初创公司尽调
尽调报告 Defense Technology / AI Autonomy Late-stage private / pre-IPO 2026-05-11

Shield AI

AI 自主能力规模化:尽职调查报告

Shield AI 是国防 AI 自主平台的领先公司,增长极快;但约 47x 收入倍数已经把 X-BAT 和 Hivemind 企业级扩张的大量执行风险计入价格。

封面要素

最近一轮融资 01
$2B Series G [CV001, CO013]
2024 财年营收 03
267 USD M [CV006]
营收同比增速 04
64 % [CV006]
2026 财年指引 05
>$540M [CV007]
已部署国家数 06
10+ [CU012]
成立时间 08
2015 [CO001]

公司概况

Shield AI 由 Ryan Tseng(CEO,美国海军退役老兵)及其兄弟 Brandon Tseng(总裁,海豹突击队退役老兵)和 Andrew Reiter(CTO)于 2015 年联合创立,总部位于圣地亚哥,是一家开发 AI 自主能力软件和自主无人机系统的国防科技公司。核心产品 Hivemind 是一个多层 AI 企业平台,包含 EdgeOS(实时中间件)、Pilot(自主飞行 AI)、Commander(多智能体协同)和 Forge(ML 模型开发)四个模块。公司同时制造 V-BAT 第三组 VTOL 无人机,并正在研发 X-BAT 第五组超音速自主飞机。2026 年 3 月,Shield AI 完成 $2B G 轮融资,融资后估值达 $12.7B,被业界广泛视为 IPO 前领先的国防 AI 公司。

官网
shield.ai
成立时间
2015-01-01
创始人
Ryan Tseng, Brandon Tseng, Andrew Reiter
创立地点
San Diego, CA
总部
San Diego, CA
产品
Hivemind 企业级 AI 平台(EdgeOS 中间件、Pilot 自主飞行、Commander 多智能体协同、Forge ML 工具链及国防基础模型),以及 V-BAT 第三组 VTOL 无人机(单价约 $1M)和 X-BAT 第五组超音速自主飞机(目标单价约 $27M,计划 2026 年秋首飞)。另于 2026 年 3 月收购仿真软件公司 Aechelon Technology。
客户
美国国防部(USAF、USN、USCG、SOCOM)、盟国政府及主要国防承包商。
商业模式
平台软件授权 + 无人机硬件销售 + 国防主承包商项目级营收。
阶段
Late-stage private / pre-IPO
融资情况
G 轮 2026 年 3 月:$2B,融资后估值 $12.7B;上一轮 F 轮 2025 年 3 月:$240M,估值 $5.3B。
[CO001, CO013, CO019, CO020, CO032, CV001, CV006, CV011]

执行摘要

主要优势

  • 独特的 AI 自主平台 Hivemind 已在 10 多个国家完成实战部署,并能在无 GPS、无通信环境下运行。
  • USAF A-GRA 协同作战飞机项目唯一主承包商;FY2024 获 USCG $198M IDIQ 和 DoD $170M 资金。
  • 收入增长突出:FY2024 同比增长 64% 至 $267M;管理层指引 FY2026 超过 $540M。
  • X-BAT Group-5 超音速 VTOL 飞机有机会把可服务市场拉高一个台阶。
  • DoD 关系深,退伍军人领导团队具备真实作战人员可信度。

主要风险

  • FY2024 约 47x 收入倍数,即使放在高增长国防科技里也偏贵;一旦增长落空,估值下修风险很大。
  • 公司尚未盈利、烧钱高,X-BAT 研发和 Hivemind 扩张仍需要持续融资。
  • 单一客户集中度高:DoD 贡献约 90%+ 收入;预算削减或项目取消都是生存级风险。
  • 出口管制 / ITAR 限制国际扩张,并带来监管压力。
  • X-BAT 研发执行风险高:首飞目标为 2026 年秋季,目前没有确认的生产合同。
  • Anduril(目标估值约 $60B)、Palantir(上市公司)以及用交叉补贴投入 AI 部门的主承包商持续施压。

