Gravis Robotics
面向现有重型设备车队的改装式自主化
Gravis Robotics 在痛点市场里拿出了可信的产品和伙伴证据,但新晋独角兽估值已经计入大量执行成果,而公开收入和毛利证据尚未验证。
封面要素
公司概况
Gravis Robotics 是一家从 ETH Zurich 孵化出来、总部位于 Zurich 的自主化初创公司,为挖掘机和其他重型设备加装感知、控制和操作员辅助系统。公司把 Gravis Rack 自主化套件与 Slate 界面组合使用,走混合车队策略,瞄准建筑、采石、采矿和基础设施流程——这些场景里的劳动力短缺、安全压力和生产率约束都很尖锐。
- 创始人
- Ryan Luke Johns, Dominic Jud, Marco Hutter
- 创立地点
- Zurich, Switzerland
- 总部
- Zurich, Switzerland
- 产品
- 用 Gravis Rack 自主化套件、Slate 操作员界面、传感器、车载计算和机器学习控制软件,改装现有重型设备。
- 客户
- 大型承包商、采石和材料运营商、基础设施建设方、采矿和工业运营商,以及改造混合重型设备车队的渠道伙伴。
- 商业模式
- 混合部署加软件模式,围绕改装安装、有监督自主化项目、支持服务,以及未来毛利率更高的车队编排或工作流软件搭建。
- 阶段
- Series A
- 融资情况
- 2025 年融资 $23M,2026 年 8 月以约 $1B 估值完成 $200M Series A 轮;公开证据显示累计融资至少 $223M。
执行摘要
主要优势
- 改装逻辑很强:先切入存量车队,而不是等待 OEM 替换周期。
- Holcim、Taylor Woodrow、Techint、Flannery 以及 OEM 演示合作,提供了可见的旗舰客户证据。
- 对一家很年轻的建筑自动驾驶公司来说,SoftBank 的资本支持异常强。
主要风险
- 在公开商业指标可见之前,估值已经反映了大量未来执行。
- 更低接触度自动化的安全、责任和保险框架,公开披露仍不足。
- OEM 既有厂商和相邻自动化供应商,可能随着时间压缩改装切口。
未决问题
- 需要收入、毛利率和部署队列指标,才能搭建真实估值模型。
- 客户集中度、续约和扩张数据未公开。
- 公开证据仍无法看清合同责任分配和保险方态度。
目录
01公司概况
1.1 身份、起源和商业模式
Gravis Robotics 现在已经清楚地从隐身实验室项目变成一家创立于 Zurich 的实体 AI 公司。官方页面、独立融资报道和合作伙伴公告都在讲同一个故事:公司 2022 年从 ETH Zurich 孵化出来,要把自主作业能力改装到现有重型设备上,而不是要求承包商更换整支车队。这个差别很关键,因为它贴合建筑公司真实的购机方式——依赖既有车队、区域服务关系和多品牌混合运营习惯。Gravis Rack 与 Slate 界面组合起来,使公司更像是给传统钢铁设备加一层自主化能力,而不是新的整机 OEM。商业化因此更可信,但 Gravis 也必须同时解决现场部署、控制集成、支持和信任,而不只是软件模型表现。按实际尽调口径,本章确立后续章节必须沿用的基本事实:Gravis 卖的是现有车队的升级路径,而不是从零设计的新机器平台。[CO001, CO002, CO003, CO016, CO017, CO018]
| 指标 | 数值 / 状态 | 证据日期 | 置信度 | 备注 |
|---|---|---|---|---|
| 成立时间 | 2022 | 2026-08-17 | 高 | ETH Zurich 孵化公司 |
| 总部 | 瑞士 Zurich | 2026-08-18 | 高 | 官方关于页面 |
| 其他办公室 | Austin 和 Oxford | 2026-08-18 | 中 | 官方关于页面 |
| 最新轮次 | $200M Series A 轮 | 2026-08-17 | 高 | SoftBank 唯一投资方 |
| 投后估值 | $1B | 2026-08-17 | 高 | 多家媒体交叉印证 |
| 公开披露累计融资 | ~$223M | 2026-08-18 | 中 | 仅基于已披露轮次 |
| 员工数 | 约 75 人 | 2026-08-17 | 中 | Inc. 报道 |
| 核心产品 | Gravis Rack + Slate | 2026-08-18 | 高 | 改装式自主化栈 |
收入和现金等未获支持的私营公司指标未被列入,而不是被猜测填充。
[CO001, CO002, CO003, CO011, CO014, CO015]公开里程碑显示,Gravis 从分拆成立到全球扩张、再到独角兽定价的 Series A,推进很快。
里程碑时间依据公开发布时间;内部合同签署日期可能不同。
[CO001, CO009, CO026, CO031, CO011, CO014]Gravis 将 ETH 研究、改装式自主化、混合车队兼容性、合作伙伴验证和 SoftBank 资本串成一条经营逻辑。
[CO001, CO017, CO019, CO027, CO015, CO035]1.2 领导班底和组织信号
公开可见的领导层集中但可信。Ryan Luke Johns 和 Dominic Jud 仍是最常出现在商业扩张叙事中的两位运营型创始人,Marco Hutter 则从 ETH Zurich 脉络中提供学术背书和董事会连续性。公开材料没有展示宽阔的高管梯队,也没有完整披露董事会,因此关键人物依赖仍然明显,尤其是 Gravis 正站在深度机器人研发和高强度运营部署的交叉点上。同时,招聘和关于页面显示,公司正在感知、自主化、平台、硬件、界面和现场应用等岗位扩充人手。投资人会期待一家从技术新奇走向可复制部署运营的公司出现这种招聘图谱。这是积极的成熟度信号,但不能替代更深入的治理披露。因此,核心领导力足以支撑运营故事,但仍然薄到需要投资人索取更完整的管理层地图和董事会权利包。[CO004, CO005, CO006, CO007, CO008, CO036]
| 人员 | 角色 | 公开背景信号 | 重要性 | 依赖程度 |
|---|---|---|---|---|
| Ryan Luke Johns | 联合创始人兼 CEO | 建筑师和机器人专家;公开商业声音 | 负责产品市场和融资叙事 | 高 |
| Dominic Jud | 联合创始人兼 CTO | 自主控制专家 | 负责技术可信度和系统行为 | 高 |
| Marco Hutter | 联合创始人兼董事会成员 | ETH Zurich 机器人学教授 | 提供机构研究连续性 | 中高 |
| Kyeni Mbiti | 工业设计负责人(被引用) | 公开页面上的硬件设计信号 | 暗示内部产品化深度 | 中 |
| Gabriel Waibel、Adam Abed Abud 与 Filippo Spinelli | 感知 / 平台 / 自主化岗位 | 可见的招聘和工程外露面 | 显示技术栈正在拓宽 | 中 |
这是基于公开表面的领导层表,不是完整高管名册或董事会名单。
[CO004, CO005, CO006, CO008, CO036]1.3 融资路径、投资方基础和验证里程碑
Gravis 的资本路径在不到一年里大幅改写。2025 年 11 月那轮融资先给出早期验证组合——IQ Capital、Zacua Ventures、Holcim 以及其他连接建筑和工业网络的支持方;2026 年 8 月 SoftBank 轮则把公司直接推到独角兽区间,投后估值约 $1B。因此,公开披露资本至少约 $223M;对一家仍未公开收入和利润率的公司来说,这个金额很大。故事最重要的不只是融资规模,而是规模出现前后的顺序。SoftBank 进入之前,Gravis 已经签下公开合作伙伴,向国际扩张,并在混合车队和真实工地上展示自主化能力。也正因为这条时间线,这轮融资更像是扩张押注,而不只是科学项目押注。同时,下一批里程碑将面对远高于 2025 年那轮的估值门槛。[CO009, CO010, CO011, CO012, CO013, CO014]
| 日期 | 事件 | 金额 / 价值 | 投资方 / 利益相关方 | 含义 |
|---|---|---|---|---|
| 2025-11 | 扩张融资 | $23M | IQ Capital、Zacua Ventures、Pear VC、Imad、Sunna Ventures、Armada Investment 与 Holcim | 为英国、美国和欧盟扩张提供资金 |
| 2025-11 | Holcim 战略投资 | 纳入本轮融资 | Holcim MAQER Ventures | 增加客户与采石场渠道价值 |
| 2026-08-17 | Series A 轮 | $200M | SoftBank | 建筑机器人领域最大 Series A 轮 |
| 2026-08-17 | 投后估值 | $1B | 由本轮融资报道推算 | 让 Gravis 跻身独角兽 |
| 2026-08-18 | 公开累计融资 | ~$223M | 仅含已披露融资 | 拉长现金跑道,但也抬高预期 |
表格仅根据公开披露重建融资事件;未披露债务、老股交易或补助可能不在清单内。
[CO009, CO010, CO011, CO012, CO014, CO015]公开记录显示融资和部署信号充足,但经审计的业务披露很少。
国家数量取自 2025-2026 年运营公告,可能比公开页面更新得更快。
[CO011, CO014, CO007, CO025, CO026, CO024]1.4 部署足迹、公开证明和仍未解决的风险
最强的公开运营证明在于覆盖面,而不是经过审计的经济性。Gravis 称其系统已在四大洲、七个国家运行,并公开提到 Holcim、Taylor Woodrow、HD Hyundai、Flannery、Hitachi、Techint 等合作伙伴。英国 CAM Pathfinder 奖项、Manchester Airport 试验路径、Argentina 管道项目,以及 CONEXPO 2026 扩张信息,都支持一个判断:公司已经越过单一演示工地阶段。即便如此,概览仍然留下未解决的承保缺口。独立报道反复指出,Gravis 必须证明改装式自主化能走出试点规模,并压过 OEM 主导的自主化栈。公司尚未发布收入、现金余额、确切客户数、毛利率或完整董事会构成。对后续尽调章节而言,这些盲点和融资头条同样重要。因此,概览能支撑强身份和里程碑叙事,但单靠它还无法证明商业耐久性。[CO025, CO026, CO027, CO028, CO029, CO030]
| 时期 | 里程碑 | 证据 | 重要性 | 待解问题 |
|---|---|---|---|---|
| 2022 | 公司作为 ETH Zurich 衍生项目成立 | 官方 Series A 页面与独立新闻报道 | 建立学术机器人出身叙事 | 公开亮相前有哪些早期试点? |
| 2025-04 | Taylor Woodrow 自动驾驶挖掘机试验 | VINCI / Highways 报道 | 证明其在英国土木工程现场测试 | 离开展示项目后,可复制性有多强? |
| 2025-11 | $23M 融资和标志性交易 | 官方新闻稿与 2025 年报道 | 释放早期商业牵引力 | 这些交易对应多少收入? |
| 2026-03 | 美国扩张与 CONEXPO 演示 | Robotics & Automation News 与 Hitachi | 显示 OEM 与渠道扩张 | 演示带来的兴趣多久能转化? |
| 2026-08 | SoftBank 领投 $200M Series A 轮,估值 $1B | 官方与多家独立媒体 | 提供规模化资金和外部验证 | SoftBank 接下来期待哪些里程碑? |
日期是公开里程碑,而非完整内部运营史。
[CO001, CO029, CO009, CO031, CO011, CO014]1.5 图表
02市场分析
2.1 市场边界和范围
Gravis 的市场应该比“建筑机器人”更窄,也比“建筑设备”更具体。公司并不是想把工地上的每一种工种都自动化。已披露的证明点集中在重复性土方、装车、开沟、整平、采石场物料处理和相关场地准备任务上,这些任务依赖长机器工时和稳定循环节奏。Gravis 进入市场的方式也是改装层,而不是 OEM 整机项目,这意味着相关预算不只是新机资本开支。它位于建筑设备、机器控制软件、远程信息处理和自主化的交叉点。这个定位很重要,因为最接近的替代品不只有其他自主化初创公司,还包括机器导引厂商、遥测平台、经销商驱动的 OEM 自主化项目,以及从不同起点解决同一买方痛点的高人工手动作业替代方案。因此,市场边界要先按工作流和买方问题界定,而不是按最宽泛的公开 TAM 类别界定。[CM001, CM002, CM003, CM004, CM005, CM006]
| 细分市场 / 类别 | 纳入支出 / 活动 | 排除支出 / 活动 | 买方 / 付款方 | 关联度 |
|---|---|---|---|---|
| 自主土方作业 | 大规模开挖、装车、整平、重复性场地准备 | 垂直建筑工种与收尾工程 | 总包商和土方分包商 | Gravis 核心切入点 |
| 工地改装式自主化 | 后装套件、传感器、计算、软件编排 | OEM 新设备制造 | 车队所有者、承包商和租赁渠道 | 核心商业模式 |
| 机器控制 / 数字工地工作流 | 方案到机器工作流、远程信息处理、进度跟踪 | 纯人工测量和纸质流程 | 项目控制与运营团队 | 邻近需求入口 |
| 租赁渠道带动的车队升级 | 借助租赁或租借设备,为混合车队加装自主能力 | 永久性车队更换周期 | 租赁公司和承包商 | Flannery 式模式验证的渠道机会 |
| OEM 集成式自主化 | Cat、Komatsu、Volvo 式机器集成自主化 | 仅后装改造方案 | 大型车队买家和 OEM 渠道 | 主要替代方案 |
| 采矿 / 运输自主化 | 非公路运输和自主物料运输 | 普通建筑工地土方流程 | 矿山运营商 | 相邻但不完全相同 |
市场边界聚焦重复性土方作业和改装式自主化,而不是所有机器人或所有建筑软件。
[CM001, CM002, CM003, CM004, CM005, CM006]市场口径从全部建筑设备,收窄到规模小得多、具备自主化改装条件的土方设备切口。
数值除建筑机器人外均以 USD billions 计;建筑机器人数字由 USD 442.49 million 折算为 0.44249 billion;SOM 层是示意性的有界切口,不是已披露市场估计。
[CM007, CM009, CM010, CM034, CM035]2.2 用多重视角测算市场
没有可获取来源给出自主土方改装的权威独立 TAM,因此单一头条数字会误导人。最佳公开证据提供的是一串相邻估算。放到最宽口径,建筑设备是一个以数千亿美元计的庞大全球市场。智能建筑设备、建筑机器人等更窄类别小得多,但仍足以容纳资金充足的新进入者。对 Gravis 重要的是,只要公司拿下混合车队中最高价值的重复性工作流,即便只占子集中的一小部分,也能做出有意义的业务。Fortune Business Insights、Global Market Insights、Future Market Insights 和 Mordor Intelligence 之间的差异,应被视为反对过度精确的警示,而不是否定投资逻辑的理由。正确结论是:存量设备基数巨大,自主化切口真实,确切支出池仍需要自下而上的尽调。[CM007, CM008, CM009, CM010, CM011, CM012]
| 发布方 | 年份 | 地区 | 数值 / 指标 | 增长 | 方法口径 | 可信度 | 局限 |
|---|---|---|---|---|---|---|---|
| Fortune Business Insights | 2026-2034 | 全球 | $183.27B 至 $310.24B 建筑设备市场 | 6.8% CAGR | 广义设备市场 | 中 | 范围超过 Gravis 切入点 |
| Global Market Insights(市场研究机构) | 2025-2035 | 全球 | $167B 至 $289.5B 建筑设备市场 | 6.1% CAGR | 广义设备市场 | 中 | 与 Fortune 口径基线不同 |
| Future Market Insights(市场研究机构) | 2025-2035 | 全球 | $24.4B 至 $81.5B 智能建筑设备 | 12.8% CAGR | 智能 / 联网设备子集 | 中 | 仍宽于改装式自主化 |
| Mordor Intelligence | 2025-2030 | 全球 | $442.49M 至 $909.53M 建筑机器人 | 15.5% CAGR | 机器人子集 | 中 | 纳入了与 Gravis 车队改装路径不同的机器人 |
| AGC / NCCER | 2025 | 美国 | 92% 承包商难以填补空缺岗位 | N/A | 用工需求压力 | 高 | 痛点指标,不是支出指标 |
| ABC | 2025 | 美国 | 行业需要近 440k 名新增工人 | N/A | 劳动力缺口估计 | 高 | 劳动力估计,不是自主化 TAM |
| CDC / BLS | 2024 或最新数据 | 美国 | 建筑业仍属高风险,坠落是主要死亡原因 | N/A | 安全成本压力 | 高 | 风险指标,不是支出指标 |
| U.S. Census | 2026 | 美国 | 持续庞大的建筑支出基数 | N/A | 宏观需求背景 | 高 | 支出不等于自主化可寻址支出 |
没有可访问公开来源单独切出自主土方改装支出;本章因此采用多重视角,而不是合成一个 TAM。
[CM007, CM008, CM009, CM010, CM011, CM012]公开市场估计差异很大,取决于视角是全部设备、智能设备还是建筑机器人。
不同发布方对品类定义不同,因此这里比较的是口径不完全相同但对决策有用的视角,而不是一条完全同口径的市场序列。
[CM007, CM008, CM009, CM010, CM013, CM031]2.3 买方分层和采用路径
公开证据指向总承包商、土方分包商、采石场运营商和设备租赁渠道,这是第一批可信买方群体。他们承担进度风险,拥有重复挖掘任务,也最受 Gravis 在现场部署中强调的操作员瓶颈影响。工业项目建设方和重型土木承包商尤其相关,因为大型制造、能源、基础设施和数据中心工地会产生可重复的场地准备工作,让自主化设备能长时间运行而不必频繁调整工作流。租赁公司在战略上很有意思,因为 Gravis 已经把改装模式接入 Flannery 的分销路径,让混合车队无需被迫永久更换设备也能获得自主化能力。开发商和业主多数时候不是直接买方,但会制造经济紧迫性:承包商如果能更快完成数据中心地坪或管道段,即使业主从不直接采购自主化,也可能赢得项目。因此,采用大概率先从承包商开始;一旦 ROI 得到证明,再扩展到更广的渠道和 OEM 伙伴关系。[CM015, CM016, CM017, CM018, CM019, CM020]
| 细分市场 | 买方 | 使用者 | 付款方 | 工作流 / 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|
| 总包商 | 运营或创新负责人 | 项目团队和现场主管 | 总包商 | 项目工期 / 利润率预算 | 压缩工期,并降低用工缺口风险 |
| 土方分包商 | 企业主 / 运营负责人 | 设备操作员和工长 | 分包商 | 土方生产率预算 | 自动化重复开挖 |
| 工业 / 制造业建设方 | 项目高管 | 现场运营 | 主承包商 | 大型场地准备合同包 | 大规模重复土方工程量 |
| 重型土木承包商 | 区域负责人 | 现场班组 | 承包商 | 基础设施项目控制 | 大型项目安全和正常运行时间 |
| 租赁公司 | 车队 / 创新负责人 | 租赁运营团队和客户 | 租赁公司或承包商 | 车队利用率预算 | 提高混合车队利用率 |
| 开发商 / 业主 | 间接经济买方 | N/A | 通过合同间接付款 | 工期和持有成本压力 | 更快交付住宅、数据中心和工厂 |
匹配度标签来自公开部署、合作伙伴证据和市场逻辑的综合判断,而非 Gravis 已披露的销售管线表。
[CM015, CM016, CM017, CM018, CM019, CM020]Gravis 的买方路径从总包商、土方分包商起步,延伸到业主的间接压力,之后可能进入租赁渠道。
契合度标签来自公开部署与市场逻辑的综合判断,不是 Gravis 披露的管线表。
[CM015, CM016, CM017, CM018, CM019, CM020]采用大概率从痛点识别走向试点审批、有人监督的部署、重复使用,最后进入车队编排。
这条流程是基于公开部署和管理层表述推导的概念性运营路径,不是已披露转化数据集。
[CM021, CM022, CM024, CM033, CM034]2.4 增长驱动因素、约束和数据缺口
最强的公开需求驱动因素很直接:劳动力短缺、安全压力,以及更快交付项目带来的经济溢价。AGC 的 2025 年调查和 ABC 的劳动力估算都显示,劳动力市场仍然结构性紧张。CDC、BLS 和 OSHA 材料进一步说明,建筑业依然高风险,因此即使不计生产率提升,自动化也有第二层逻辑。但同一组证据也说明,采用不会自动发生。建筑工地临时、动态且社会关系复杂;买方常常会先部署机器控制工具、增加操作员或调整排期,再承诺采用自主化。公开市场数据也仍然令人沮丧地不精确。我们知道宏观市场很大、痛点真实,但还没有一个干净的公开数据集能切出自主化预算、试点到生产的转化率,或让无操作员作业成为必需品的 ROI 门槛。后续财务和估值章节需要一直盯住这些问题。[CM021, CM022, CM023, CM024, CM025, CM026]
| 驱动因素 / 约束 | 方向 | 时间 | 影响 | 尽调问题 |
|---|---|---|---|---|
| 劳动力短缺 | 正向驱动 | 眼下 | 提高测试自动化的意愿 | 用工痛点转化为有预算试点的比例有多高? |
| 安全 / 死亡事故压力 | 正向驱动 | 眼下 | 支撑更安全工地的 ROI 主张 | Gravis 能否证明事故减少? |
| 数据中心和工厂建设 | 正向驱动 | 近期 | 工期压缩更有价值 | 这些垂直领域贡献多少需求? |
| 临时工地基础设施限制 | 约束 | 眼下 | 利好低基础设施部署 | 每个工地需要哪些搭建? |
| 信任和变革管理 | 约束 | 近期 | 放慢从有人监督转向无人操作的速度 | 需要哪些操作员培训? |
| 竞争性机器控制工具 | 约束 | 眼下 | 可能在没有全自主化的情况下满足部分买方 | 自主化与现有软件之间的 ROI 差距有多大? |
| 估计分散 / 数据缺口 | 约束 | 当前 | 让标题级 TAM 说法不可靠 | 哪种客户自下而上测算可以替代自上而下 TAM? |
| 车队编排上行空间 | 正向驱动 | 中期 | 创造超出单台机器的平台价值 | 有哪些多机协同证据? |
本风险 / 驱动图有意保持不完整,因为保险、劳动规则和采购预算的公开证据,弱于用工痛点和安全需求证据。
[CM021, CM022, CM023, CM024, CM025, CM026]03竞争格局
3.1 谁在和 Gravis 竞争,为什么
