SR006, SR007
CR033 Public sources still do not disclose revenue, gross margin, cash balance, debt load, or burn rate, so financial-model risk remains materially under-observed. SR005, SR006, SR008
CR034 In a bubble-prone market, opaque economics increase the probability that valuation and burn discipline can decouple from real deployment traction. SR009, SR010, SR025
CR035 EP-style industrial tasks and warehouse motion are also addressable by simpler automation, so embodied-brain vendors face substitution and price-pressure risk before full humanoid adoption arrives. SR008, SR011, SR021
CR036 Deloitte’s March 2026 NVIDIA collaboration release argues that simulation-led testing and secure edge AI can reduce downtime and accelerate safe deployment. SR022
CR037 Those mitigants matter for Xingyuanzhi, but public evidence does not yet show company-specific benchmark packs, model cards, or validation tooling maturity. SR003, SR017, SR022
CR038 Public materials do not clearly disclose board composition, product-liability insurance, or a named governance structure deep enough to evaluate key-person and oversight risk. SR004, SR030
CR039 A small 50-150 person organization pursuing platform R&D, integrations, and ecosystem coverage is likely dependent on scarce technical and deployment talent. SR006, SR019, SR021
CR040 The minimum public-market diligence package should include chip sourcing and ECCN review, safety incident logs, customer concentration schedule, and an org chart covering leadership and technical owners. SR013, SR017, SR018, SR025
CR041 The most practical thesis-break triggers are anchor-customer internalization, inability to source or qualify advanced compute, failure to publish credible deployment metrics before the next financing event, and any meaningful safety or compliance incident. SR013, SR021, SR024, SR025
CR042 Overall residual risk remains high: market tailwinds and strong funding help, but they do not eliminate the company’s dependencies, disclosure gaps, or commercialization uncertainty. SR010, SR011, SR025
CV001 Public sources consistently indicate Xingyuanzhi raised roughly ¥1 billion across its first ten months, with the latest financing discussed in June 2026. SV004, SV005, SV006, SV007
CV002 36Kr explicitly frames Xingyuanzhi as a new unicorn, supporting the view that the company entered the unicorn band by mid-2026. SV005
CV003 China Biz Insider’s 2026 unicorn article reinforces that embodied-AI valuations in China reached about $1.4 billion territory in H1 2026, providing a plausible market anchor for Xingyuanzhi’s band even if not a precise term-sheet value. SV010, SV005
CV004 Public evidence does not disclose enough revenue, burn, cash, or margin data to support a conventional bottom-up valuation model. SV004, SV005, SV007
CV005 That data gap makes a buy recommendation premature even if the company’s strategic position is attractive. SV003, SV004, SV010
CV006 The most defensible recommendation today is research-more rather than buy or outright avoid. SV004, SV010, SV028
CV007 The current risk rating is high because valuation evidence is weaker than market enthusiasm and because chip, partner, and execution dependencies remain material. SV010, SV026, SV027, SV028
CV008 The current valuation stance is stretched relative to public evidence quality if the company is already being priced at or above the unicorn band. SV003, SV005, SV010
CV009 Feed the AI’s 2026 tracker shows physical-AI and robotics rounds ranging from about $145 million to more than $500 million, placing Xingyuanzhi’s roughly $140 million total raise near the lower end of headline mega-rounds rather than at the frontier. SV008, SV004
CV010 EmbodiedGlobal’s ¥93.5 billion H1 2026 funding figure shows that Xingyuanzhi is fundraising into an unusually well-capitalized Chinese embodied-AI market. SV009
CV011 Precedence Research and Global Market Insights both forecast substantial humanoid-robot market growth through the 2030s, supporting the existence of real strategic upside. SV013, SV014
CV012 Those forecasts diverge sharply in starting size and long-term magnitude, so they support directionally large upside but not a narrow present-day point estimate. SV013, SV014
CV013 Public robotics market caps span a very wide range, from Serve Robotics at about $0.43 billion to UBTECH at $5.36 billion and Symbotic at $24.13 billion as of August 2026. SV015, SV016, SV017
CV014 That spread shows that stage, business model, and proof depth matter far more than the generic label “robotics company.” SV015, SV016, SV017, SV028
CV015 UBTECH is a more relevant public reference than Symbotic or NVIDIA because it is a China-rooted humanoid robotics company rather than a warehouse-automation systems leader or global semiconductor platform. SV015, SV016, SV018
CV016 Symbotic is better interpreted as an upper-bound execution comp showing how high valuations can go once deployment proof and enterprise scale are much deeper than Xingyuanzhi’s current public record. SV016, SV028
CV017 Serve Robotics demonstrates that public markets can place sub-$1 billion values on robotics companies when proof is narrower or sentiment cools. SV017