未决问题

  • 毛利率和 EBITDA 未公开;软件 / 硬件利润率组合未知。
  • Hivemind 企业业务的净留存率(NRR)和合同续约率未披露。
  • X-BAT 研发成本和量产准备时间表需要 DoD 确认。
  • IPO 时间表和 Series G 清算优先权不清晰;稀释压力无法量化。
  • 国际客户(盟国)拆分和收入占比未公开。

目录

Chapter 01

01公司概况

1.1 公司定位与使命

Shield AI 总部位于加利福尼亚州圣地亚哥,是一家深度科技国防公司,使命是以智能自主系统保护服役人员和平民。公司由 Brandon Tseng(前美国海豹突击队员)、其兄弟 Ryan Tseng(工程师兼企业家)和 AI 技术专家 Andrew Reiter 于 2015 年创立,创立契机源自 Brandon Tseng 在阿富汗作战期间亲眼目睹的战场漏洞——AI 飞行员这一概念由此诞生:能够在 GPS 拒止和通信拒止环境下无需人工操控自主执行任务的软件,可在人类控制受限或反应过慢时挽救生命。 Shield AI 将自身定位不仅是无人机制造商,更是美国国防部及盟军的"自主能力基础设施"运营商。旗舰软件平台 Hivemind Enterprise 让开发者和国防机构能够规模化构建、测试和部署自主任务能力。公司同时制造战术飞机,主要包括 V-BAT VTOL 无人机和新推出的 X-BAT VTOL 打击战斗机,两款产品均以 Hivemind 作为 AI 飞行引擎。Shield AI 目前处于后期私募阶段(G 轮),在圣地亚哥、达拉斯、华盛顿特区、阿布扎比(阿联酋)、基辅(乌克兰)和墨尔本(澳大利亚)设有办公室,反映出其支持美国及盟军现役军事行动的全球化布局。 公司商业模式横跨软件授权(Hivemind Enterprise 部署于第三方平台)、软硬件一体化产品(V-BAT、X-BAT)和政府服务三个层面。Shield AI 将 Hivemind 定位为军事自主能力的"操作系统"——类似 GPS 曾为导航构建的那道持久软件护城河。创始团队始终聚焦使命;Ryan Tseng 曾亲口讲述一个关键早期故事:一位湾区投资人要求公司转向商业市场,他宁可放弃该轮融资也不妥协于国防优先的创业使命。 [CO001, CO005, CO006, CO019, CO020, CO034]

Shield AI 关键指标快照(截至 2026 年 5 月)
指标数值/状态日期置信度差距/注意事项
最新估值(投后)$12.7B2026-03-26私营公司;无独立审计
累计融资总额~$3.52B(17 轮)2026-03-26来源:Tracxn、TechCrunch;早期轮次部分细节不完整
最新一轮融资$2B Series G2026-03-26由 Advent International、JPMorganChase 领投
预测 2026 年营收~$540M2026第三方分析师估算;非公司披露数据
ARR / 收入运行率未公开披露私营公司;无公开申报文件
员工数~1,422 名员工2026-03-26来源:Tracxn;可能不完整(不含承包商及海外雇员)
总部San Diego, CA2015在 San Diego 创立并设总部
阶段Series G(后期私募)2026-03-26截至报告日期暂无 IPO 申报披露
运行 Hivemind 的平台数量15+2026公司自报;未经独立核实

营收及 ARR 均未公开披露;2026 年约 5.4 亿美元营收为第三方分析师估算。员工数来自 Tracxn:2024 年 12 月法律实体备案数据为 757 人,2026 年 3 月 LinkedIn/Tracxn 数据为 1,422 人。估值来源:2026 年 3 月 26 日 Series G 公告。

[CO010, CO013, CO017, CO027]
FO003: Shield AI 关键指标快照(2026 年 5 月)

截至 2026 年 5 月报告时点,Shield AI 关键业绩指标,数据来源于 G 轮融资公告、Tracxn 及分析师估算。

营收数据为第三方分析师估算,非公司披露。员工数来自 Tracxn,截至 2026 年 3 月。估值来自 2026 年 3 月 26 日 G 轮融资新闻稿。

[CO013, CO017, CO032]