Gravis 的竞争集合比“做建筑机器人的初创公司名单”更宽。最接近的类比对象,是那些解决同一个买方问题的公司——在更少依赖稀缺操作员的情况下,让重型设备产出更安全、更稳定。实际可以分成三类。第一类是 Built Robotics 这样的初创公司类比对象,它同样讲建筑自动化故事,但更集中在太阳能工作流。第二类是 Caterpillar 等 OEM 巨头,它们可以把自主化直接嵌入基础机器,并带来经销商覆盖、服务能力和已建立的信任。第三类是 Hexagon、Pronto、Polymath 等相邻自主化或工作流玩家,它们通过软件、数据、运输或平台工具切入市场,而不是采用 Gravis 的承包商共研模式。只有在比较工作流匹配度、渠道控制和上线就绪度时,Gravis 的位置才说得通;把所有公司塞进一个宽泛机器人桶里并不能解释竞争。[CP001, CP002, CP003, CP004, CP005, CP006]
| 公司 | 主攻方向 | 设备 / 工作流 | 商业化路径 | 重要性 |
|---|---|---|---|---|
| Gravis Robotics | 面向重型建筑的改装式自主化 | 开挖 / 场地准备 | 与承包商共研 | 基准行 |
| Built Robotics | 机器人化光伏施工 | 打桩 / 光伏流程 | 产品化机器人设备 | 最接近的创业公司类比,但工作流更窄 |
| Caterpillar | 建筑领域 OEM 自主化 | 装载机、挖掘机、推土机、矿用卡车 | 机器 + 经销商渠道 | 最大在位者威胁 |
| Hexagon | 数字工作流和自主化邻近软件 | 工地数据 / 采矿 / 定位 | 企业软件和传感器 | 在机器行为上游竞争 |
| Pronto | 自主运输 | 非公路卡车 | 自主系统层 | 验证非公路自主化需求 |
| Polymath Robotics | 非公路车辆自主化中间件 | 多类非公路车辆 | 软件 / 系统层 | 邻近自主化平台竞争者 |
画像行强调公开可见的商业重点,并不声称完整覆盖每家公司的产品。
[CP001, CP002, CP003, CP004, CP005, CP006]Gravis 位于改装比重高、面向建筑的象限,OEM 与相邻自主化供应商则占据市场的其他角落。
x 值越高,表示 OEM 无关 / 软件层定位越强;y 值越高,表示越直接贴近主流建筑买家。
[CP001, CP002, CP003, CP004, CP005, CP006]3.2 功能广度、工作流匹配和渠道深度
Gravis 最强的产品层差异,是不绑定 OEM 的改装姿态。公开报道显示,公司把系统安装到现有挖掘机上,并在承包商活跃工地部署,而不是要求客户购买一整套全新机器生态。这不同于 Caterpillar 的模式,后者靠对机器和服务渠道的完整控制强化自主化层;也不同于 Hexagon 的模式,后者更看重工作流数据和工地系统,而不是直接改装机器。Built Robotics 展示了另一种战略极端:深挖一个可重复的建筑工作流,这能做出更标准化的产品,但会缩窄可触达用例。Pronto 和 Polymath 重要,是因为它们证明自主化能力可以跨越非公路车辆类别流动,即使没有 Gravis 这种精确工地焦点。因此,Gravis 竞争的核心不是原始功能数量,而是产品能否在真实承包商条件下干净嵌入重复性土方工作流。[CP007, CP008, CP009, CP010, CP011, CP012]
| 能力 | Gravis | Built | Caterpillar | Hexagon | Pronto | Polymath |
|---|---|---|---|---|---|---|
| 不限 OEM 的改装 | 高 | 中 | 低 | N/A | 中 | 高 |
| 开挖聚焦 | 高 | 低 | 中 | 低 | 低 | 中 |
| 经销商 / 服务渠道 | 低 | 低 | 高 | 中 | 低 | 低 |
| 工作流软件深度 | 中 | 中 | 中 | 高 | 中 | 中 |
| 重复性施工任务的公开现场验证 | 高 | 太阳能场景高 | 中 | 低 | 低 | 低 |
| 车队编排叙事 | 高 | 低 | 高 | 中 | 中 | 中 |
功能评分是基于公开材料做出的定性综合标签,而不是厂商提供的基准测试。
[CP007, CP008, CP009, CP010, CP011, CP012]Gravis 的优势在工作流契合和改装灵活性;在位厂商赢在服务渠道深度,相邻供应商赢在平台广度。
能力标签是基于公开材料的定性综合判断,不是已披露基准测试。
[CP007, CP008, CP009, CP010, CP011, CP012]3.3 包装方式、商业形态和购买摩擦
定价是竞争格局中最不透明的部分之一。Gravis 没有公布标价,说明当前商业动作仍围绕试点、工地和客户特定部署范围定制。这并不意味着业务弱,只意味着尽调还无法把 Gravis 与竞争对手放在一张干净的同口径价格表上比较。Built Robotics 在太阳能设备包装上似乎更产品化;Caterpillar 则能把自主化与机器销售和服务支持打包。Hexagon 可以靠软件和工作流 ROI 竞争,自主化平台玩家有时也能在不拥有车辆本体的情况下为系统层定价。对投资人而言,主要含义是:部署证明和买方信任目前比名义标价更有信息量。在商业条款可见之前,判断这个类别更应看安装摩擦、现场支持和重复使用证明,而不是只看标价。[CP013, CP014, CP015, CP016, CP017, CP029]
| 厂商 | 公开打包方式信号 | 公开定价透明度 | 渠道模式 | 含义 |
|---|---|---|---|---|
| Gravis | 定制部署 / 试点驱动 | 低 | 直接对接承包商 | 当下有灵活性,但买方看不清 |
| Built Robotics | 面向特定场景的机器人工作流产品 | 中低 | 解决方案直销 | 比 Gravis 更标准化 |
| Caterpillar | 整机与自主作业集成 | 中 | 经销商渠道 | 可把自主作业打包进机器生命周期 |
| Hexagon | 软件、传感器和工作流工具 | 中 | 企业销售 | 可能靠工作流 ROI 竞争,而不是靠替换机器 |
| Pronto / Polymath | 自主作业系统层 | 低 | 直销或伙伴主导 | 显示软件层打包的灵活性 |
此类产品公开定价仍然稀少,因此本表比较的是打包方式和商业透明度,而不是具体标价。
[CP013, CP014, CP015, CP016, CP017]Gravis 的竞争就绪度在现场验证上最强,在定价透明度和渠道深度上最弱。
这些 KPI 标签只概括公开证据;私有安装基数或续约数据可能显著改写图景。
[CP013, CP018, CP019, CP029, CP034, CP035]3.4 护城河耐久性和竞争风险
Gravis 正在形成的护城河不是某一项专利或硬件形态,而是现场数据、承包商集成和工作流专业知识的组合;部署扩大后,这些要素可以复利。这很有希望,但还不稳。OEM 仍是最大威胁,因为它们控制机器平台、质保边界和服务渠道;如果它们决定积极进入同样的重复性土方用例,Gravis 的改装优势可能收窄。同时,初创公司和软件层竞争者也显示,自主化栈本身可能随时间变得更可替换。Gravis 今天最好的防守,是证明承包商信任它、系统能以最小扰动贴合客户工作流,并且有监督作业产生的现场数据能让产品改进速度快过对手追赶。换句话说,Gravis 的护城河由学习速度驱动。它可以变得耐久,但前提是客户转化和部署重复先于巨头补上差距。这个类别仍足够年轻,执行速度极其关键。[CP018, CP019, CP020, CP021, CP022, CP023]
| 风险或护城河 | 方向 | 重要性 | 当前证据 | 尽调问题 |
|---|---|---|---|---|
| 现场数据护城河 | 优势 | 真实工地学习可能随时间复利 | Gravis 强调正在承包商项目中部署 | 标注数据集的专有性有多强? |
| OEM 渠道权力 | 风险 | OEM 掌握机器、质保和服务 | Cat 已经在推广自主作业 | 后装系统能否与 OEM 政策共存? |
| 工作流专精 | 优势 | 狭窄重复任务更容易先拿下 | 大规模土方开挖是最强公开切入点 | 开挖之后下一个工作流是什么? |
| 功能趋同 | 风险 | 软件层对手能在自主作业栈上追上来 | 非公路自主作业市场很碎片化 | Gravis 能多快推出改进? |
| 客户信任循环 | 优势 | 与承包商共创能让采用更有粘性 | 多条承包商引述已经公开 | 是否有复购或扩张数据? |
| 定价不透明 | 风险 | 很难在厂商之间比较 ROI | 没有清晰公开定价数据 | 收集报价和 SOW |
本登记表同时放入耐久因素和攻击面,因为 Gravis 的护城河仍在形成,还没有完全锁住。
[CP018, CP019, CP020, CP021, CP022, CP023]04财务情况
4.1 变现模式和收入形态
Gravis 的公开材料不像一家标准软件公司,因为产品并非纯靠代码交付。公司在客户工地改装重型设备,这意味着除任何经常性自主化软件收费外,收入里至少会有一些安装、校准和部署服务。时间拉长后,经济承诺大概率会转向软件、远程监控和多机编排,尤其是在 Gravis 成功从有监督单机部署走向协同车队之后。但现有证据指向混合模式:先有服务较重的收入把机器跑起来,随后如果客户让系统留在生产环境中,再形成经常性价值。这个组合在战略上有吸引力,因为它绑定真实工地 ROI;但也意味着公司今天很可能还没有软件式利润率。承保时的关键不是 Gravis 究竟是“软件”还是“硬件”,而是重复部署能多快把业务推向杠杆更高的经常性收入结构。[CI001, CI002, CI003, CI004, CI005, CI006]
| 收入流 | 公开支撑 | 当前可见度 | 存在原因 | 置信度 |
|---|---|---|---|---|
| 部署 / 安装费 | 公开资料描述了后装和现场设置 | 推断 | 安装和启动调试需要人工与硬件工作 | 中 |
| 经常性自主作业软件 | 公开资料突出实时智能和车队工具 | 推断 | 软件价值在安装后仍然持续 | 中 |
| 支持 / 监控 | 客户需要正常运行时间和现场支持 | 推断 | 保证机器持续、安全运行 | 中 |
| 工作流 / 编排工具 | B 轮叙事强调联网车队 | 推断 | 长期可能成为毛利更高的一层 | 中 |
| 扩张部署 | 伙伴计划和多工地测试已经公开 | 推断 | 重复部署能让收入复利 | 中 |
各项收入流都没有公开定价;本表区分的是可能的变现组成部分和已经披露的财务结果。
[CI001, CI002, CI003, CI004, CI005]| 问题 | 公开答案 | 可能方向 | 风险 | 下一步尽调 |
|---|---|---|---|---|
| 是否公布标价? | 否 | 定制报价 | 透明度低 | 收集报价 |
| 定价基准 | 未披露 | 机器 / 工地 / 支持组合 | ROI 难以比较 | 审阅客户 SOW |
| 订阅部分 | 未披露 | 长期可能会有 | 短期可能较小 | 询问收入拆分 |
| 试点折扣 | 未披露 | 当前可能不小 | 可能夸大长期经济性 | 比较试点与复购交易 |
| 客户回本框架 | 未披露 | 劳动力 + 工期 + 安全 ROI | 收益可能随工地类型而变 | 按工作流测算回本周期 |
由于公开披露尚未触及真实合同经济性,本表有意围绕未解答的变现问题展开。
[CI006, CI007, CI008, CI009, CI010]Gravis 可能先靠部署工作起量,随后逐步转向经常性软件和编排价值。
数值是方向性权重分数,不是已披露美元收入;图里展示的是结构,而不是报告口径收入组合。
[CI001, CI002, CI003, CI004, CI005, CI026]4.2 单位经济性和成本驱动因素
尽管数字不公开,单位经济性逻辑很直观。Gravis 把传感器、计算和控制系统安装到现有机器上,这意味着硬件和人工会进入销货成本,而纯 SaaS 公司不会这样。现场运营和客户支持也很重要,因为公司的公开证明仍由部署驱动,并且处在有监督阶段。这是短期负担。长期上行在于,重复挖掘工作流正是那类可以通过反复安装剧本、更好软件和更低监督力度逐步改善利润率的运营环境。如果 Gravis 能标准化更多安装流程、降低监督负担,并复制相似工地,毛利率应该朝正确方向移动。但如果每个项目都仍是定制化现场集成,业务会比估值叙事暗示的更偏服务。[CI011, CI012, CI013, CI014, CI015, CI028]
| 驱动因素 | 方向 | 重要性 | 公开证据 | 含义 |
|---|---|---|---|---|
| 传感器 + 计算硬件 | 成本上行 | 后装套件需要实体部件 | Equipment World 对硬件的描述 | 毛利率起步低于 SaaS |
| 安装与校准人工 | 成本上行 | 部署需要按工地做适配 | 后装 + 现场部署报道 | 早期利润率呈重服务特征 |
| 现场运营 / 支持 | 成本上行 | 客户需要安全、可靠的正常运行时间 | 意味着需要工地现场持续支持 | 利润率取决于可重复性 |
| 重复工作流相似度 | 利润率上行 | 标准化工地减少定制工作 | 大规模土方开挖验证具备重复性 | 最利于打出贡献毛利的切入点 |
| 有人监督与无操作员模式 | 利润率随时间上行 | 减少人工监督会改善单位经济模型 | 无操作员模式仍是前瞻目标 | 近期利润率可能仍处过渡期 |
公司未披露部署损益,因此单位经济模型分析属于推断;本表突出最可能起关键作用的变量。
[CI011, CI012, CI013, CI014, CI015]早期,硬件和现场支持压住毛利率;之后,可重复性和监督减少会改善模型。
桥接数值是概念性的贡献驱动因素,不是已披露毛利率百分比。
[CI011, CI012, CI013, CI014, CI015, CI033]4.3 资本充足性和现金跑道逻辑
Gravis 公开拥有的东西是资本。公司在 2025 年 7 月完成 $80M 启动融资,七个月后又完成 $200M Series A 轮,使公开披露累计融资至少约 $223M。这给了它比多数早期自主化初创公司强得多的现金垫。它也告诉投资人一件重要的事:Gravis 获得的是资本密集型扩张公司的融资方式,而不是小额融资的软件实验。对一家需要硬件、安全验证、客户部署团队以及潜在库存的公司来说,这很合理。未解决的问题是充足性,而不是绝对金额。没有烧钱速度、员工数或现金余额披露,外部投资人仍无法判断当前弹药库能支撑两年有纪律的执行,还是在部署快速扩张时只够更短的现金跑道。股权结构表广度显示 Gravis 大概率还能再融资,但未来融资筹码取决于当前资本能否转化为可复制的商业证据。[CI016, CI017, CI018, CI019, CI020, CI029]
| 主题 | 公开事实 | 重要性 | 置信度 | 缺口 |
|---|---|---|---|---|
| B 轮规模 | $270M | 为产品和部署扩张提供资金 | 高 | 资金用途未完全详述 |
| 累计融资 | >$350M | 降低短期融资风险 | 高 | 现金余额未披露 |
| 初始融资 | $80M 种子轮 + A 轮 | 显示公开发布前已有投资者支持 | 高 | 入场估值未披露 |
| 资本密集度 | 可能较高 | 硬件 + 现场运营需要现金 | 中 | 需要烧钱预测 |
| 后续融资选项 | 可能较强 | 多元股东结构能支撑未来融资 | 中 | 需要投资者 pro-rata 细节 |
融资历史有充分支撑;在 Gravis 披露烧钱速度和招聘计划前,充足性判断必然带有推断。
[CI016, CI017, CI018, CI019, CI020]现金必须先从融资流向硬件、现场运营、安全验证和重复部署,之后才可能出现软件式杠杆。
这条流程描述财务结构,不是历史现金流量表项目。
[CI016, CI018, CI019, CI020, CI028, CI029]4.4 公开缺口和承保边界
本章的核心限制在于,Gravis 对融资的披露远比经营表现清楚。公开来源没有提供收入、ARR、利润率、客户数、全公司员工数或烧钱速度。因此,今天无法诚实地套用传统收入倍数或按毛利率调整的框架。最有用的公开承保框架反而更简单:公司是否有足够资本推进路线图,现场证据积累速度是否足以支撑下一步估值?这个基础弱于投资人理想状态,但对现阶段私营公司仍有信息量。它迫使后续估值工作保持情景化,而不是假装精确。Gravis 可能成为高度可扩展的自主化平台,但仅凭公开证据还无法把这种结果与一个资金充足的试点项目区分开。缺失指标不是脚注,而是剩余尽调的主线。这种不确定性应直接计入推荐置信度。[CI021, CI022, CI023, CI024, CI025, CI027]
| 缺失指标 | 公开状态 | 为何阻碍投资测算 | 可用代理指标 | 尽调路径 |
|---|---|---|---|---|
| 收入 / ARR | 未披露 | 无法检验规模或可重复性 | 已签约部署数量 | 索取已签约收入和实际运行收入 |
| 毛利率 | 未披露 | 无法与软件或机器人同行比较 | 部署成本模型 | 审阅毛利率拆解 |
| 客户数 | 未披露 | 集中度风险未知 | 已命名伙伴清单 | 索取活跃客户名单 |
| 烧钱速度 / 现金跑道 | 未披露 | 无法评估现金是否足够 | 仅有融资金额 | 索取现金计划 |
| 员工数 | 未披露 | 无法对标产出效率或烧钱 | 招聘页面 / 领导层招聘 | 索取组织层面人员配置数据 |
本表有意列出这些未知项;仅靠公开证据,常规私营公司投资测算无法完成。
[CI021, CI022, CI023, CI024, CI025]公开证据对融资和估值区间的支撑远强于对任何经营指标区间的支撑。
融资与估值是公开报道区间;收入被刻意展示为基本不可得,而不是猜测值。
[CI016, CI017, CI021, CI022, CI023, CI024]05产品与技术
5.1 产品到底是什么
Gravis 的产品最好理解为改装式自主化栈,而不是一台新的 OEM 机器。公司自己的材料把 Gravis Operator 描述为可加装到现有重型设备上的传感器加软件系统。公开部署报道补上了更多细节:LiDAR、GPS、惯性传感器、摄像头和车载计算装在机器上,远程进度可视化则帮助把自主化接入工地运营。这个组合很重要,因为它告诉投资人产品边界真正在哪里。Gravis 卖的是让今天的车队换一种方式运转,而不是卖一支新车队。因此,产品必须同时解决机器人问题和部署工程问题。硬件、机器集成和软件都是产品的一部分;这提高了复杂度,但如果 Gravis 能让承包商觉得改装是常规动作,也会形成更强切口。招聘页面还显示,工程深度仍在快速扩张。[CE001, CE002, CE003, CE004, CE005, CE026]
| 模块 / 资产 | 公开证据 | 角色 | 重要性 | 置信度 |
|---|---|---|---|---|
| 传感器 | LiDAR、GPS、IMU、摄像头已有公开描述 | 感知与定位 | 机器安全感知的核心 | 高 |
| 车载计算 | 驾驶室内计算机已有公开描述 | 在本地运行自主作业栈 | 快速响应所必需 | 高 |
| Gravis Operator 软件 | 官方网站点名 | 自主作业和编排层 | 定义产品身份 | 高 |
| 实时智能层 | 公开资料突出进度跟踪 | 监控与监督 | 把自主作业接入项目管理 | 高 |
| 后装安装套件 | 公开资料描述可在数小时级可逆安装 | 把产品带入现有车队 | 关键商业化切入点 | 高 |
本表只反映公开描述的组件;内部模型架构和底层控制设计仍未披露。
[CE001, CE002, CE003, CE004, CE005]Gravis 的架构结合传感、车载计算、机器学习软件、监督和改装安装。
这张架构图把技术栈简化为公开层级,并不意味着完整内部系统图。
[CE001, CE002, CE003, CE011, CE012, CE013]5.2 工作流匹配和运营模式
公开证据对 Gravis 今天最适合的场景非常一致:大型工地上的重复挖掘和装车。这是特点,不是限制。重复性工作流最能让承包商感到劳动力短缺,也最能让机器通过更长工时、更低疲劳和更可预测的循环时间产生可量化 ROI。Gravis 的合作伙伴和媒体报道还显示,公司很努力地嵌入现有承包商运营,而不是强迫客户改用全新工作方式。安装套件,跑有监督作业,衡量进度,再复制。对一家年轻自主化公司来说,这是一条合理运营路径,因为它让客户在验证表现时仍能把人留在闭环附近。下一个问题是,这条路径会自然扩展到更广泛的工地自主化,还是只在狭窄的重挖掘任务中最有力。这个过渡将决定 Gravis 是工作流解决方案,还是更广的平台。[CE006, CE007, CE008, CE009, CE010, CE028]
| 工作流 | 公开验证 | 当前匹配度 | 匹配原因 | 约束 |
|---|---|---|---|---|
| 大规模土方开挖 | 是 | 高 | 重复且可衡量 | 需要安全地与卡车交互 |
| 卡车装料 | 是 | 高 | 周期重复、目标明确 | 要求铲斗动作精确 |
| 通用场地准备 | 是 | 中高 | 大型工地有可重复的移动模式 | 工地差异性 |
| 远程 / 用工受限工地 | 隐含 | 中高 | 操作员稀缺提升 ROI | 支持与后勤 |
| 完全无操作员车队运营 | 仅为前瞻性 | 未来 | 若跑通,上行最大 | 安全与成熟度门槛 |
公开证据最强的是有监督的重复挖掘任务;更广泛的自主化仍主要停留在路线图层面。
[CE006, CE007, CE008, CE009, CE010]产品嵌入承包商工作流:先改装安装,再进入有人监督的作业,最终目标是更低人工介入的自主化。
运营阶段综合了发布材料和现场部署报道。
[CE004, CE006, CE007, CE008, CE009, CE010]5.3 技术架构和关键依赖
即使公司没有发布技术白皮书,Gravis 的架构逻辑也很清楚。创始人相信,在 ETH 和现场发展出的数据驱动自主化技术可以迁移到建筑业;在这里,机器必须实时理解地形、移动资产和工地目标。这个挑战比直线导航更难,因为建筑设备不只是穿过世界;它一边工作一边改变世界。这意味着感知、规划和控制都必须跟上动态地形,也要跟上附近作业的人和卡车。它还意味着现场运营会成为技术系统的一部分,因为部署质量、校准和客户信任都会影响软件能否发挥作用。对 Gravis 来说,产品架构和运营架构不可分割。因此,数据、现场支持和承包商共研都不是可选附加项,而是依赖。产品必须同时在技术和运营上跑通。[CE011, CE012, CE013, CE014, CE015, CE017]
| 层级 | 公开描述 | 依赖 | 风险 | 含义 |
|---|---|---|---|---|
| 感知 | 地形、障碍物、作业区感知 | 传感器 + 校准 | 粉尘 / 遮挡 / 杂物 | 稳健感知是任务关键 |
| 规划 | 目标驱动的自主作业执行 | 项目计划 + 状态估计 | 意外工地变化 | 工作流适配很关键 |
| 控制 | 精准机器执行和循环可重复性 | 机器接口 | 延迟 / 机器差异 | 改装集成质量是关键 |
| 监督 / 监控 | 实时进度可见性和监督 | 遥测和 UI | 告警疲劳 / 界面弱 | 人的信任取决于可见性 |
| 部署 / 设置 | 小时级安装和可逆改装 | 现场运营流程 | 设置摩擦过大 | 部署工程也是产品的一部分 |
架构来自公开描述和工地报道推断,而不是来自已发布的技术白皮书。
[CE011, CE012, CE013, CE014, CE015]产品要成功,传感质量、机器集成、现场运营、客户信任和安全验证必须同步推进。
这些依赖是方向性、概念性的;它们说明商业化需要哪些环节协同跑通,而不是内部工程组织图。
[CE015, CE016, CE017, CE018, CE019, CE020]5.4 信任、安全和产品成熟度