CV018 NVIDIA’s trillions-scale market cap is not a direct valuation comparable, and Tesla’s much larger market cap likewise reflects a very different maturity and ambition set, but together they illustrate how much value can accrue to layers that become indispensable at scale. SV018, SV030, SV031
CV019 Xingyuanzhi’s brain-layer positioning gives it a plausible horizontal-platform upside case if it can become a standard software and controller layer across multiple OEMs. SV004, SV019, SV030
CV020 BAAI incubation materially improves technical credibility and talent signaling, but it does not substitute for revenue, retention, or contract disclosure. SV019, SV005
CV021 Public customer proof with AgiBot and Beijing ecosystem partners is strategically meaningful, but it still lacks the revenue specificity needed to support a premium price today. SV001, SV022, SV023, SV024
CV022 Because Xingyuanzhi is not selling the robot body, it may deserve a software-style strategic premium if it proves cross-embodiment attach rates and recurring economics. SV004, SV006, SV030
CV023 Spirit AI’s reported nearly ¥2 billion financing and stronger published industrial proof limit the premium Xingyuanzhi can command purely on narrative today. SV020, SV021
CV024 The 2026 Chinese embodied-AI boom reduces scarcity premium because investors have many capitalized alternatives, not just Xingyuanzhi. SV009, SV010
CV025 The best valuation method here is milestone-and-scenario analysis rather than a straight revenue multiple because the revenue denominator is not publicly visible. SV004, SV013, SV015
CV026 A plausible bull case requires Xingyuanzhi to become a neutral brain standard across major Chinese OEMs, expand beyond a few named logos, and publish credible deployment and retention evidence. SV001, SV005, SV019, SV030
CV027 Under that bull case, a valuation range around $2.0-2.8 billion is arguable, but only if market sentiment stays supportive and commercialization proof deepens materially. SV010, SV014, SV016
CV028 The base case is a China-first platform supplier that retains key partners, expands selectively, and raises again without proving broad recurring economics. SV001, SV009, SV023
CV029 That base case supports an illustrative valuation range around $1.2-1.7 billion, close to the current unicorn band but not obviously cheap. SV010, SV015, SV023
CV030 The bear case combines partner internalization, compute friction, and cooling investor sentiment before Xingyuanzhi publishes durable economic proof. SV021, SV026, SV027
CV031 That bear case can justify a sub-unicorn valuation range around $0.7-1.0 billion. SV017, SV021, SV026
CV032 All scenario values in this chapter are illustrative ranges, not precise price targets, because term-sheet structure and financial statements are absent from the public record. SV003, SV004, SV013
CV033 Thesis-break triggers include loss of advanced-compute access, flagship partner internalization, failure to publish credible deployment metrics before the next financing event, and any material safety or compliance incident. SV022, SV026, SV027, SV028
CV034 The most valuable diligence requests are current revenue and burn, customer concentration, contract economics with flagship partners, and the exact cap-table and preference stack. SV003, SV004, SV021
CV035 Bubble commentary is relevant: China Biz Insider and the broader funding statistics both imply real down-round or multiple-compression risk if sector sentiment turns before proof catches up. SV009, SV010
CV036 Advanced-compute export controls and future cross-border AI compliance requirements cap the plausible upside multiple because they can narrow TAM and increase execution cost. SV026, SV027, SV028
CV037 Near-term exit readiness appears low because the company remains privately financed, young, and under-disclosed relative to IPO-quality expectations. SV003, SV004, SV011
CV038 Strategic M&A optionality exists if Xingyuanzhi proves that its embodied-brain layer materially improves attach rate or operating economics for larger OEMs and industrial platforms. SV018, SV023, SV030
CV039 A price below or near the lower end of the unicorn band would be materially more interesting than paying a clear premium above it without new disclosures. SV010, SV017, SV029
CV040 At or above roughly $1.4 billion, the public record supports strategic interest and tracking, but not a buy call. SV005, SV010, SV029
CV041 Overall conviction should remain medium at best because the strategic thesis is coherent but many valuation-critical inputs remain private. SV004, SV010, SV030
CV042 The chapter’s final posture is research-more, medium confidence, high risk, and stretched valuation stance unless diligence produces a step-change in economic proof. SV004, SV008, SV010
来源
编号出版方标题引文
SO001 Xingyuanzhi Robotics 星源智 北京星源智机器人科技有限公司成立于2025年8月1日,系北京智源人工智能研究院孵化的具身智能公司。
SO002 Xingyuanzhi Robotics 关于我们 - 星源智 与北京亦庄机器人科技产业发展有限公司签署战略合作协议,三年内合作完成不低于5亿元订单,共建生态闭环。
SO003 Xingyuanzhi Robotics XYZ - About AgiBot released Genie G2, equipped with XYZ’s Embodied Brain Domain Controller T5.