1.2 创始人与领导团队

Shield AI 由三位背景互补的联合创始人共同创立,赋予公司少有的组合优势:军事实战信誉、工程深度和 AI 研究能力。Brandon Tseng 毕业于美国海军学院,曾以海豹突击队员身份在阿富汗和太平洋战区执行战斗任务,随后联合创立 Shield AI,现任总裁兼首席增长官,负责全球战略拓展。Ryan Tseng 是工程与商业架构者,从创立之初担任 CEO 至 2025 年 5 月,期间带领 Shield AI 员工规模突破 900 人、累计融资逾 $1B,估值达 $5.3B;2025 年 5 月转任总裁兼首席战略官,专注于美国及国际政策与国防高层关系。Andrew Reiter 是 AI 技术联合创始人,构建了 Hivemind 的初始算法基础。 2025 年 3 月,Shield AI 宣布领导层更迭:Gary Steele 获任 CEO,于 2025 年 5 月 13 日正式上任,此前他曾担任 Cisco (2024 年以 $28B 收购 Splunk)市场侧总裁,同时出任 Shield AI 董事会成员。其履历包括创立 Proofpoint 并将其从零带至 IPO、达成 $12B 企业价值,以及领导 Splunk 实现 58% 营收增长并在 Cisco 收购前完成全年 GAAP 盈利。这一任命标志着 Shield AI 着力扩展企业软件分发能力,推动 Hivemind Enterprise 走向传统国防主承包商以外的更广泛市场。 领导团队其他成员包括:Nathan Michael(CTO)、Kingsley Afemikhe(CFO)、Sharyl Givens(首席人事官)、Bob Harward(国际业务拓展执行副总裁)、Michael Yang(首席法务官)和 Chandra Rangan(首席营销官)。对联合创始人的关键人依赖——尤其是 Brandon Tseng 的军事人脉和 Ryan Tseng 的战略关系——仍是值得关注的重要尽调因素,Steele 加入后形成的职业管理层在一定程度上缓解了这一风险。 [CO002, CO003, CO004, CO007, CO008, CO009]

管理层与创始人概览
姓名职位背景创始人与市场契合度关键人物依赖
Brandon Tseng总裁兼首席增长官(联合创始人)美国海军学院毕业;海豹突击队队员;曾赴阿富汗及太平洋战区执行作战任务深厚的军方关系网络;在 DoD 及海外军方具有高度公信力;主导塑造了早期产品愿景高——军事关系与任务公信力难以复制
Ryan Tseng总裁兼首席战略官(联合创始人)工程师背景;创办 WiPower(后任职 Qualcomm 工程高管);2015—2025 年担任 Shield AI CEO从零搭建公司;推动估值从零增长至 53 亿美元;深耕国防行业关系网络高——在美国及外国政策领导层中构建了战略人脉网络
Andrew Reiter联合创始人AI 技术专家;构建了 Hivemind 早期算法基础创立期的 AI/自主技术公信力中——奠基性贡献;规模扩大后日常参与度有所降低
Gary Steele首席执行官(2025 年 5 月加入)创办 Proofpoint(IPO,企业价值 120 亿美元);担任 Splunk CEO(2024 年以 280 亿美元出售给 Cisco);加州大学伯克利分校本科毕业企业软件规模化、B2B 市场拓展、盈利纪律中——职业经理人 CEO;带来外部软件与商业化专业经验
Nathan Michael首席技术官学术型 AI/机器人背景;曾任职多所高校在自主系统研究领域拥有深厚技术公信力中——技术领导力的持续性至关重要
Kingsley Afemikhe首席财务官财务高管背景后期私募公司的运营财务与投资者关系管理低至中
Bob Harward国际业务拓展执行副总裁美国海军退役上将;前副国家安全顾问美国及盟友军事政策领域的高层人脉关系中——国际业务拓展部分依赖 Harward 的个人关系网络

管理层数据来源:PR Newswire 新闻稿(2025 年 3 月)、GovConWire 及 Craft.co(截至 2026 年)。Andrew Reiter 当前运营职能在近期公开资料中未有详述。

[CO002, CO003, CO004, CO007, CO008, CO009]