Gravis 的产品故事在公开证明和成熟度重合的地方最强:与真实承包商伙伴一起推进有监督挖掘自主化。这足以支撑技术可信度,但不等于广泛商业成熟。安全仍是核心,公司围绕作业区感知和减少意外的表述,也隐含承认自主化买方会先按风险评判产品。有监督部署的存在说明 Gravis 理解这一点,并把人工监督当作成熟度和信任的桥梁。OSHA 和 CDC 提供的外部安全语境也强化了这种做法的合理性。真正的成熟度测试还在前面:Gravis 能否从有监督成功走向无操作员、少人工介入的商业部署,同时不引入足以吓退客户的摩擦或风险?产品看起来有前景,设计方向也合理,但仍处在自主化成熟曲线很陡的阶段。因此,验证速度几乎和原始技术野心一样重要。[CE016, CE017, CE018, CE019, CE021, CE022]
| 信任维度 | 公开信号 | 重要性 | 当前状态 | 尽调问题 |
|---|---|---|---|---|
| 安全叙事 | 超人级安全 / 作业区感知表述 | 核心买方信任 | 营销说法 + 合作伙伴背书 | 需要客观安全指标 |
| 有监督部署 | 是 | 产品成熟期保持审慎 | 公开证据强 | 需要推进标准 |
| 与承包商共创 | 是 | 改善工作流适配和可信度 | 公开证据强 | 需要重复转化数据 |
| 监管对齐 | OSHA/CDC 背景相关 | 建筑安全审查严格 | 外部压力高 | 需要合规运营模型 |
| 机器可逆改装 | 是 | 降低采用顾虑 | 已公开说明 | 需要真实操作员使用数据 |
公开信任证据更多来自叙事和合作伙伴引述,正式安全披露更弱。
[CE016, CE017, CE018, CE019, CE020]| 能力 | 当前阶段 | 公开证据 | 下一关口 | 风险 |
|---|---|---|---|---|
| 有监督挖掘自主化 | 进行中 | 多个公开工地报道 | 扩展到更多工地 | 中等 |
| 卡车装载流程 | 进行中 | Phoenix 项目证据 | 更高利用率和一致性 | 中等 |
| 多合作伙伴部署计划 | 进行中 | 合作伙伴名单扩大 | 将合作伙伴转为重复项目 | 中等 |
| 无操作员挖掘机部署 | 已设目标 | 已披露 2026 目标 | 安全和可靠性签批 | 高 |
| 广泛多机器编排 | 初现概念 | Series B 叙事 | 展示车队协同 | 高 |
该表区分已公开演示的能力与仍停留在路线图表述的内容。
[CE021, CE022, CE023, CE024, CE025]Gravis 在有人监督的挖掘上最强,在广泛无人值守车队自主化上还不成熟。
成熟度标签是基于 Gravis 已公开展示内容与仍面向未来能力的定性综合判断。
[CE021, CE022, CE023, CE024, CE025, CE032]06客户情况
6.1 客户是谁
Gravis 的公开客户故事从承包商开始,而不是从开发商、市政机构或设备 OEM 开始。这说得通,因为公司解决的是工地上的工作流问题:谁拥有机器、谁难以配齐操作员、任务更快完成时谁获得回报。因此,总承包商和土方专业公司是最清晰的第一批细分市场。具名合作伙伴名单也支持这一点,核心是 Sundt、Zachry、Champion Site Prep 和 Capitol Aggregates。租赁公司还不能算已证明的客户,但在战略上重要,因为改装产品比 OEM 锁定系统更容易跨混合车队流动。大型 EPC 和超级项目建设方也重要,因为它们运营的是资本密集型工地,进度压力、劳动力稀缺和重复场地作业会创造最高的自主化 ROI。这些细分市场给了 Gravis 一个合理的客户排序策略,也说明企业销售纪律会很早变得重要。[CU001, CU002, CU003, CU004, CU005, CU030]
| 客群 | 公开证据 | 买方逻辑 | 适配原因 | 当前置信度 |
|---|---|---|---|---|
| 总承包商 | 高 | 承担进度风险 | 需要提升场地准备吞吐量并撬动劳动力杠杆 | 高 |
| 土方承包商 | 高 | 重复性挖掘流程 | 最匹配已披露用例 | 高 |
| 骨料 / 材料运营商 | 中 | 重型机械重复作业 | 逻辑上相邻适配 | 中 |
| 租赁公司 | 低 | 混合车队渠道潜力 | 改装模式兼容 | 中低 |
| 大型 EPC / 超级项目建造商 | 间接 | 大规模场地准备和基建工程 | 大客户机会 | 中 |
细分表把已确认的公开证据,与战略上合理但尚未公布的渠道分开。
[CU001, CU002, CU003, CU004, CU005]Gravis 目前的客户旅程从问题识别开始,进入伙伴式测试、有人监督的部署、验证,最后才可能扩张。
这张旅程图反映公开市场进入证据,而不是已披露内部 CRM 漏斗。
[CU001, CU006, CU007, CU008, CU010, CU026]6.2 采用证据和具名客户证明
客户证据强于典型早期初创公司,但仍不同于成熟企业软件的客户账本。Gravis 有具名伙伴、公开工作流引述,以及来自 Phoenix 真实工地的运营指标。65,000 立方码这个数字很重要,因为它把客户证明从抽象兴趣转化为可衡量活动。同时,公司没有公布单客户收入、按账户划分的部署数量或任何标准化转化漏斗。因此,正确解读是“可信且正在改善的证明”,而不是“采用已经完全去风险”。参考客户质量确实有帮助。Sundt 和 Austin Bridge 在重型土木和场地工程中有真实分量,Champion 则展示了专业挖掘需求。今天的客户证明是运营性和证言性的;经济证明是缺失的一层。这个区分应压住任何轻易的增长叙事。投资人仍需要把参考质量和收入质量分开看。[CU006, CU007, CU008, CU009, CU010, CU011]
| 阶段 | 公开信号 | 证据 | 含义 | 置信度 |
|---|---|---|---|---|
| 启动合作伙伴组合 | 上线时四家公司 | 官方 + TechCrunch | 初始客户足迹 | 高 |
| Phoenix 证据 | 130-acre 工地 | Equipment World + ENR | 运营可信度 | 高 |
| 已搬运物料 | 65,000+ cubic yards | Equipment World + ENR | 具体产出证据 | 高 |
| 合作伙伴扩张 | 新增 Austin / Maverick / Haydon | Equipment World + ENR | 更广泛商业兴趣 | 中 |
| 收入转化 | 未披露 | 无公开来源 | 最大采用缺口 | 低 |
采用证据真实存在,但仍以部署为中心,而非以收入为中心。
[CU006, CU007, CU008, CU009, CU010]| 账户 / 合作伙伴 | 公开证据 | 验证内容 | 来源质量 | 含义 |
|---|---|---|---|---|
| Sundt Construction | 引述 + 现场部署报道 | 缓解重复卡车装载,并证明可在活跃工地使用 | 高 | 最强公开客户证据 |
| Zachry | CEO 引述 | 安全和进度目标 | 中 | 高管级验证 |
| Champion Site Prep | CEO 引述 | 车队协同和班组人效放大 | 中 | 土方专业客户证据 |
| Austin Bridge & Road | 官方合作伙伴公告 | 工人保护和精度 | 中 | 新近合作伙伴验证 |
| Capitol Aggregates | 具名合作伙伴 | 骨料 / 重型设备相邻场景 | 中 | 拓宽客群地图 |
经济细节稀疏,但具名证据覆盖大型承包商和土方专业公司。
[CU011, CU012, CU013, CU014, CU015]公开采用路径似乎先从具名合作伙伴开始,再到有人监督的部署指标,之后才走向未知的收入转化。
由于 Gravis 未披露客户转化指标,漏斗后段仍属推断。
[CU006, CU007, CU008, CU009, CU010, CU017]具名验证在工作流减压和安全表述上最强,经济性验证仍然偏薄。
这张矩阵刻意把验证质量和已披露经济性拆开看;所有具名客户的经济性披露仍很稀疏。
[CU011, CU012, CU013, CU014, CU015, CU027]6.3 留存、耐久性和扩张逻辑
留存是公开证据很快断掉的地方。没有披露来源提供续约率、NRR、流失率或账户级扩张模式。今天最好的代理指标,是参考伙伴是否继续加深合作,以及 Gravis 能否在不损失早期部署运营质量的情况下新增承包商。这有用,但不能替代队列数据。建筑科技可以赢下一个强试点,但如果培训负担、支持负荷或工作流扰动居高不下,仍可能难以变成重复运营预算项。Gravis 的承诺是,在需要时保留人工监督,同时帮助施工队处理重复性土方。如果承诺成立,扩张应该可能;如果不成立,客户关系可能停留在浅层、项目特定状态。眼下,耐久性更多是尽调问题,而不是公开事实。投资人应把留存视为未解决,而不是默认存在。可重复性仍是公开层面缺失的商业门槛。[CU016, CU017, CU018, CU019, CU020, CU029]
| 信号 | 公开状态 | 最佳代理指标 | 重要性 | 缺口 |
|---|---|---|---|---|
| 续约率 | 未披露 | 工地重复使用 | 显示持久性 | 无数据 |
| 账户内扩张 | 未披露 | 合作伙伴计划扩张 | 显示账户增长 | 无账户级数据 |
| 客户满意度 | 仅基于引述 | 客户背书质量 | 落地后扩张所必需 | 无调研数据 |
| 运营可重复性 | 部分可见 | 大规模挖掘重复性 | 支撑 ROI 叙事 | 仍依赖具体工地 |
| 多年持久性 | Unknown | None | 检验客户是否留下 | 无队列数据 |
留存证据刻意稀疏,因为公司尚未披露填补该表所需的队列数据。
[CU016, CU017, CU018, CU019, CU020]续约和 NRR 未披露,公开证据只能支撑早期概念性队列视角。
这是基于公开证据的概念性队列图,不是公司披露的留存表。
[CU016, CU017, CU018, CU019, CU020, CU033]6.4 集中度和渠道风险
由于公开具名账户集合仍然很小,集中度风险今天几乎肯定有意义。对这个年龄的公司来说,这并不罕见,但它很重要,因为少数设计伙伴关系会塑造路线图、参考质量和近期收入。Gravis 降低这种风险的最佳机会,是把强参考账户变成飞轮,打开相邻承包商,并最终打开租赁公司等渠道伙伴。数据中心和本土制造等终端市场尤其有吸引力,因为它们把进度紧迫性和大规模场地准备范围结合起来;但同样的大型项目往往也伴随严苛采购流程。因此,客户章节和财务章节落在同一个结论:Gravis 的证明足以支持继续关注,但公开转化数据还不足以假设广泛、耐久的客户采用。渠道杠杆是接下来要看的关键上行。集中度和扩张必须放在一起评估,不能分开看。这个框架对承保纪律很重要。[CU021, CU022, CU023, CU024, CU025, CU032]
07风险
7.1 监管和法律风险
任何把自主系统装到重型机械上的公司,都继承了很高的举证负担。建筑业本来就是危险行业,OSHA、CDC 和 BLS 材料清楚显示,即使还没有加入自主化,危险也持续存在。因此,Gravis 不能只因为说自己的系统更安全就获得信用。它必须用监管方、客户和保险方都能信任的方式证明安全。Frontiers 和 ILO 的外部研究进一步强化这一点:机器人可以同时降低某些危险,也引入新的危险。对 Gravis 来说,眼前的法律问题不是建筑业是否需要更好的安全工具——显然需要。问题在于,Gravis 从有监督部署走向少人工介入作业时,能否建立可复制的责任和合规框架。公开层面这仍未解决。法律清晰度可能在相当长时间里落后于技术曲线,法院和保险方在实践中也可能适应缓慢。[CR001, CR002, CR003, CR004, CR005, CR032]
| 风险 | 重要性 | 公开证据 | 当前严重性 | 尽调问题 |
|---|---|---|---|---|
| 机器人安全合规 | 自主设备带来独特风险 | OSHA 机器人指南 | 高 | Gravis 如何让运营对齐 OSHA 要求? |
| 建筑行业死亡事故基线 | 行业高危,抬高了出错容忍门槛 | CDC + BLS | 高 | Gravis 如何衡量安全改善? |
| 新自动化风险 | 可能引入机械和心理社会风险 | Frontiers + ILO | 中高 | 哪些风险在主动跟踪? |
| 责任 / 保险不确定性 | 索赔责任分配可能不清 | OSHA + ILO 背景 | 高 | 谁承担哪些责任? |
| AI 治理与问责 | 建筑 AI 可能造成问责缺口 | RICS | 中 | 安全关键变更由谁签批? |
该表把直接监管内容与更广泛的机构级风险分析合并,因为 Gravis 自身没有披露法律框架细节。
[CR001, CR002, CR003, CR004, CR005]监管、运营、商业化风险都不小;现阶段没有哪一类可以安全忽略。
热度标签是基于公开证据的综合判断,不是公司发布的风险评分。
[CR001, CR002, CR006, CR011, CR016, CR026]7.2 运营和依赖风险
Gravis 的运营风险来自一个事实:它的系统必须在临时、杂乱、变化中的工地上工作,而不是在受控工厂里工作。粉尘、地形变化、移动卡车和施工人员都会提高感知、规划和现场运营的负担。公开证明令人鼓舞,但仍处在有监督阶段,因此不同于完全成熟的产品。依赖图谱会进一步放大这一点。Gravis 需要承包商伙伴来学习和证明,需要现场团队保证部署质量,还需要持续兼容并非自己制造的机器。资本也是一种依赖,因为一家全栈自主化公司可能在商业经济性清晰之前就大量花钱。这并不让业务不可行,但意味着规模化路径更少依赖纯软件分发,而更依赖在多个外部约束下有纪律地执行系统。运营卓越不仅是成本中心,也是风险控制。[CR006, CR007, CR008, CR009, CR010, CR011]
| 风险 | 机制 | 证据 | 严重性 | 缓释思路 |
|---|---|---|---|---|
| 感知失效 | 粉尘 / 遮挡 / 杂乱 | 公开技术栈 + Frontiers | 高 | 冗余感知与校验 |
| 安装 / 校准负担 | 临时工地持续变化 | Gravis + 部署报道 | 中高 | 完善安装作业手册 |
| 支持强度 | 异常过多,需要人工介入 | 有监督部署 | 中高 | 提升自动化可靠性 |
| 工作流脆弱性 | 复杂工地会打破窄场景假设 | 施工场景 | 中 | 聚焦可重复任务 |
| 安全 / 遥测短板 | 远程监督依赖可信数据流 | 实时监控叙事 | 中 | 审计连接性和数据处理 |
把安全风险列入概念框架,是因为远程监控和机器遥测会制造数据依赖,即便公开资料没有披露泄露证据。
[CR006, CR007, CR008, CR009, CR010]| 依赖项 | 重要性 | 当前信号 | 风险 | 尽调问题 |
|---|---|---|---|---|
| 承包商合作伙伴 | 提供工地和学习闭环 | 强 | 集中度 | 每个合作伙伴有多少活跃工地? |
| OEM 兼容性 | 加装技术栈会触及既有机器 | Unknown | 保修或接口摩擦 | 是否有 OEM 限制? |
| 现场运营团队 | 部署质量决定信任 | 关键 | 执行瓶颈 | 现场运营可扩展性如何? |
| 资本市场 | 自主化扩张会烧钱 | 目前支持 | 未来融资冲击 | 增长放缓时现金跑道多长? |
| 终端市场需求 | 客户紧迫性取决于项目管线 | 当下强 | 宏观放缓 | ROI 对需求有多敏感? |
这些依赖项不在软件栈内,但仍会决定产品能否成功商业化。
[CR011, CR012, CR013, CR014, CR015]一次安全或可靠性失效,可能连锁冲击客户信任、责任成本和融资。
这张图展示的是可能的业务传导链条,不是已报告事故。
[CR001, CR005, CR010, CR024, CR027, CR028]Gravis 的产品要跑通,客户、OEM 兼容性、现场运营和资本基础必须同时稳住。
依赖不只是技术问题,也关乎战略和运营。
[CR011, CR012, CR013, CR014, CR015, CR031]7.3 人才、劳动力和采用风险
自主化采用从来不只是技术问题。它会改变工作组织方式,改变哪些人感到受威胁或被赋能,也改变客户在依赖系统前必须投入多少培训和信任建设。Gravis 的合作伙伴引述很聪明,把产品描述为释放熟练操作员,让他们去做更有价值的任务,而不是简单替代他们。即便如此,Brookings、St. Louis Fed 和 ILO 都显示,工人替代叙事可能变成真实的采用障碍。公司内部也面对经典初创公司执行风险:公开身份紧密绑定少数创始人,招聘很快,管理层梯队仍在追赶估值暗示的规模。如果变革管理或劳动力接受度落后于产品路线图,即使技术继续改善,客户扩张也会放慢。人的因素可能成为隐藏瓶颈。[CR016, CR017, CR018, CR019, CR020, CR030]
7.4 缓释因素和止损条件
Gravis 确实有可见的缓释因素。有监督部署在产品成熟期间保留了一层人工安全保障。可逆改装降低了买方焦虑,因为机器可以退回手动作业。合作伙伴共研确保产品在真实工作流上训练,而不是在合成演示里训练。这些都是有意义的正面因素。但它们不是无限保护。Gravis 最终必须证明,有监督成功能转化为更安全、少人工介入、经济上可重复的运营模式。严重事故模式、无法把伙伴转化为耐久项目,或 OEM 快速追赶,都应成为投资逻辑的真实止损条件。因此,正确的投资姿态不是因为风险高就否定公司,也不是因为痛点真实就忽视风险。正确姿态是要求证据证明 Gravis 的学习曲线跑赢风险曲线。这是下一次刷新时的核心风险测试,也是最清晰的董事会级监控议程。[CR021, CR022, CR023, CR024, CR025, CR031]
08估值
8.1 推荐逻辑
Gravis 值得认真放进投资人观察名单,因为它用可信的自主化人才、可见的旗舰合作伙伴,以及 SoftBank 领投的强劲 $200M Series A 轮,进攻一个巨大而痛苦的建筑问题。话虽如此,公开记录还不足以支撑高确信度的看多承保。原因很直接:Gravis 的估值已经反映品类领导者野心,但公开证据在收入质量、利润率结构、客户转化深度和部署可重复性上仍然滞后。投资人能看到融资、改装逻辑和几个强客户品牌;还看不到商业队列、留存,或价值中有多少正在变成软件式而不是部署重。这个组合要求推荐纪律。证据足以继续跟进,但仅凭公开信息还不足以承保一个强买入。继续研究是合适结论,因为公司有前景,价格也已经不低。[CV001, CV002, CV003, CV004, CV005, CV006]
| 维度 | 评估 | 原因 | 置信度 | 含义 |
|---|---|---|---|---|
| 建议 | 继续研究 | 痛点和验证有吸引力,经济性尚不完整 | 中 | 保持跟进,但要求更深入尽调 |
| 置信度 | 中 | 关键事实真实,但运营指标缺失 | 中 | 避免虚假精确 |
| 风险评级 | 高 | 安全、执行和商业化都重要 | 中 | 坚持下行约束 |
| 估值立场 | 偏高 | 公开收入证据前已是独角兽价格 | 中 | 需要里程碑约束 |
| 主要支撑 | 合作伙伴 + 投资人背书强 | 存在真实部署和大额资本 | 高 | 投资逻辑仍成立 |
| 主要阻碍 | 财务披露薄弱 | 仅靠公开记录难以建模回报 | 高 | 需要更深入尽调 |
该表把公开证据转化为投资人立场,而不是假装现有披露足以支撑完整估值模型。
[CV001, CV002, CV003, CV004, CV005, CV006]| 立场 | 表述 | 证据 | 重要性 | 置信度 |
|---|---|---|---|---|
| 投资逻辑 | 市场痛点大且明显 | 劳动力短缺、安全压力、生产率拖累 | 支撑需求 | 中 |
| 投资逻辑 | 加装模式可能比依赖更换周期的硬件扩张更快 | 可用于既有设备 | 支撑 ROI 逻辑 | 中 |
| 投资逻辑 | 按阶段看,SoftBank 和合作伙伴背书异常强 | $200M Series A 轮加上可见现场背书 | 支撑可信度 | 高 |
| 反向逻辑 | 仍可能是一门重部署业务 | 没有公开证据证明软件式经济性 | 限制扩张信心 | 中 |
| 反向逻辑 | OEM 和相邻领域可能挤压切入口 | 在位者掌握渠道和机器关系 | 护城河变窄 | 中 |
| 反向逻辑 | 估值可能跑在证据前面 | 公开经济性有限 | 降低新投资人的上行空间 | 中 |
反向逻辑并非为了看空而看空;它抓住的是当前估值似乎已经假定不存在的问题。
[CV007, CV008, CV009, CV010, CV011, CV012]建议遵循一条简单链路:问题足够大,验证有可信度,关键缺口仍在,价格已经拉高,因此维持中等信心的继续研究立场。
这条流程反映的是本报告的判断逻辑,不是公司发布的决策框架。
[CV001, CV002, CV003, CV004, CV005, CV035]公开 KPI 在融资和验证上较强,但经济性和耐久性仍弱。
KPI 组刻意排除了未披露的收入、利润率和留存数据。
[CV001, CV004, CV005, CV006, CV035]8.2 乐观 / 基准 / 悲观情景框架
本章采用情景分析,因为点估值会暗示一种公开证据并不支持的精确性。乐观情景下,Gravis 把旗舰客户和 OEM 参考转化为跨多个机器品牌和工地类型的可重复、少人工介入项目,让投资人相信软件和数据杠杆更耐久。基准情景下,它成为一家有价值但运营仍然较重的自主化专科公司,持续获得战略支持,并保持不错的部署动能。悲观情景下,客户继续喜欢演示和试点结果,却没有转化为足够广泛或盈利的规模化项目,使当前估值跑在证明之前。关键问题不是某个精确数字,而是区分这些路径的一组里程碑:安全和责任准备度、工地泛化表现、客户扩张,以及部署劳动在价值交付中占比下降的速度。情景纪律可以避免虚假精确,也能讲清投资人下一步应监控什么。[CV013, CV014, CV015, CV016, CV017, CV031]
| 情景 | 核心假设 | 运营结果 | 估值含义 | 必须成立的条件 |
|---|---|---|---|---|
| 乐观 | 多工地项目可重复 + 部署触点降低 + 数据杠杆 | 成为加装式自主化品类领导者 | 当前估值之外仍有上行 | 里程碑快速落地 |
| 基准 | 有战略支持的有用自主化细分市场 | 公司不错,但运营仍重 | 估值大致站得住,但不便宜 | 客户证据稳定 |
| 悲观 | 试点无法稳定转化,软件杠杆仍弱 | 演示强,规模经济性弱 | 当前估值显得过早 | 商业耐久性仍弱 |
| 乐观 / 悲观摆动因素 | 客户扩张和可重复性 | 决定是软件式形态还是重服务形态 | 最敏感变量 | 需要队列数据 |
| 乐观 / 悲观摆动因素 | 安全 / 责任准备度 | 决定无人值守部署速度 | 可能抬高或压缩估值 | 需要事故和保险证据 |
该情景表有意按里程碑驱动,因为公开数据不足以支持点估值。
[CV013, CV014, CV015, CV016, CV017]估值对重复部署转化、软件杠杆、安全 / 责任准备度最敏感。
1–5 档敏感性评分基于留存的公开证据,不是统计模型。
[CV006, CV013, CV014, CV015, CV016, CV017]公开证据只能支撑围绕当前估值的宽区间;上行需要打出品类领导者级别的执行,下行则会在转化或安全验证停滞时很快显现。
情景区间是从里程碑信心推导出的示意结果,不是按市场交易可比公司得出。
[CV013, CV014, CV015, CV031, CV032, CV033]8.3 可比框架及其边界