SO004 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 公司成立10个月内完成三轮融资,累计融资突破10亿元人民币。
SO005 Xingyuanzhi Robotics 产品中心 - 星源智 具身大脑算力平台 N5。
SO006 Xingyuanzhi Robotics Plug-and-Play Deployment Across Robot Embodiments - XYZ The 2560-core NVIDIA Blackwell architecture GPU with 96 fifth-generation Tensor Cores...
SO007 Xingyuanzhi Robotics 新闻动态 - 星源智 星源智完成Pre-A轮融资,用“世界模型”加速具身智能代际跃迁。
SO008 Xingyuanzhi Robotics 合作伙伴 - 星源智 北京市海淀区海淀大街3号鼎好大厦A座2层203-1
SO009 Xingyuanzhi Robotics XYZ Embodied AI - English homepage XYZ Embodied AI Co., Ltd. was founded on August 1, 2025. Incubated by the Beijing Academy of Artificial Intelligence.
SO010 Xingyuanzhi Robotics XYZ - News News and Information.
SO011 BAAI Community 2026智源大会议程公开 | 具身智能CEO华山论剑,产业爆发前的关键判断 刘东丨星源智创始人&CEO,智源研究院具身脑研究中心PI
SO012 Baidu Baike 北京星源智机器人科技有限公司 2026年6月3日,公司完成Pre-A轮融资。公司自2025年8月1日成立以来,在10个月内累计融资额达10亿元人民币。
SO013 Baidu Baike Beijing Xingyuan Zhi Robot Technology Co., Ltd. Its legal representative is Liu Dong. It is a technology company incubated by the Beijing Zhiyuan Research Institute.
SO014 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SO015 36Kr Xingyuanzhi completed a new round of financing So far, it has raised a total of RMB 1 billion.
SO016 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time On the commercialization front, Xingyuanzhi’s hardware-software integrated solutions cover more than 70% of the leading embodied intelligence companies.
SO017 36Kr “星源智”完成新一轮融资 本轮融资将重点投入三大方向:下一代具身大脑与世界模型的核心技术研发、产品规模化量产落地、顶尖人才引进与团队建设。
SO018 36Kr “星源智”完成新一轮融资(移动版) 至今已累计融资10亿元人民币。
SO019 RobotToday Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology Xingyuanzhi Robot, a robotics company based in Beijing, has successfully secured 1 billion yuan in funding over the past 10 months.
SO020 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms Most startups in the cohort carry cash runways of only 18 to 24 months.
SO021 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge total investment reaching ¥93.5 billion across 322 financing deals
SO022 Feed The AI Robotics Funding Tracker (2026) Robotics funding is entering its physical AI era.
SO023 Goldman Sachs Research The AI Accelerant: Humanoid Robots the global market for humanoid robots may reach a market size of at least US$6bn in 10-15 years
SO024 CNBC Investors bet humanoid robots will transform industry and homes over the next decade China also “dominates the production and deployment of humanoid robots.”
SO025 Edge AI and Vision Alliance / IDTechEx Humanoid Robots 2026-2036: Technologies, Markets, and Opportunities IDTechEx forecasts the humanoid robot market will reach ~US$29.5 billion by 2036.
SO026 Axis Intelligence Humanoid Robots Deployment 2026: Case Studies The transition from prototype to production deployment represents the defining moment for humanoid robotics.
SO027 Faxiangongchang China Humanoid Robot 2026 By end of 2025, global humanoid robot shipments exceeded 17,000 units, with Chinese manufacturers contributing approximately 14,400 units.
SO028 LAVX News Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker The revenue numbers are still tiny relative to the funding.
SO029 News Globe Now Embodied AI Startup Xingyuan Zhi Raises 1 Billion Yuan The T5 platform, based on Nvidia computing chips, began volume production in 2025, shipping several hundred units with revenues exceeding 10 million yuan.
SO030 AGIBOT AGIBOT Innovation (Shanghai) Technology Co., Ltd. AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SM001 Xingyuanzhi Robotics 星源智 构建物理世界的通用具身大脑。
SM002 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 北京亦庄机器人、智元机器人、中力股份等为战略合作伙伴与生态。
SM003 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SM004 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time Xingyuanzhi’s hardware-software integrated solutions cover more than 70% of the leading embodied intelligence companies.
SM005 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms Most startups in the cohort carry cash runways of only 18 to 24 months.
SM006 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge total investment reaching ¥93.5 billion across 322 financing deals
SM007 Feed The AI Robotics Funding Tracker (2026) Robotics funding is entering its physical AI era.