1.3 产品与技术组合

Shield AI 的产品组合以 Hivemind Enterprise——核心 AI 自主能力软件平台——为核心,辅以一系列软硬件一体化产品,后者兼具营收来源和 Hivemind 能力展示载体的双重作用。 Hivemind Enterprise 由四个模块组成:EdgeOS(自主执行运行时环境)、Pilot(自主能力目录与 AI 代理库)、Commander(多平台协同的指挥控制工具包)和 Forge(训练与仿真的自主能力工厂)。四者共同构成 Shield AI 所称的"任务自主能力开发者平台",使 OEM 厂商、国防主承包商和政府机构能够将自主行为集成到自身平台。Hivemind 已部署于逾 15 个平台,包括 Avenger(MQ-20)、X-62A VISTA(改装型 F-16)、BQM-177A 靶机和 Kratos 的 MQM-178 Firejet。该平台自 2019 年起经过实战验证,包括在乌克兰活跃战场环境中的部署。 V-BAT 是 Shield AI 的旗舰 VTOL 战术无人机,于 2021 年 7 月通过收购 Martin UAV 获得,在德克萨斯州达拉斯生产。采用导管重型燃料发动机(JP5),续航时间最长 11 小时,可携带 25 磅有效载荷,起降面积最小仅需 4.6 m × 4.6 m——包括在行进中的舰船甲板上。V-BAT 已入选美国海岸警卫队海上无人机项目,被全部七支海军陆战队远征部队使用,并已向印度、乌克兰、荷兰和亚美尼亚等盟军出售。其 ViDAR(宽域运动成像)传感器系统源自 2024 年 4 月收购 Sentient Vision Systems,每小时覆盖面积达 3,140 平方海里——是最近竞争对手的 2.5 倍以上。 X-BAT 于 2025 年 10 月发布,是全球首款 AI 驾驶的 VTOL 战斗机——重 23,000 磅,采用尾坐式设计,专为 GPS 拒止和通信拒止环境打造。X-BAT 代表 Shield AI 进军协同作战飞机(CCA)领域,设计用于在单一操控员指挥下以自主蜂群模式飞行作战。 2026 年 3 月,Shield AI 收购了 Aechelon Technology——一家为美军飞行员训练提供高保真军事仿真软件的公司,以加速 Hivemind 的训练与仿真能力。Booz Allen Hamilton 战略合作伙伴关系(2025 年 3 月宣布)借助 Booz Allen 的使命工程关系,将 Hivemind Enterprise 的触达范围扩展至更广泛的国防部企业。 [CO019, CO020, CO021, CO022, CO023, CO024]

FO002: Shield AI 商业模式与产品逻辑

Shield AI 核心软件平台(Hivemind)如何与硬件产品联动,并为国防部客户及商业 OEM 合作伙伴创造价值。

[CO019, CO020, CO021, CO031, CO032]

1.4 融资历程与投资方构成

Shield AI 自成立以来共完成 17 轮融资,累计融资约 $3.52B。首轮外部融资为 2015 年的 $600K 天使轮,公司于 2021 年跻身独角兽行列。融资历程折射出估值的快速跃升,驱动因素包括 V-BAT 在国防部的强劲普及、Hivemind 作战自主能力的实证,以及 2022 年后国防科技投资热情的持续高涨。 最近一轮重大融资为 $2B G 轮(2026 年 3 月 26 日),由 Advent International 和 JPMorganChase 安全与韧性倡议领投,另有 Blackstone 以固定回报优先股形式提供 $500M,以及一笔 $250M 信贷额度。G 轮确立了 $12.7B 融资后估值——较十二个月前的 $5.3B 上涨 140%,主要源于 Shield AI 作为 Hivemind 自主能力软件供应商入选美国空军协同作战飞机项目。其他 G 轮投资方包括 Snowpoint Ventures、InnovationX、Riot Ventures、Disruptive Ventures 和 Apandion。 上一轮 F 轮(含 2025 年 3 月 6 日宣布的 $240M F-1 子轮,合计约 $592M)由战略投资方 L3Harris(美国最大国防承包商之一)和韩华宇航(韩国国防主承包商)领投,Andreessen Horowitz、Booz Allen Hamilton、US Innovative Technology Fund 和 Washington Harbour Partners 参与跟投。战略投资方构成——两家国防主承包商加一家顶尖 VC——既是商业验证,也是与美国及盟友国防生态深度供应链整合的信号。 Shield AI 作为非上市公司未披露营收或 ARR 数据,但第三方分析引用的 2026 年预计营收约为 $540M(相当于较 2025 年估算基数约同比增长 80%)。公司尚未报告盈利,具体烧钱速度和现金跑道数据在公开来源中不可查阅。 [CO010, CO011, CO012, CO013, CO014, CO015]