Gravis 没有整齐的公开可比公司集合。Built Robotics 有用,因为它展示了一家初创公司如何围绕窄工作流做出建筑自动化切口,并随后仍然面对艰难的规模化选择。Caterpillar 和 Komatsu 重要,因为它们展示了现有 OEM 能带到自主化中的渠道、融资和服务力量。Hexagon 和 Trimble 重要,因为工作流控制层即使不拥有完整机器栈,也能变得有力量。Teleo 和相关车队现代化厂商重要,因为它们验证了买方愿意升级传统车队,而不是等待完全新的设备周期。但这些都不是干净的倍数可比。它们的产品、上市路径、资本需求和客户经济性差异太大。因此,本章把可比对象当作原型,而不是假装一张公开倍数表能解决问题。Gravis 应该按它可能成为的样子估值——重型设备的跨品牌自主化层;同时也要承认,它可能永远达不到投资人今天隐含期待的规模或利润率。[CV018, CV019, CV020, CV021, CV022, CV023]
| 可比原型 | 示例 | 相关性 | 不完美之处 | 结论 |
|---|---|---|---|---|
| 聚焦工作流的初创公司 | Built Robotics | 显示窄施工自动化切入口的价值 | 转向太阳能聚焦后的演进不是干净类比 | 有用的方向性可比 |
| OEM 在位者 | Caterpillar / Komatsu | 体现机器、渠道和服务力量 | 上市 OEM 经济性不可比 | 是威胁,不是干净的倍数可比 |
| 工作流软件在位者 | Hexagon / Trimble | 体现工作流和定位控制的价值 | 机器自主化关联较弱 | 重要相邻领域 |
| 存量设备现代化 | Teleo | 显示升级存量设备的需求 | 远程操作模式不同于自主化逻辑 | 仅部分可比 |
| 成长资本基准 | SoftBank / CapitalG / Georgian / 8VC | 传递雄心和扩张预期 | 投资人名望不是运营证明 | 不要过度解读股权结构质量 |
Gravis 几乎没有公开近似对标公司,因此可比估值更偏原型法,而不是统计法。
[CV018, CV019, CV020, CV021, CV022, CV023]8.4 打破投资逻辑的触发点和最终尽调要求
最终投资判断应转向少数几个决定性事实。如果 Gravis 能证明安全、可重复的部署扩张、改善中的软件杠杆和可信的客户队列经济性,当前估值仍可能说得通。反过来,如果出现安全事故、客户停在试点阶段,或相邻巨头在 Gravis 规模化之前更快补上差距,投资人就应假设估值过高。这里的纪律,是在下一轮叙事动能到来之前先定义止损触发点。因此,最终尽调问题应该务实而不是学术:收入和毛利率队列、安全与保险包、从有监督到少人工介入自主化的过渡里程碑,以及能明确下行保护的股权结构表或优先权细节。没有这些,信心最多只能保持中等。有了这些,Gravis 可能从有意思的建筑科技自主化押注,变成可投资的高确信度案例——也可能变成更清晰的放弃。眼下,里程碑应比故事本身更能决定定价。[CV024, CV025, CV026, CV027, CV028, CV029]
| 触发因素 | 重要性 | 早期预警信号 | 严重性 | 投资人应对 |
|---|---|---|---|---|
| 安全或可靠性事故反复出现 | 削弱信任、保险可承保性和推广速度 | 介入次数增加、工地撤回或事故披露 | 危急 | 暂停投资评估 |
| 试点转项目转化弱 | 显示商业耐久性弱 | 试点多,规模化部署少 | 高 | 下调估值 / 要求需求证明 |
| OEM 或相邻玩家追赶 | 缩小加装切入口 | 客户偏好捆绑式或更简单替代方案 | 高 | 重新评估护城河 |
| 烧钱但没有软件杠杆 | 稀释回报并抬高融资风险 | 大额支出,但队列证据有限 | 高 | 要求更严格里程碑 |
| 客户集中度冲击 | 一两个客户贡献过多价值 | 当前标杆客户之外扩张缓慢 | 中高 | 压测下行 |
该表列出最能打破当前投资逻辑的事件,而不是罗列每一种通用初创公司风险。
[CV024, CV025, CV026]| 问题 | 当前为何重要 | 可回答什么 | 优先级 | 负责人 |
|---|---|---|---|---|
| 收入 + 毛利率队列 | 估值中最大的缺失环节 | 商业韧性与经营杠杆 | 紧急 | 财务 |
| 安全 / 责任 / 保险组合 | 低人工介入扩张前必须补齐 | 责任、保险方态度、铺开节奏 | 紧急 | 运营 + 法务 |
| 低人工介入自主化路线图里程碑 | 情景取决于时间点 | 乐观 / 基准 / 悲观权重 | 高 | 产品 |
| 股权结构表、优先权与集中度数据 | 大额集中融资会影响下行与稀释 | 回报框架与下行保护 | 高 | 财务 / 投资人 |
| 可比公司基准包 | 原型可比项仍然粗略 | 回报预期与价格纪律 | 中 | 企业发展 / 投资人 |
这些尽调请求刻意保持务实、面向投资人;要显著提高推荐置信度,最小数据集就是这些。
[CV027, CV028, CV029, CV030]免责声明
本报告是基于公开证据的尽调快照,不构成投资建议。关键财务、法律、技术和合同事实仍未公开;任何投资决策前,都应向管理层和一手文件直接核验。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Gravis Robotics was founded in 2022 as an ETH Zurich spinout focused on autonomous heavy machinery. | 中 | SO003, SO008, SO013 |
| CO002 | Gravis Robotics identifies Zurich, Switzerland as its headquarters. | 中 | SO002 |
| CO003 | Public company materials list Austin, Texas and Oxford, UK as Gravis office locations in addition to Zurich. | 中 | SO002 |
| CO004 | Ryan Luke Johns is the public-facing CEO and co-founder of Gravis Robotics. | 中 | SO003, SO008, SO013 |
| CO005 | Dominic Jud is Gravis Robotics’ co-founder and CTO. | 中 | SO003, SO008, SO013 |
| CO006 | ETH Zurich robotics professor Marco Hutter is a co-founder and board member. | 中 | SO008, SO013, SO012 |
| CO007 | Inc. described Gravis as a roughly 75-person company in August 2026. | 中 | SO013 |
| CO008 | The careers page shows Gravis still recruiting across interface, field application, hardware, perception, platform, autonomy and simulation roles. | 中 | SO007, SO002 |
| CO009 | Gravis announced $23 million of funding in November 2025 to expand in the UK, US and EU. | 中 | SO004, SO017, SO018 |
| CO010 | The 2025 financing was co-led by IQ Capital and Zacua Ventures, with Pear VC, Imad, Sunna Ventures, Armada Investment and Holcim participating. | 中 | SO004, SO020, SO018 |
| CO011 | SoftBank invested $200 million in Gravis Robotics in August 2026. | 中 | SO003, SO008, SO009 |
| CO012 | SoftBank was the sole investor in the 2026 Series A round. | 中 | SO003, SO008, SO014 |
| CO013 | Gravis and multiple publications described the 2026 round as the largest Series A in construction robotics history. | 中 | SO003, SO009, SO015 |
| CO014 | Multiple 2026 publications place the post-money valuation at about $1 billion. | 中 | SO008, SO010, SO013 |
| CO015 | Based on the disclosed 2025 and 2026 rounds, Gravis has publicly raised at least about $223 million. | 中 | SO004, SO003 |
| CO016 | Gravis positions itself as a retrofit autonomy company rather than a new-machine OEM. | 中 | SO001, SO003, SO012 |
| CO017 | The flagship product is the Gravis Rack, an autonomous control kit that bolts onto existing heavy equipment. | 中 | SO001, SO005, SO009 |
| CO018 | Gravis also sells or deploys the Slate tablet interface for operator guidance, remote operation and task definition. | 中 | SO006, SO005, SO011 |
| CO019 | Gravis says its autonomy stack works across mixed fleets rather than locking customers into one manufacturer. | 中 | SO003, SO015, SO012 |
| CO020 | Public materials list Caterpillar, Case, Develon, John Deere, JCB, Hitachi, Sumitomo, Yanmar and Volvo among supported brands. | 中 | SO003, SO010, SO012 |
| CO021 | Gravis says the technology has been adapted for more than a dozen brands, makes and models. | 中 | SO008, SO013 |
| CO022 | Public sources place Gravis deployments from roughly 10-tonne machines to substantially larger excavators. | 中 | SO008, SO010, SO013 |
| CO023 | Publicly described autonomous tasks include trench digging, bulk excavation, truck loading, stockpile management and machine driving. | 中 | SO008, SO013, SO022 |
| CO024 | Gravis repeatedly claims up to 30 percent productivity gains versus peak manual operation. | 中 | SO003, SO008, SO004 |
| CO025 | The company says its systems are deployed across four continents. | 中 | SO003, SO008, SO013 |
| CO026 | Robotics & Automation News reported Gravis was live in seven countries by late 2025 and again cited seven countries during the 2026 US expansion. | 中 | SO018, SO026 |
| CO027 | Named partners and customers in public materials include Holcim, Taylor Woodrow, HD Hyundai and Flannery Plant Hire. | 中 | SO004, SO008, SO010 |
| CO028 | Holcim is both a strategic investor from the 2025 round and a public user of Gravis automation in construction and quarry applications. | 中 | SO020, SO025 |
| CO029 | Taylor Woodrow publicly described a successful autonomous excavator trial and planned deployment at Manchester Airport. | 中 | SO021, SO021 |
| CO030 | Gravis and Flannery were selected to lead an $8 million UK CAM Pathfinder project across six excavators. | 中 | SO003, SO022, SO015 |
| CO031 | In March 2026 Gravis said it was commercially expanding into the US and showing live CONEXPO demos with Hitachi and Develon. | 中 | SO026, SO023, SO024 |
| CO032 | Gravis’ US expansion announcement cited a 60-mile autonomous pipeline project in Argentina with Techint Group. | 中 | SO026 |
| CO033 | The US expansion announcement cited up to 97 percent bucket fill rates and estimated annual net savings of more than $74,000 per machine. | 中 | SO026 |
| CO034 | The company consistently frames labor scarcity and infrastructure build-outs as the reason autonomy is timely. | 中 | SO003, SO009, SO015 |
| CO035 | Independent coverage notes Gravis still has to prove its retrofit approach can scale beyond pilots and beat OEM and startup rivals. | 中 | SO009, SO013 |
| CO036 | Public materials do not disclose revenue, gross margin, customer count, board composition beyond Marco Hutter, or cash balance. | 中 | SO001, SO003, SO013 |
| CM001 | Gravis’ practical market is autonomous and semi-autonomous earthmoving on active construction sites rather than the entire robotics market. | 中 | SM001, SM003 |
| CM002 | The company’s retrofit approach places it in the aftermarket autonomy layer rather than in the new-machine OEM market. | 中 | SM003, SM004 |
| CM003 | Gravis sits adjacent to machine-control and telematics workflows because its product translates plans, sensor data, and progress information into machine behavior. | 中 | SM003, SM004 |
| CM004 | Equipment rental and fleet-upgrade channels matter because retrofit economics work best when contractors can modernize machines already in circulation. | 中 | SM002, SM001 |
| CM005 | OEM autonomy programs from Caterpillar, Komatsu, Volvo, Hitachi, and Develon are substitutes for the same buyer problem even when their routes to market differ from retrofit vendors. | 中 | SM025, SM001 |
| CM006 | Mining and haulage autonomy are adjacent markets that validate off-road autonomy demand but do not fully solve construction’s dynamic worksite problem. | 中 | SM025, SM001 |
| CM007 | Fortune Business Insights projects the global construction equipment market to grow from $183.27 billion in 2026 to $310.24 billion in 2034. | 中 | SM010 |
| CM008 | Global Market Insights pegs the construction equipment market at $167 billion in 2025 and $289.5 billion by 2035, illustrating estimate dispersion but similar order of magnitude. | 中 | SM011 |
| CM009 | Future Market Insights estimates the smart construction equipment segment at $24.4 billion in 2025 and $81.5 billion by 2035. | 中 | SM012 |
| CM010 | Mordor Intelligence estimates the construction robots market at $442.49 million in 2025 and $909.53 million by 2030. | 中 | SM013 |
| CM011 | The market evidence supports a large underlying equipment base but a much smaller near-term wedge for autonomy-specific spend. | 中 | SM010, SM013 |
| CM012 | Gravis’ public narrative points to labor shortage and infrastructure backlog rather than a discrete published TAM as the immediate demand driver. | 中 | SM001, SM002 |
| CM013 | Construction spending remains large enough to support autonomy experimentation because the U.S. Census still tracks a massive ongoing construction outlay base. | 中 | SM014 |
| CM014 | No accessible public source cleanly isolates autonomous earthmoving retrofit as a standalone market line item. | 中 | SM010, SM011 |
| CM015 | General contractors are the main economic buyer because they own schedule risk and can justify productivity tools that compress project duration. | 中 | SM005, SM002 |
| CM016 | Earthmoving subcontractors are a primary user segment because repetitive excavation, trenching, and truck loading are the first public Gravis use cases. | 中 | SM001, SM002 |
| CM017 | Large infrastructure, quarry, energy, and industrial builders are attractive early adopters because Gravis’ proof points cluster around heavy site-prep and materials workflows. | 中 | SM002, SM001 |