SM008 Goldman Sachs Research The AI Accelerant: Humanoid Robots the global market for humanoid robots may reach a market size of at least US$6bn in 10-15 years
SM009 CNBC Investors bet humanoid robots will transform industry and homes over the next decade the size of the market today is really small, it’s 2 to 3 billion [dollars], but we see it going up to $200 billion in 2035.
SM010 Edge AI and Vision Alliance / IDTechEx Humanoid Robots 2026-2036: Technologies, Markets, and Opportunities IDTechEx forecasts the humanoid robot market will reach ~US$29.5 billion by 2036.
SM011 Axis Intelligence Humanoid Robots Deployment 2026: Case Studies The transition from prototype to production deployment represents the defining moment for humanoid robotics.
SM012 Faxiangongchang China Humanoid Robot 2026 China 2025 humanoid robot shipments: ~14,400 units, 84.7% global share; 2026E: ~62,500 units.
SM013 International Federation of Robotics World Robotics 2025 China is by far the world’s largest market in 2024, representing 54% of global deployments.
SM014 The Robot Report IFR: industrial robot deployments have doubled in 10 years China’s operational robot stock exceeded the 2 million mark in 2024.
SM015 RobotToday China’s Humanoid Robot and Embodied Intelligence Standard System (HEIS 2026) HEIS 2026 is structured around six primary categories, covering 22 secondary domains and more than 80 granular sub-standards.
SM016 RobotToday China’s 15th Five-Year Plan (2026–2030): Embodied Intelligence as National Industrial Strategy Embodied intelligence now commands its own dedicated inset box among the plan’s ten priority future-industry tracks.
SM017 State Council Information Office China’s first national standard system for humanoid robotics poised to spur industry development It was developed collaboratively by over 120 research institutions, enterprises and industry users.
SM018 Beijing Municipal Government 2026 Humanoid Robot Half-Marathon The winning team will be rewarded with industrial order(s) exceeding 1 million yuan.
SM019 Deloitte Insights AI for industrial robotics, humanoid robots, and drones Deloitte estimates annual unit shipments to be in the range of 5,000 to 7,000 in 2025, which may increase to 15,000 in 2026.
SM020 Deloitte Transformation of Warehouses and Manufacturing: How Humanoid Robots Will Change Automotive Supply Chains Humanoids are primarily used for simple, repetitive tasks such as handling metal sheets or other components.
SM021 Deloitte AI goes physical: Navigating the convergence of AI and robotics The reality gap, trust, safety, regulation, data complexity, and human acceptance remain key challenges.
SM022 Deloitte Australia Deloitte unveils physical AI solutions built with NVIDIA Omniverse simulation-led testing and secure edge AI can reduce downtime and support faster decision-making.
SM023 36Kr “星源智”完成新一轮融资 本轮融资将重点投入三大方向:下一代具身大脑与世界模型的核心技术研发、产品规模化量产落地、顶尖人才引进与团队建设。
SM024 AGIBOT AGIBOT Innovation (Shanghai) Technology Co., Ltd. AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SM025 Xingyuanzhi Robotics 关于我们 - 星源智 发布具身大脑算力平台T5,搭载于智元全新精灵G2系列。
SP001 Xingyuanzhi Robotics 星源智 构建物理世界的通用具身大脑。
SP002 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 与北京亦庄机器人、智元机器人、中力股份等形成战略合作生态。
SP003 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SP004 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time hardware-software integrated solutions cover more than 70% of the leading embodied intelligence companies.
SP005 Lavx Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker Robot makers have a strong incentive to own the brain themselves.
SP006 Spirit AI Spirit AI News Spirit AI Secures Nearly ¥2 Billion in Funding.
SP007 RobotToday Spirit AI Moz humanoid robots achieved insertion success rates consistently exceeding 99% on CATL battery production lines.
SP008 X Square Robot X Square Robot Research WALL-SS ... achieving calibrated task outcomes and consistent policy rankings across 600+ matched sim-to-real closed-loop rollouts.
SP009 RobotToday GigaAI an initial batch of around 100 SeeLight S1 robots into real homes.
SP010 Astribot Astribot S1 首创面向AI的软硬件一体化系统架构。
SP011 RobotToday Astribot The Astribot S1 ... supports autonomous navigation with real-time mapping and obstacle avoidance.
SP012 AGIBOT AGIBOT AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SP013 NVIDIA Developer NVIDIA Isaac This open robotics development platform consists of simulation and robot learning frameworks... for AMRs, robot arms, manipulators, and humanoids.
SP014 NVIDIA Jetson Orin Jetson AGX Orin ... 275 TOPS.
SP015 NVIDIA Docs Jetson Orin Series Software Features Linux supports these software features ... complete package to bring up Linux on Jetson AGX Orin.
SP016 EP Automation EP Automation - Innovative warehouse automation Start automation the easy way ... with no fixed infrastructure and no complex setup.
SP017 EP Automation Technology – EP Automation DAS integrates with existing ERP/WMS ... Start small and expand.
SP018 EP Automation XP15 – EP Automation 995€/mo ... 25.000€ + (1.500€ set up fee).
SP019 EP Equipment XP15 Wins Innovative Robotics Award ensuring a return on investment in less than a year.
SP020 Unitree Robotics Advanced Quadruped Inspection Solutions new power intelligent inspection solution ... dangerous, urgent and repetitive tasks.
SP021 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge 22 unicorns emerged in just six months.
SP022 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms capital is abundant, but proof of durable commercialization is much thinner.
SP023 Deloitte Insights AI for industrial robotics, humanoid robots, and drones AI humanoid robots are likely to be deployed in some industrial settings first.
SP024 RobotToday China’s 15th Five-Year Plan (2026–2030): Embodied Intelligence as National Industrial Strategy Embodied intelligence now commands its own dedicated inset box among the plan’s ten priority future-industry tracks.
SP025 World Robot Conference 世界机器人大会 2025世界机器人大会 ... 让机器人更智慧,让具身体更智能。
SI001 Xingyuanzhi Robotics 星源智 构建物理世界的通用具身大脑。
SI002 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 中力股份、智元机器人、北京亦庄机器人等形成合作生态。
SI003 Xingyuanzhi Robotics 关于我们 - 星源智 发布具身大脑算力平台T5,搭载于智元全新精灵G2系列。
SI004 Xingyuanzhi Robotics 产品中心 - 星源智 具身大脑算力平台 N5。
SI005 Xingyuanzhi Robotics Product - XYZ Compact Jetson Thor computing platform for on-device deployment.
SI006 Xingyuanzhi Robotics 合作机构 - 星源智 合作机构。
SI007 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SI008 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time raised $1 billion in just 10 months
SI009 36Kr “星源智”完成新一轮融资 重点投入三大方向:下一代具身大脑与世界模型的核心技术研发、产品规模化量产落地、顶尖人才引进与团队建设。
SI010 Lavx Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker The company says it shipped hundreds of T5 units in 2025 and booked over ¥10 million in revenue with a team of about 50 people, more than 90% of them in R&D.
SI011 News Globe Now Embodied AI Startup Xingyuan Zhi Raises 1 Billion Yuan shipping several hundred units with revenues exceeding 10 million yuan ... nearly 10,000 units in 2026.
SI012 企查查 北京星源智机器人科技有限公司 - 企查查 北京市海淀区海淀大街3号鼎好大厦A座2层203-1
SI013 PitchBook Xingyuanzhi 2026 Company Profile: Valuation, Funding & Investors | PitchBook Valuation, Funding & Investors
SI014 EP Equipment EP Equipment - World leading material handling 90% of Components
SI015 EP Automation EP Automation - Innovative warehouse automation 4-week rollout ... 1-week delivery
SI016 EP Automation Technology – EP Automation DAS integrates with existing ERP/WMS ... Start small and expand.
SI017 EP Automation XP15 – EP Automation 995€/mo ... 25.000€ + (1.500€ set up fee).
SI018 EP Equipment XP15 Wins Innovative Robotics Award return on investment in less than a year.
SI019 NVIDIA Jetson Orin Jetson AGX Orin ... 275 TOPS.
SI020 NVIDIA Docs Jetson Orin Series Software Features complete package to bring up Linux on Jetson AGX Orin
SI021 Deloitte Insights AI for industrial robotics, humanoid robots, and drones AI humanoid robots are likely to be deployed in some industrial settings first.
SI022 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge ¥93.5 billion across 322 financing deals
SI023 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms Most startups in the cohort carry cash runways of only 18 to 24 months.
SI024 AGIBOT AGIBOT AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SI025 Beijing Municipal Government 2026 Humanoid Robot Half-Marathon The winning team will be rewarded with industrial order(s) exceeding 1 million yuan.
SI026 NVIDIA Robotics and Edge AI Robotics and Edge AI
SI027 NVIDIA Isaac Robotics Robotics and Edge AI
SI028 Google Patents Google Patents Advanced Search Google Patents Advanced Search
SI029 USPTO Patent Public Search This online tool provides public access to search U.S. patents and published applications.
SI030 AGIBOT AGIBOT News AGIBOT Unveils Four New Products at WAIC...