| CM018 | Heavy civil contractors matter because Gravis positions itself around large-scale earthmoving, infrastructure, and site-prep workflows. | 中 | SM001, SM002 |
| CM019 | Equipment rental companies are strategic channels because Gravis has already tied product distribution to Flannery’s plant-hire model. | 中 | SM001, SM002 |
| CM020 | Developers and owners influence demand indirectly by rewarding contractors that can finish housing, factory, energy-grid, transit, and data-center projects faster. | 中 | SM001, SM002 |
| CM021 | AGC reported that 92% of contractors had a hard time filling open positions in its 2025 workforce survey. | 中 | SM015 |
| CM022 | ABC said the construction industry needed to attract nearly 440,000 new workers in 2025 to meet expected demand. | 中 | SM006 |
| CM023 | CDC, BLS, and OSHA all reinforce that construction remains a high-risk operating environment with persistent safety pressure. | 中 | SM009, SM007, SM008 |
| CM024 | Safety pressure strengthens the value proposition for automation even before productivity gains are counted. | 中 | SM008, SM009 |
| CM025 | Gravis’ public materials tie demand to housing, energy grids, transit, climate-resilient infrastructure, and data centers. | 中 | SM001, SM002 |
| CM026 | Public deployment reporting suggests repetitive trenching, truck loading, stockpile work, and bulk excavation are easier early wedges than highly variable multi-trade building tasks. | 中 | SM001, SM003 |
| CM027 | Estimate dispersion across market-research firms means valuation work should use multiple lenses instead of one headline TAM number. | 中 | SM010, SM011, SM013 |
| CM028 | Because construction jobsites are temporary, autonomy systems that avoid heavy site-infrastructure requirements have an adoption advantage. | 中 | SM003, SM025 |
| CM029 | The most credible near-term market framing is not all construction, but the subset of repetitive earthmoving tasks where autonomy can extend machine hours and reduce operator bottlenecks. | 中 | SM001, SM003 |
| CM030 | Schedule compression is the dominant value proposition because owners increasingly care about time-to-completion for data centers, manufacturing, energy, and infrastructure projects. | 中 | SM001, SM005 |
| CM031 | The market is demand-rich but evidence-poor: buyer pain is well documented, while willingness-to-pay and budget carve-outs for autonomy remain less transparent. | 中 | SM005, SM006, SM021 |
| CM032 | Gravis benefits from a favorable macro backdrop but still has to prove that autonomy ROI beats machine-control software, telematics, extra crews, and staffing workarounds. | 中 | SM003, SM004, SM005 |
| CM033 | Construction autonomy adoption is likely to progress from Copilot-style guidance and supervised workflows toward broader multi-machine orchestration only after safety and trust thresholds are met. | 中 | SM003, SM023, SM024 |
| CM034 | The gap between the broad construction-equipment market and the small construction-robots market implies that autonomy penetration is still early. | 中 | SM013, SM010 |
| CM035 | For Gravis, the relevant serviceable market is probably measured in specialized excavation, quarry, and infrastructure fleets rather than in total global equipment shipments. | 中 | SM001, SM003, SM002 |
| CP001 | Gravis positions itself as a retrofit autonomy layer for heavy construction equipment already in contractor and quarry fleets. | 中 | SP001, SP002 |
| CP002 | Built Robotics currently emphasizes AI-powered tools for solar construction, especially pile-driving workflows, rather than general earthmoving retrofits. | 中 | SP006, SP008 |
| CP003 | Caterpillar is bringing semi-autonomous and autonomous capabilities into construction from a deep OEM and mining-autonomy base. | 中 | SP021, SP022 |
| CP004 | Hexagon competes more from digital workflows, positioning, and autonomy-enabling site systems than from a Gravis-like retrofit excavator program. | 中 | SP023, SP024 |
| CP005 | Pronto.ai focuses on autonomous haulage and off-road vehicle systems, making it adjacent rather than identical to Gravis’ excavator-heavy wedge. | 中 | SP013, SP014 |
| CP006 | Polymath Robotics markets autonomy, retrofits, and safety systems for off-highway vehicles, giving it a platform-level adjacency to Gravis. | 中 | SP016, SP018 |
| CP007 | Gravis’ clearest differentiation is OEM-agnostic retrofit installation across existing excavator and mixed-equipment fleets. | 中 | SP003, SP002 |
| CP008 | Built Robotics demonstrates strong productization in a narrow solar workflow, which reduces direct overlap with Gravis’ broader earthmoving thesis. | 中 | SP006, SP007 |
| CP009 | Caterpillar’s advantage is end-to-end control of the base machine, embedded automation, and dealer support. | 中 | SP021, SP022 |
| CP010 | Hexagon’s advantage is software and workflow integration across construction and mining rather than direct machine retrofits. | 中 | SP023, SP024 |
| CP011 | Pronto’s architecture is proven in off-road haulage, which validates the general autonomy stack but not Gravis’ excavator manipulation challenge. | 中 | SP013, SP014 |
| CP012 | Polymath competes at the autonomy middleware layer and could partner with OEMs or fleet owners without owning a full Gravis-style contractor program. | 中 | SP019, SP017 |
| CP013 | Gravis has not publicly disclosed pricing, which suggests its commercial model is still customized around deployments rather than standardized catalog pricing. | 中 | SP005, SP003 |
| CP014 | Built Robotics sells specialized robotic construction equipment for solar tasks, implying more productized packaging than Gravis’ current mixed-workflow offering. | 中 | SP006, SP008 |
| CP015 | Caterpillar can package autonomy through machine sales, dealer channels, and integrated software services. | 中 | SP021, SP022 |
| CP016 | Hexagon typically monetizes through software, workflow tools, sensors, and enterprise integration rather than through one contractor-specific autonomy kit. | 中 | SP023, SP024 |
| CP017 | Pronto and Polymath both illustrate that autonomy can be sold as a system layer even when the vehicle platform is provided by someone else. | 中 | SP013, SP016 |
| CP018 | Gravis’ moat rests on field data, contractor workflows, mixed-fleet integrations, and installation know-how more than on exclusive machine manufacturing. | 中 | SP002, SP004 |
| CP019 | OEM incumbents remain the most serious competitive threat because they already control the machine platform, service channel, and installed customer base. | 中 | SP021, SP025 |
| CP020 | Built Robotics demonstrates how a construction-automation startup can narrow its scope and become excellent in one repetitive workflow. | 中 | SP006, SP007 |
| CP021 | Platform autonomy players such as Pronto and Polymath show that software-layer competition could intensify even without identical jobsite focus. | 中 | SP013, SP016 |
| CP022 | Hexagon shows that Gravis may also face competition from workflow incumbents that already sit upstream of machine behavior through data and site-control systems. | 中 | SP023, SP024 |
| CP023 | Caterpillar’s three-decade autonomy history means Gravis cannot rely on first-mover rhetoric as a durable defense. | 中 | SP022, SP021 |
| CP024 | Gravis’ strongest competitive wedge is that it attacks existing contractor fleets without asking buyers to re-platform onto a single OEM. | 中 | SP003, SP002 |
| CP025 | The hardest part of Gravis’ product is not driving from A to B but manipulating terrain and material safely around crews, trucks, and changing topography. | 中 | SP002, SP005 |
| CP026 | Built and Gravis share a common autonomy-for-construction narrative, but their public commercial focus has diverged materially. | 中 | SP006, SP002 |
| CP027 | Caterpillar, Komatsu, and Trimble are much larger organizations, which gives them channel reach but can also slow the kind of fast contractor co-development Gravis emphasizes. | 中 | SP021, SP025, SP024 |
| CP028 | Because public pricing is scarce across the category, customer success and deployment proof are currently better competitive signals than list-price comparison. | 中 | SP009, SP003 |
| CP029 | Gravis’ latest public differentiation claims are grounded in mixed-fleet excavation evidence rather than in abstract autonomy rhetoric. | 中 | SP003, SP005 |
| CP030 | Teleo’s messaging shows another route into the same labor-constrained market: one operator supervising multiple machines in safer conditions. | 中 | SP009, SP011 |
| CP031 | The category remains fragmented enough that Gravis can matter without being the only autonomy vendor in off-road environments. | 中 | SP009, SP013, SP016 |
| CP032 | If OEMs improve quickly or offer low-cost autonomy bundles, Gravis’ retrofit advantage could narrow. | 中 | SP005, SP022 |
| CP033 | If Gravis converts partner testing into repeatable programs, its field-data loop could become a more durable moat than static feature checklists. | 中 | SP004, SP002 |
| CP034 | Competitive success likely depends on owning the repetitive-work wedge before broader autonomy platforms converge on the same contractor accounts. | 中 | SP003, SP009, SP016 |
| CP035 | No public evidence suggests Gravis has exclusive OEM partnerships today, so interoperability remains a strength and a risk at the same time. | 中 | SP003, SP005 |
| CI001 | Public materials imply Gravis monetizes through customer deployments on heavy equipment rather than through consumer software or new-machine sales. | 中 | SI003, SI004 |
| CI002 | Because Gravis retrofits existing fleets, upfront deployment and installation services are a likely revenue component. | 中 | SI004, SI003 |
| CI003 | Recurring software, monitoring, and support subscriptions are plausible follow-on revenue streams once machines are active on site. | 中 | SI001, SI002 |
| CI004 | Professional services tied to site setup, workflow tuning, and customer success are likely important while the product remains deployment-intensive. | 中 | SI002, SI001 |
| CI005 | Multi-machine supervision and future orchestration could become a higher-margin software layer if Gravis advances beyond single-machine tasks. | 中 | SI004, SI001 |
| CI006 | Gravis has not publicly disclosed pricing or contract structure. | 中 | SI003, SI001 |
| CI007 | The current commercial motion looks customized around pilots, channel partners, and deployments rather than around standardized SaaS list pricing. | 中 | SI002, SI007 |
| CI008 | A retrofit model gives Gravis flexibility to price around machine count, site scope, upgrade path, and support intensity. | 中 | SI004, SI002 |
| CI009 | Because Gravis is still building customer proof, pricing likely needs to clear against labor savings, schedule compression, and safety improvement rather than against a software seat metric. | 中 | SI009, SI001 |
| CI010 | The lack of public pricing increases diligence risk because customers may view autonomy as capex, software, or an outsourced service depending on contract form. | 中 | SI007, SI001 |
| CI011 | Hardware on the machine includes sensors, compute, and installation labor, making Gravis more capital intensive than pure software vendors. | 中 | SI004, SI012 |
| CI012 | Field deployments require operations staff and customer success support, which likely depress near-term gross margins. | 中 | SI005, SI002 |
| CI013 | Machine uptime, operator handoff efficiency, and deployment repetition are likely the most important drivers of contribution margin. | 中 | SI004, SI002 |
| CI014 | Because Gravis still emphasizes supervised deployments and copilot workflows, labor savings must currently be shared between the product and human oversight layers. | 中 | SI001, SI007 |
| CI015 | Gravis’ best unit-economics scenario likely comes from repeat deployments on similar excavation, trenching, and quarry workflows rather than one-off bespoke jobsites. | 中 | SI020, SI019 |
| CI016 | Gravis announced a $200 million Series A on 2026-08-17. | 中 | SI001, SI007, SI008 |
| CI017 | The public capital base disclosed across the 2025 and 2026 rounds totals at least about $223 million. | 中 | SI001, SI002 |
| CI018 | The company announced a $23 million financing in November 2025 to expand in the UK, US, and EU. | 中 | SI002, SI011, SI010 |
| CI019 | The rapid sequence from a $23 million growth round to a $200 million Series A suggests investors expect capital-intensive scale-up rather than a lightly funded software rollout. | 中 | SI002, SI001 |
| CI020 | A retrofit autonomy business likely needs large capital reserves for hardware inventory, field operations, safety validation, and customer support. | 中 | SI004, SI005, SI020 |
| CI021 | Gravis does not publicly disclose revenue run-rate. | 中 | SI006 |
| CI022 | Gravis does not publicly disclose gross margin or contribution margin. | 中 | SI006, SI007 |
| CI023 | Gravis does not publicly disclose customer count or ARR. | 中 | SI006, SI003 |
| CI024 | Gravis does not publicly disclose company-wide burn rate or cash on hand. | 中 | SI001, SI005 |
| CI025 | The absence of audited financial statements means investors cannot independently verify runway or cash conversion. | 中 | SI015 |
| CI026 | The most plausible near-term model is a blend of deployment revenue and recurring software-like revenue layered onto active machines. | 中 | SI004, SI001 |
| CI027 | Gravis’ public proof points are still too early to support a strong revenue-multiple framework. | 中 | SI006, SI007 |
| CI028 | Compared with pure software startups, Gravis likely trades lower gross-margin potential for a larger operational ROI if it succeeds on site. | 中 | SI015, SI004 |
| CI029 | The company’s financing pace reduces short-term solvency risk but raises the bar for disciplined capital deployment. | 中 | SI001, SI007 |
| CI030 | Investor diversity across strategic backers, sector VCs, and now SoftBank suggests Gravis can likely raise follow-on capital if technical progress continues. | 中 | SI013, SI001, SI009 |
| CI031 | The biggest financial diligence question is not whether Gravis can fund pilots today, but whether pilots convert into repeatable, profitable deployment programs. | 中 | SI007, SI001, SI020 |
| CI032 | Because the company emphasizes 24/7 productivity and schedule compression, its ROI case likely improves most on labor-constrained, high-urgency jobsites. | 中 | SI018, SI001 |
| CI033 | Custom installation and support work can create strong customer value while also slowing the path to software-like margins. | 中 | SI004, SI005 |
| CI034 | Without public renewal, expansion, or deployment-cohort data, revenue durability remains unproven. | 中 | SI006, SI003 |
| CI035 | A useful underwriting frame is capital adequacy plus conversion evidence, not headline valuation alone. | 中 | SI001, SI002, SI015 |
| CE001 | Gravis Rack is a retrofit sensor, compute, and controls system for existing heavy construction equipment. | 中 | SE005, SE001 |
| CE002 | The public hardware stack includes LiDAR, GNSS RTK, cameras, machine telemetry, and onboard automotive-grade edge compute. | 中 | SE005, SE011, SE012 |
| CE003 | Gravis highlights real-time surveying, hazard mapping, and terrain visualization as part of the product value proposition. | 中 | SE005, SE006 |
| CE004 | The product strategy depends on working across existing contractor fleets rather than only on one machine platform. | 中 | SE003, SE001 |
| CE005 | Gravis markets the system as compatible across more than a dozen brands and machine classes. | 中 | SE003, SE013 |
| CE006 | The clearest public use cases are trenching, bulk excavation, truck loading, stockpile management, and grading-adjacent earthmoving. | 中 | SE005, SE010, SE012 |
| CE007 | Public partner and customer quotes emphasize repetitive earthmoving as a workflow where autonomy can free skilled operators for harder tasks. | 中 | SE004, SE010 |
| CE008 | The product is designed to integrate with existing jobsite workflows instead of forcing a wholly new operating model. | 中 | SE006, SE005 |
| CE009 | Gravis frames the operator role as supervisory, exception-handling, and optional manual takeover rather than as fully absent today. | 中 | SE003, SE009 |
| CE010 | A likely expansion path is from one repetitive task to broader multi-machine and multi-workflow coordination. | 中 | SE010, SE003 |
| CE011 | Gravis explicitly describes simulation-trained machine learning as central to its autonomy system. | 中 | SE003, SE013 |
| CE012 | The founding thesis is that robot-control methods proven in research can be adapted to heavy equipment that changes the terrain as it works. | 中 | SE003, SE012 |
| CE013 | Environmental understanding is a core technical requirement because the machine must interpret soil, slope, buried utilities, trucks, and people in real time. | 中 | SE006, SE011, SE009 |
| CE014 | Gravis’ architecture blends onboard sensing and compute with tablet-based supervision and remote visibility rather than relying only on cloud control. | 中 | SE006, SE009 |
| CE015 | The hardest technical challenge is not simple navigation but precise earth shaping in dynamic environments around people, trucks, and changing ground conditions. | 中 | SE012, SE014 |
| CE016 | Gravis repeatedly markets the system around safety improvement, people detection, and work-zone awareness. | 中 | SE003, SE008 |
| CE017 | Public safety workflows still include emergency stop, safety boundaries, and human supervision, which implies a defense-in-depth posture rather than pure unattended autonomy. | 中 | SE009, SE008 |
| CE018 | The company’s public deployment model is still supervised, which is itself a quality and trust control while full autonomy matures. | 中 | SE003, SE010 |
| CE019 | Hitachi’s 2026 CONEXPO announcement suggests Gravis’ technology is mature enough for live public demos with mainstream OEM equipment. | 中 | SE008, SE009 |
| CE020 | The Equipment World walkthrough shows the interface supports CAD imports, live terrain coloring, AR overlays, and bucket-defined excavation zones. | 中 | SE009, SE006 |
| CE021 | Public proof is strongest for supervised autonomy on excavation tasks, not for broad multi-machine autonomous sites without oversight. | 中 | SE010, SE014 |
| CE022 | Copilot is positioned as the commercial bridge product that gives contractors immediate machine guidance while keeping the fleet autonomy-ready. | 中 | SE010, SE006 |
| CE023 | The product is more mature on repetitive excavation and loading than on generalized construction autonomy. | 中 | SE003, SE023 |
| CE024 | Real-world generalization across sites and machines is a central technical hurdle inherited from the mixed-fleet retrofit thesis. | 中 | SE003, SE013 |
| CE025 | Retrofit installation is strategically important because it removes the need for customers to wait for OEM roadmaps. | 中 | SE005, SE008 |
| CE026 | The system’s value proposition combines safety, schedule compression, operator augmentation, data capture, and progress visibility rather than only autonomous driving. | 中 | SE001, SE006, SE005 |
| CE027 | Product-market fit appears strongest where the same loading or trenching pattern repeats for long hours on large sites. | 中 | SE010, SE003 |
| CE028 | Gravis’ public architecture claims emphasize simulation, sensor fusion, and learning-based control more than classical rule-based robotics alone. | 中 | SE013, SE012 |
| CE029 | The ability to move between copilot, remote orchestration, and autonomous task execution lowers buyer anxiety about adoption. | 中 | SE006, SE009 |
| CE030 | A durable advantage would come from compounding labeled field data and contractor-specific workflow knowledge across many sites and machines. | 中 | SE005, SE004 |
| CE031 | The current product still depends on human oversight, so safety claims are stronger for assisted-supervised autonomy than for unattended fleet operation. | 中 | SE014, SE009 |
| CE032 | Competitor materials such as Teleo, Pronto, and Polymath show that the broader category is also converging on retrofit-friendly autonomy stacks and trust surfaces. | 中 | SE015, SE018, SE020 |
| CE033 | Gravis’ architecture must work with changing terrain and temporary infrastructure, which makes deployment engineering a core product feature, not a side service. | 中 | SE005, SE024 |
| CE034 | The strongest near-term product narrative is automation that fits today’s crews and fleets, not fully unmanned greenfield jobsites. | 中 | SE006, SE005, SE003 |