SE001 Xingyuanzhi Robotics 星源智 构建物理世界的通用具身大脑。
SE002 Xingyuanzhi Robotics 关于我们 - 星源智 发布具身大脑算力平台T5,搭载于智元全新精灵G2系列。
SE003 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 场景涵盖货架拣选、巡检导览、康养服务、餐饮服务、智能装卸。
SE004 Xingyuanzhi Robotics 产品中心 - 星源智 具身大脑算力平台 N5。
SE005 Xingyuanzhi Robotics Product - XYZ Compact Jetson Thor computing platform for on-device deployment.
SE006 XYZ Embodied AI XYZ Embodied AI Co., Ltd. Building a highly generalizable, cross-embodiment general-purpose brain
SE007 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology the T5 computing platform ... an AI brain platform for robots
SE008 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time hardware-software integrated solutions cover more than 70% of the leading embodied intelligence companies.
SE009 36Kr “星源智”完成新一轮融资 下一代具身大脑与世界模型的核心技术研发
SE010 Lavx Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker a high-performance domain controller paired with general-purpose embodied AI models that run inference on edge hardware in real time
SE011 NVIDIA Jetson Orin Jetson AGX Orin ... 275 TOPS.
SE012 NVIDIA Docs Jetson Orin Series Software Features Linux supports these software features ... complete package to bring up Linux on Jetson AGX Orin
SE013 GitHub / NVIDIA-ISAAC-ROS isaac_ros_common Essential utilities, packages, scripts, and testing infrastructure for Isaac ROS packages.
SE014 GitHub / NVIDIA-ISAAC-ROS isaac_ros_visual_slam high-performance ... ROS 2 package for VSLAM
SE015 GitHub / NVIDIA-ISAAC-ROS isaac_ros_nvblox GPU-accelerated 3D reconstruction library
SE016 GitHub / NVIDIA-ISAAC-ROS isaac_ros_pose_estimation pose estimation and tracking
SE017 NVIDIA Omniverse Docs Isaac Sim Documentation Import robots and scenes from URDF, MJCF, Onshape CAD, or USD.
SE018 NVIDIA Developer Isaac GR00T open reference platform for general-purpose humanoid robots
SE019 Google Patents Patent search for 北京星源智机器人科技有限公司 404. That’s an error.
SE020 企查查 北京星源智机器人科技有限公司 - 企查查 北京市海淀区海淀大街3号鼎好大厦A座2层203-1
SE021 AGIBOT AGIBOT AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SE022 Beijing Municipal Government 2026 Humanoid Robot Half-Marathon industrial order(s) exceeding 1 million yuan
SE023 World Robot Conference 世界机器人大会 让机器人更智慧,让具身体更智能。
SE024 X Square Robot X Square Robot Research 600+ matched sim-to-real closed-loop rollouts.
SE025 Baidu Baidu patent search page 很抱歉,您要访问的页面不存在!
SU001 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 合作机构包括智元机器人、北京亦庄机器人、中力股份等。
SU002 Xingyuanzhi Robotics 合作机构 - 星源智 合作机构。
SU003 Xingyuanzhi Robotics 星源智 构建物理世界的通用具身大脑。
SU004 Xingyuanzhi Robotics 关于我们 - 星源智 发布具身大脑算力平台T5,搭载于智元全新精灵G2系列。
SU005 Lavx Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker Customers and partners named so far include AgiBot ... and Beijing Yizhuang Robot.
SU006 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SU007 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time solutions cover more than 70% of the leading embodied intelligence companies.
SU008 News Globe Now Embodied AI Startup Xingyuan Zhi Raises 1 Billion Yuan major strategic clients ... Agibot and EP Equipment.
SU009 AGIBOT AGIBOT Unveils Four New Products at WAIC 2026, Showcasing Embodied AI in Real-World Operations more than 60 AGIBOT robots are operating across WAIC venues
SU010 PR Newswire AGIBOT Unveils New Generation of Embodied AI Robots and Models, Accelerating Real-World Deployment of Physical AI In March 2026, AGIBOT announced the rollout of its 10,000th robot.
SU011 AGIBOT AGIBOT AGIBOT Innovation (Shanghai) Technology Co., Ltd.
SU012 EP Automation Fiege Logistics – EP Automation completed in a single-day, one-shot installation.
SU013 EP Automation Active Ants – EP Automation already seeing clear improvements in efficiency and productivity.
SU014 EP Automation aalbers|wico – EP Automation eliminates repetitive walking and idle travel time
SU015 EP Equipment XP15 AMR: What Is It, Use Cases, Benefits boost productivity by up to 300%
SU016 IFToMM WRC 2026 World Robot Conference 2026
SU017 IEEE Robotics and Automation Society 2026 World Robot Conference (WRC) 2026 World Robot Conference
SU018 ChinaPower / CSIS Is China Leading the Robotics Revolution? China is rapidly increasing its use of industrial robots.