| CE035 | No public whitepaper, trust center, or formal certification library was visible in the reviewed Gravis materials, leaving diligence gaps on documented assurance processes. | 中 | SE001, SE003 |
| CU001 | Large contractors and infrastructure builders are Gravis’ clearest customer segment because named partners such as Taylor Woodrow, Techint, Boskalis, and Holcim run large site-prep programs. | 中 | SU001, SU009, SU023 |
| CU002 | Earthmoving-heavy operators are strong early adopters because repetitive excavation, trenching, loading, and stockpile work are Gravis’ public sweet spots. | 中 | SU002, SU009 |
| CU003 | Materials and aggregates operators such as Holcim matter because they link heavy-equipment operations with repetitive loading and quarry workflows. | 中 | SU007, SU008 |
| CU004 | Rental and plant-hire channels are real go-to-market paths because Gravis has already tied autonomy distribution to Flannery Plant Hire. | 中 | SU001, SU011, SU015 |
| CU005 | Large OEM and dealer relationships matter because mixed-fleet retrofit adoption still benefits from machine-maker support and open interfaces. | 中 | SU025, SU020 |
| CU006 | By late 2025 Gravis said it was live in seven countries across the UK, EU, US, Latin America, and Asia. | 中 | SU010, SU009 |
| CU007 | By August 2026 the company was describing systems deployed across four continents with global infrastructure leaders. | 中 | SU002, SU023, SU024 |
| CU008 | Public proof includes a Taylor Woodrow autonomous excavator trial and planned Manchester Airport deployment in the UK. | 中 | SU004, SU006 |
| CU009 | Public proof also includes a 60-mile autonomous pipeline project in Argentina with Techint Group. | 中 | SU009, SU017 |
| CU010 | Holcim is both an investor and a public quarry/construction user, making it one of Gravis’ strongest named proof points. | 中 | SU007, SU013 |
| CU011 | The adoption story still centers on supervised deployments, partner programs, and channel expansion rather than on a large installed base of disclosed paying accounts. | 中 | SU001, SU009 |
| CU012 | Taylor Woodrow publicly emphasized productivity, safety, and reduced rework as reasons to pilot the autonomous excavator. | 中 | SU004, SU006 |
| CU013 | Techint publicly praised Gravis’ field presence, feature responsiveness, and tough-ground performance during the Argentina pipeline project. | 中 | SU009, SU017 |
| CU014 | Holcim publicly framed Gravis as a way to increase output consistency, improve safety, and optimize machine selection across quarry operations. | 中 | SU007, SU013, SU014 |
| CU015 | Gravis’ public customer proof remains quote-based and deployment-based rather than revenue-based. | 中 | SU022, SU001 |
| CU016 | No public source discloses renewal rate, churn, GRR, or NRR for Gravis. | 中 | SU021, SU022 |
| CU017 | Repeat deployment across multiple geographies and customer archetypes is the best visible proxy for early customer satisfaction. | 中 | SU010, SU009 |
| CU018 | Gravis’ partner expansion suggests customer references are helping it win additional pilot and production contexts even without public ARR metrics. | 中 | SU010, SU001 |
| CU019 | Because the current deployments are operationally intensive, customer satisfaction likely depends heavily on field support quality. | 中 | SU009, SU001 |
| CU020 | The absence of public multi-year cohort data means durability of customer relationships remains unproven. | 中 | SU021, SU022 |
| CU021 | Customer concentration risk is likely high today because the publicly named account set is still small. | 中 | SU022, SU021 |
| CU022 | Gravis appears best suited to large, repetitive projects, which could narrow the customer base even as deal size rises. | 中 | SU002, SU001 |
| CU023 | Data-center, factory, energy, quarry, and infrastructure buildouts are attractive end markets because owners care intensely about schedule compression. | 中 | SU002, SU009 |
| CU024 | If Gravis sells mostly to large contractors and industrial operators, enterprise adoption could be powerful but procurement cycles may also be slow. | 中 | SU021, SU022 |
| CU025 | Rental channels can reduce concentration risk over time if Gravis proves interoperability and ROI on mixed fleets. | 中 | SU011, SU012, SU015 |
| CU026 | Gravis’ current customer strategy is better described as co-development with lead partners than as broad-market sales coverage. | 中 | SU001, SU009 |
| CU027 | The best customer proof is operational rather than brand-based: live jobsites, earth moved, and quotes about workflow relief and consistency. | 中 | SU009, SU004 |
| CU028 | Because construction adoption is conservative, named customer advocates are more valuable than abstract claims about a giant TAM. | 中 | SU007, SU004, SU009 |
| CU029 | Partner quotes repeatedly emphasize freeing scarce skilled operators for higher-value work rather than removing humans entirely. | 中 | SU009, SU004 |
| CU030 | The company’s strongest early demand likely comes from labor-constrained, large-scale site prep, pipeline, quarry, and excavation programs rather than from all construction categories. | 中 | SU002, SU008, SU009 |
| CU031 | Flannery shows how large the eventual channel opportunity could be if autonomy-ready fleets become rentable at scale. | 中 | SU011, SU015 |
| CU032 | The CAM Pathfinder project is particularly relevant because it connects rental distribution, government funding, and repeatable earthmoving workflows. | 中 | SU011, SU012, SU019 |
| CU033 | Commercial adoption risk remains meaningful because no public source yet shows repeat revenue or standardized deployment conversion across customers. | 中 | SU022, SU021 |
| CU034 | Customer expansion will likely depend on how quickly Gravis can move from closely supported pilots to repeatable operating programs. | 中 | SU001, SU009 |
| CU035 | A slow-moving construction market can still support Gravis if each successful reference account unlocks adjacent contractors, operators, or project owners. | 中 | SU004, SU009, SU007 |
| CR001 | OSHA maintains dedicated robotics guidance because robot systems create distinctive workplace hazards that require formal hazard recognition and evaluation. | 中 | SR009, SR010 |
| CR002 | CDC and BLS both show construction remains a dangerous industry, which raises the evidentiary bar for any autonomous-equipment safety claim. | 中 | SR013, SR012 |
| CR003 | Frontiers’ construction-robotics review says automation can improve productivity and safety while also introducing new mechanical and psychosocial risks. | 中 | SR014 |
| CR004 | ILO argues that AI and digitalization can reduce hazards but also create new oversight, ergonomics, and worker-protection risks. | 中 | SR015 |
| CR005 | Because Gravis operates around heavy machinery, legal and insurance scrutiny will likely increase before fully operator-less deployments scale broadly. | 中 | SR020, SR023 |
| CR006 | Dynamic terrain, dust, occlusion, and changing work zones are core operational risks for Gravis’ perception and planning stack. | 中 | SR002, SR006 |
| CR007 | The company’s strongest public proof still uses supervised autonomy and copilot workflows, which indicates technical and operational guardrails remain important. | 中 | SR004, SR007 |
| CR008 | OSHA’s robotics manual emphasizes that hazard recognition must be followed by engineered controls and operating procedures, not just awareness. | 中 | SR010, SR011 |
| CR009 | Construction sites punish brittle setup assumptions because network, calibration, and workflow conditions change rapidly from one site to another. | 中 | SR002, SR019 |
| CR010 | A supervised deployment can still fail commercially if support burden and exception handling stay too high. | 中 | SR004, SR005 |
| CR011 | Gravis depends heavily on contractor, OEM, and channel partners for field data, workflow learning, and reference quality. | 中 | SR007, SR029, SR030 |
| CR012 | If a few partners dominate deployment learning, roadmap concentration can become a hidden strategic dependency. | 中 | SR030, SR007 |
| CR013 | OEMs remain external dependencies because retrofit autonomy has to coexist with machine interfaces, warranties, and service realities not controlled by Gravis. | 中 | SR029, SR004 |
| CR014 | Temporary-site execution means field operations are part of the product, increasing dependency on a high-quality deployment team. | 中 | SR005, SR003 |
| CR015 | Capital markets are also a dependency because a hardware-plus-software autonomy company can burn cash faster than a pure software startup. | 中 | SR025, SR026 |
| CR016 | Ryan Luke Johns and Dominic Jud are key-person risks because Gravis’ public identity is tightly bound to founder credibility. | 中 | SR004, SR001 |
| CR017 | The company is young enough that leadership depth below the founders is still developing even as hiring accelerates. | 中 | SR003, SR001 |
| CR018 | St. Louis Fed and Brookings both highlight labor-market dislocation risk around automation, which can create workforce resistance to adoption. | 中 | SR017, SR016 |
| CR019 | RICS highlights AI governance, data quality, and accountability as wicked problems in construction, which maps directly to Gravis’ execution risk. | 中 | SR018 |
| CR020 | A startup can have strong technology and still fail if customer education, training, and change management lag behind engineering progress. | 中 | SR005, SR028 |
| CR021 | Supervised deployment is currently a mitigation because it keeps humans in the loop while Gravis gathers real-world evidence. | 中 | SR004, SR007 |
| CR022 | Retrofit reversibility and manual takeover are mitigations because customers can return machines to human control if needed. | 中 | SR002, SR002 |
| CR023 | Partner co-development is a mitigation because it exposes the product to real workflows before broad commercialization. | 中 | SR005, SR030 |
| CR024 | A true stop condition would be repeated safety incidents or failure to move from supervised to lower-touch deployments on schedule. | 中 | SR006, SR004 |
| CR025 | Another stop condition would be if OEMs or workflow incumbents close the product gap faster than Gravis can scale customer proof. | 中 | SR006, SR029 |
| CR026 | Construction autonomy creates a paradox: the labor and safety crisis makes automation attractive, but the same risk intensity makes customer proof harder to earn. | 中 | SR028, SR012 |