SU019 Rest of World How China is using human labor to win the humanoid robot data race Chinese robotics companies are building large data-collection operations to train physical AI.
SU020 Beijing Municipal Government Visitor Guide to 2026 World Robot Conference The 2026 World Robot Conference opened today in Beijing E-Town.
SU021 Beijing Municipal Government 2026 Humanoid Robot Half-Marathon industrial order(s) exceeding 1 million yuan
SU022 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms a reality check looms
SU023 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge 22 unicorns emerged in just six months.
SU024 企查查 北京星源智机器人科技有限公司 - 企查查 北京市海淀区海淀大街3号鼎好大厦A座2层203-1
SU025 AGIBOT AGIBOT News AGIBOT Ranks No.1 ... AGIBOT Unveils Four New Products at WAIC 2026
SR001 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 合作机构包括智元机器人、北京亦庄机器人、中力股份等。
SR002 Xingyuanzhi Robotics 关于我们 - 星源智 发布具身大脑算力平台T5。
SR003 Xingyuanzhi Robotics 产品 - 星源智 面向具身智能的端侧算力平台。
SR004 企查查 北京星源智机器人科技有限公司 - 企查查 成立日期 2025-08-01。
SR005 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SR006 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time solutions cover more than 70% of the leading embodied intelligence companies.
SR007 Lavx Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker RoboBrain Pro built on NVIDIA’s Orin X chip.
SR008 News Globe Now Embodied AI Startup Xingyuan Zhi Raises 1 Billion Yuan major strategic clients include Agibot and EP Equipment.
SR009 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms a reality check looms
SR010 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge 22 unicorns emerged in just six months.
SR011 International Federation of Robotics China Makes AI-powered Robots Core of National Strategy actual capabilities in real-world production scenarios are currently limited to demonstrators or pilot projects.
SR012 The China Strategy China’s New Five-Year Plan Prioritizes Robotics—The World Should Pay Attention This is less an industrial policy for robots than an industrial policy through robots.
SR013 U.S. International Trade Administration China - U.S. Export Controls In particular, exporters should be aware of license requirements introduced in October 2022 and expanded or clarified in October 2023, April 2024, and December 2024 regarding certain advanced computing integrated circuits.
SR014 Bureau of Industry and Security Guidance Regarding Enforcement of License Requirements for Advanced Computing Items a license is required to export advanced computing items to entities headquartered in Country Group D:5.
SR015 Hill Dickinson Humanoid robots and the law: preparing for a new era of risk Who is responsible if a robot causes harm - the manufacturer, the operator, or the software provider?
SR016 European Commission Regulatory framework proposal on artificial intelligence high-risk AI systems will be subject to strict obligations before they can be put on the market.
SR017 NIST AI Risk Management Framework developed a framework to better manage risks to individuals, organizations, and society associated with artificial intelligence.
SR018 CISA Artificial Intelligence Guidelines for Secure AI System Development.
SR019 OSHA Robotics many robot accidents occur during non-routine operating conditions.
SR020 euRobotics euRobotics European culture can bring to the ethics and application of robotics.
SR021 Deloitte AI Goes Physical: Navigating the Convergence of AI and Robotics The “reality gap”: robots trained in simulations can still perform differently in the real world.
SR022 Deloitte Australia Deloitte and NVIDIA expand collaboration on physical AI solutions simulation-led testing and secure edge AI can reduce downtime and support faster decision-making.
SR023 Rest of World How China is using human labor to win the humanoid robot data race Chinese robotics companies are building large data-collection operations to train physical AI.
SR024 AGIBOT AGIBOT Unveils Four New Products at WAIC 2026, Showcasing Embodied AI in Real-World Operations more than 60 AGIBOT robots are operating across WAIC venues
SR025 PR Newswire AGIBOT Unveils New Generation of Embodied AI Robots and Models AGIBOT announced the rollout of its 10,000th robot.
SR026 Beijing Municipal Government Visitor Guide to 2026 World Robot Conference The 2026 World Robot Conference opened today in Beijing E-Town.
SR027 Beijing Municipal Government 2026 Humanoid Robot Half-Marathon industrial orders exceeding 1 million yuan
SR028 IEEE Robotics and Automation Society 2026 World Robot Conference (WRC) 2026 World Robot Conference
SR029 ChinaPower / CSIS Is China Leading the Robotics Revolution? China is rapidly increasing its use of industrial robots.
SR030 Xingyuanzhi Robotics 星源智 构建物理世界的通用具身大脑。
SV001 Xingyuanzhi Robotics 关于我们 / 公司概况 - 星源智 合作机构包括智元机器人、北京亦庄机器人、中力股份等。
SV002 Xingyuanzhi Robotics 关于我们 - 星源智 发布具身大脑算力平台T5。
SV003 企查查 北京星源智机器人科技有限公司 - 企查查 成立日期 2025-08-01。
SV004 Pandaily Xingyuanzhi Robot Raises ¥1 Billion in 10 Months for Embodied AI Brain Technology It doesn’t make robot hardware but only focuses on the “brain” of robots.
SV005 36Kr Zhipu AI Creates a New Unicorn: Raises $1 Billion in Just 10 Months for the Second Time new unicorn
SV006 Lavx Xingyuanzhi Bets ¥1 Billion That Robots Need a Brain Vendor, Not Another Body Maker RoboBrain Pro built on NVIDIA’s Orin X chip.
SV007 News Globe Now Embodied AI Startup Xingyuan Zhi Raises 1 Billion Yuan embodied AI startup ... raises 1 billion yuan
SV008 Feed the AI Robotics funding tracker 2026 The biggest robotics startup funding rounds of 2026 so far show investors moving beyond software-only AI and into machines that can see, move, manipulate, assemble, deliver, sort, inspect, and operate in the real world.
SV009 EmbodiedGlobal China Embodied AI Funding Hits ¥93.5B in H1 2026, 22 Unicorns Emerge 22 unicorns emerged in just six months.
SV010 China Biz Insider China Embodied AI Unicorns Surge: 15 Startups Hit $1.4B Valuation in H1 2026, But a Reality Check Looms 15 startups hit $1.4B valuation in H1 2026
SV011 International Federation of Robotics China Makes AI-powered Robots Core of National Strategy actual capabilities in real-world production scenarios are currently limited to demonstrators or pilot projects.
SV012 The China Strategy China’s New Five-Year Plan Prioritizes Robotics—The World Should Pay Attention This is less an industrial policy for robots than an industrial policy through robots.
SV013 Precedence Research Humanoid Robot Market Size The global humanoid robot market size is calculated at USD 1.84 billion in 2025 and is predicted to increase from USD 2.16 billion in 2026 to approximately USD 8.78 billion by 2035.
SV014 Global Market Insights Humanoid Robot Market Size The market is expected to grow from USD 10.9 billion in 2026 to USD 54.2 billion in 2031 & USD 192.7 billion in 2035.
SV015 CompaniesMarketCap UBTECH Robotics market cap As of August 2026 UBTECH Robotics has a market cap of $5.36 Billion USD.
SV016 CompaniesMarketCap Symbotic market cap As of August 2026 Symbotic has a market cap of $24.13 Billion USD.
SV017 CompaniesMarketCap Serve Robotics market cap As of August 2026 Serve Robotics has a market cap of $0.43 Billion USD.
SV018 CompaniesMarketCap NVIDIA market cap As of August 2026 NVIDIA has a market cap of $5.253 Trillion USD.
SV019 BAAI BAAI智源研究院 BAAI智源研究院
SV020 Spirit AI Spirit AI News Spirit AI Secures Nearly ¥2 Billion in Funding.
SV021 RobotToday Spirit AI Moz humanoid robots achieved insertion success rates consistently exceeding 99% on CATL battery production lines.
SV022 PR Newswire AGIBOT Unveils New Generation of Embodied AI Robots and Models AGIBOT announced the rollout of its 10,000th robot.
SV023 AGIBOT AGIBOT Unveils Four New Products at WAIC 2026, Showcasing Embodied AI in Real-World Operations more than 60 AGIBOT robots are operating across WAIC venues
SV024 Beijing Municipal Government Visitor Guide to 2026 World Robot Conference The 2026 World Robot Conference opened today in Beijing E-Town.
SV025 ChinaPower / CSIS Is China Leading the Robotics Revolution? China is rapidly increasing its use of industrial robots.
SV026 U.S. International Trade Administration China - U.S. Export Controls exporters should be aware of license requirements ... regarding certain advanced computing integrated circuits
SV027 Bureau of Industry and Security Guidance Regarding Enforcement of License Requirements for Advanced Computing Items a license is required to export advanced computing items to entities headquartered in Country Group D:5.
SV028 Deloitte AI Goes Physical: Navigating the Convergence of AI and Robotics The field is also moving from small pilots to large-scale production.
SV029 Deloitte Australia Deloitte and NVIDIA expand collaboration on physical AI solutions simulation-led testing and secure edge AI can reduce downtime and support faster decision-making.
SV030 Xingyuanzhi Robotics 产品 - 星源智 面向具身智能的端侧算力平台。
SV031 CompaniesMarketCap Tesla market cap As of August 2026 Tesla has a market cap of $1.377 Trillion USD.