| CR027 | Publicly disclosed deployment success does not eliminate the long tail of rare but serious edge cases that regulators and customers will care about. | 中 | SR007, SR006 |
| CR028 | Gravis’ biggest technical risk is not that autonomy is impossible, but that robust operation on messy temporary sites may take longer than investors expect. | 中 | SR002, SR014 |
| CR029 | Gravis’ biggest commercial risk is that customers continue to like pilots but hesitate to operationalize them at scale. | 中 | SR004, SR006 |
| CR030 | Worker-acceptance risk should not be ignored because automation can be framed as both a safety tool and a labor substitute. | 中 | SR017, SR016 |
| CR031 | Insurance and liability frameworks may evolve more slowly than the technology itself, delaying large-scale unattended deployment. | 中 | SR021, SR023 |
| CR032 | Because Gravis is privately held, outsiders cannot yet observe whether internal safety culture scales as quickly as deployment ambition. | 中 | SR001, SR004 |
| CR033 | The company’s strongest mitigation is learning speed on live jobsites, but that only works if incidents stay low and partner trust stays high. | 中 | SR005, SR007 |
| CR034 | A downturn in construction demand or funding appetite could amplify technical and customer risks by stretching deployment payback periods. | 中 | SR028, SR025 |
| CR035 | Overall risk is high but not fatal: the company is attacking a hard, painful problem with credible talent, yet still has to prove safe scalable execution. | 中 | SR004, SR006, SR007 |
| CR036 | Gravis publishes product and marketing materials, but public legal documents do not yet explain how autonomous-equipment liability is allocated in commercial contracts. | 中 | SR023, SR020 |
| CR037 | BLS injury and fatality datasets reinforce that construction hazard monitoring is continuous and nationally visible, increasing reputational consequences of any incident. | 中 | SR012, SR008 |
| CR038 | The NIOSH and HSE excavation materials frame worker-centered design and excavation discipline as essential to safe automation adoption in construction. | 中 | SR019, SR011 |
| CR039 | Gravis’ hiring posture suggests the company is still building the organizational depth needed for safe multi-site scale. | 中 | SR003, SR001 |
| CR040 | Public legal and safety context remains ahead of Gravis’ disclosed contract framework, which is a meaningful governance gap before unattended deployments. | 中 | SR020, SR023, SR001 |
| CV001 | Gravis Robotics’ reported $1 billion post-money valuation and $200 million SoftBank-led Series A are real and well corroborated, but public commercialization evidence remains thinner than the headline implies. | 高 | SV002, SV006, SV008 |
| CV002 | The company addresses a painful market problem—labor scarcity, safety pressure, and productivity drag in heavy construction—that is large enough to support a meaningful upside case if execution works. | 中 | SV018, SV019, SV020 |
| CV003 | Public product proof is credible but still concentrated in supervised or tightly managed autonomy settings rather than broad unattended mixed-fleet operations. | 中 | SV004, SV005, SV016 |
| CV004 | Public financial disclosure is not strong enough to justify a precision valuation model for Gravis today. | 中 | SV002, SV003 |
| CV005 | The right public-evidence recommendation is research-more rather than an outright bullish underwriting call. | 中 | SV006, SV002, SV007 |
| CV006 | Valuation stance is stretched because Gravis cleared unicorn status before public revenue, retention, or margin evidence became visible. | 中 | SV006, SV002, SV009 |
| CV007 | Thesis: Gravis could become the leading retrofit autonomy layer for mixed-fleet heavy equipment if it turns pilots and flagship accounts into repeatable programs. | 中 | SV004, SV014, SV015 |
| CV008 | Thesis: the retrofit model can create strong ROI for customers because it targets installed fleets instead of forcing new machine purchases. | 中 | SV004, SV017, SV006 |
| CV009 | Thesis: SoftBank’s $200 million endorsement plus visible customer and OEM proof gives Gravis unusual credibility for such a young construction-robotics company. | 中 | SV002, SV026, SV016 |
| CV010 | Anti-thesis: Gravis may remain a well-funded but operationally heavy deployment business rather than a scalable software-led platform. | 中 | SV003, SV006, SV007 |
| CV011 | Anti-thesis: OEM incumbents and adjacent autonomy vendors can compress Gravis’ wedge by bundling control, data, and service into broader equipment or workflow offerings. | 中 | SV021, SV022, SV023, SV024 |
| CV012 | Anti-thesis: the $1 billion mark may already discount much of the upside before public economics are available. | 中 | SV002, SV006, SV007 |
| CV013 | Bull case requires repeat multi-site programs, lower-touch deployments, and evidence that autonomy performance generalizes across brands and use cases. | 中 | SV004, SV014, SV015 |
| CV014 | Base case assumes Gravis wins a useful niche in retrofit autonomy with continued strategic backing but still carries meaningful field-service weight. | 中 | SV003, SV009, SV026 |
| CV015 | Bear case assumes customers like the technology yet scale adoption more slowly than investors expect, making the current valuation look early. | 中 | SV006, SV007, SV025 |
| CV016 | In the bull case, Gravis could earn premium platform status because retrofit interoperability across incumbent fleets would matter more than manufacturing ownership. | 中 | SV004, SV005, SV014 |
| CV017 | In the bear case, the company could still retain strategic or acquisition value without delivering venture-scale returns from the current entry price. | 中 | SV025, SV021 |
| CV018 | Built Robotics is a useful workflow-focused startup comparable, but its solar concentration and product evolution make it an imperfect analog for Gravis’ broader heavy-equipment thesis. | 中 | SV025, SV006 |
| CV019 | Caterpillar and Komatsu are relevant autonomy benchmarks, but their public-company OEM scale and disclosure make their valuation frameworks incomparable to Gravis. | 中 | SV021, SV022 |
| CV020 | Hexagon and Trimble are useful workflow-software adjacencies, but they compete from positioning, workflow, and site-control layers rather than from full retrofit autonomy. | 中 | SV023, SV024 |
| CV021 | Teleo and similar remote-operation startups validate buyer appetite for modernizing legacy fleets, even if their operating model differs from Gravis’ autonomy ambition. | 中 | SV030, SV004 |
| CV022 | SoftBank’s involvement signals ambition and access to capital, not proof that Gravis has already solved commercialization or unit economics. | 中 | SV026, SV002 |
| CV023 | The cleanest comparable set is therefore archetypal rather than statistical: startup construction automation, private retrofit autonomy, OEM incumbent, workflow software incumbent, and capital-provider benchmark. | 中 | SV025, SV021, SV023, SV027 |
| CV024 | A thesis-break trigger would be safety incidents or reliability failures that reduce customer and insurer trust materially. | 中 | SV006, SV002, SV003 |
| CV025 | Another thesis-break trigger would be weak pilot-to-program conversion or expansion beyond a small set of reference accounts. | 中 | SV014, SV015, SV003 |
| CV026 | Another thesis-break trigger would be faster-than-expected convergence from OEMs or adjacent autonomy platforms. | 中 | SV021, SV022, SV023 |
| CV027 | The first diligence ask is revenue, gross margin, and deployment-cohort data that can tie the current valuation to commercial reality. | 中 | SV002, SV006 |
| CV028 | The second diligence ask is safety, liability, and insurance documentation that can show how autonomous-equipment risk is contractually governed. | 中 | SV002, SV026 |
| CV029 | The third diligence ask is a roadmap showing how Gravis moves from supervised deployments to lower-touch autonomy with better software leverage. | 中 | SV004, SV005, SV006 |
| CV030 | The fourth diligence ask is cap-table, preference, and concentration detail, especially given the size and concentration of the SoftBank round. | 中 | SV002, SV026, SV028 |
| CV031 | The current valuation can still work for new investors if Gravis compounds proof quickly, but the margin for execution error is already thin. | 中 | SV006, SV002, SV007 |
| CV032 | Gravis’ upside is asymmetrical to the positive because a successful autonomy layer for existing heavy-equipment fleets could capture large workflow value without building new machines. | 中 | SV004, SV019, SV020 |
| CV033 | Gravis’ downside is also real because deployment-heavy economics can produce an operationally valuable business that still struggles to justify venture-scale software multiples. | 中 | SV003, SV006, SV025 |
| CV034 | Scenario analysis is more honest than multiples analysis at this stage because too many core metrics remain private. | 中 | SV002, SV021, SV022 |
| CV035 | A medium-confidence recommendation is appropriate because Gravis’ strategic logic is strong while its commercial and financial evidence remains incomplete. | 中 | SV002, SV006, SV026 |
| CV036 | Public filings and investor-relations materials from Caterpillar, Komatsu, and SoftBank are useful reminders of how much disclosure, scale, and balance-sheet depth separate Gravis from mature incumbents and sponsors. | 高 | SV021, SV022, SV026 |
| CV037 | Growth-investor benchmark sources such as 8VC, CapitalG, and Georgian show the kind of scaling posture growth capital celebrates, but investor prestige is not a substitute for unit-economics proof. | 中 | SV027, SV028, SV029 |
| CV038 | If Gravis executes well, SoftBank support may accelerate hiring, global expansion, and partner access; if execution slips, cap-table prestige will not protect valuation. | 中 | SV026, SV001, SV002 |
| CV039 | The valuation debate is therefore less about whether Gravis is interesting and more about whether today’s entry price leaves enough upside for new capital. | 中 | SV006, SV002, SV007 |
| CV040 | Until commercial cohorts are visible, downside protection comes more from disciplined milestones and entry terms than from comparative multiples. | 中 | SV002, SV026, SV021 |