股东与投资方概览
利益相关方角色控制权/经济重要性尽调事项
Advent InternationalSeries G 领投方(私募股权基金)最新一轮最大股权持有方;约 10 亿美元国防科技授权投资逻辑、董事会席位、治理权利及反稀释条款
JPMorganChase (Security & Resiliency Initiative)Series G 联合领投方共同领投 20 亿美元轮次;与金融服务战略方向协同战略回报与财务回报预期;董事会观察员权利
BlackstoneSeries G 优先股5 亿美元固定回报优先股;清算优先权清算优先权、分红条款及转换权利
L3Harris TechnologiesSeries F 战略领投方(国防大型承包商)重要股权持仓及供应链整合;年营收 190 亿美元的国防大型承包商募资用途战略协同、独家协议或授权条款及董事会权利
Hanwha AerospaceSeries F 战略联合领投方(韩国国防大型承包商)股权持仓;进入韩国国防市场的战略合作关系技术转让条款、合资协议及出口合规
Andreessen Horowitz(a16z,风险投资机构)Series D/F 参投方风险投资机构;科技生态背书投资依据及对软件授权与硬件利润率的预期
Booz Allen HamiltonF-1 轮战略投资方及合作伙伴Booz Allen 迄今最大企业风险投资;DoD 渠道合作伙伴合作独家范围、收入分成条款及 DoD 合同联合竞标规则
Ryan Tseng联合创始人兼总裁持有创始阶段大量股权;董事会成员;战略领导核心董事会构成、创始人投票权及 CEO 交接后的归属时间表
Brandon Tseng联合创始人兼总裁(首席增长官)持有创始阶段大量股权;掌握运营与战略影响力留任与竞业条款;接班计划

股权比例未公开披露。持股数据根据融资轮次参与情况推算;Tracxn 共列出 36 家机构投资方。Blackstone 以优先股而非普通股形式参与。

[CO010, CO011, CO012, CO013, CO014, CO015]

1.5 公司里程碑与组织规模

Shield AI 从两人创业的特种作战衍生公司发展为估值 $12.7B 的国防科技企业,经历了四个清晰阶段:创立原型期(2015—2018 年)、A 轮至 D 轮扩张期(2019—2022 年)、平台化与规模化阶段(2022—2024 年)及企业化与全球化阶段(2025 年至今)。 创立期诞生了 Nova 四轴飞行器——2018 年首款随美国特种作战司令部在战场实战部署的 AI 飞行员产品。Nova 专为 GPS 拒止环境下在士兵前方清扫建筑威胁而设计,联合创始人频繁援引该项目"挽救的生命"作为公司最有说服力的成果佐证。2021 年收购 Heron Systems(DARPA AlphaDogfight 冠军)和 Martin UAV(V-BAT 制造商)是两次关键规模化举措,使 Shield AI 同时获得了第五组飞机自主能力信誉和在海军/海军陆战队市场具有吸引力的战术 VTOL 产品。 2023 年里程碑——Hivemind 在 DARPA 认可的 AI 对人类格斗模拟中击败 F-16 飞行员——引发了最广泛的公众关注,验证了难度最高的自主能力主张(空中格斗机动)。2024 年,V-BAT 在乌克兰激烈电子战环境下的实战部署,展现了远超实验室条件的真实对抗性能,为盟军客户提供了极具说服力的参考。截至 2026 年 3 月,Shield AI 员工约 1,422 人——较 2025 年 5 月 CEO 交接时报告的约 900 人增长约 60%。 值得关注的风险维度:与所有致命自主系统研发商一样,Shield AI 面临日益上升的监管压力和人道主义压力。人权观察(2026 年 3 月)指出,美国五角大楼正加速推进完全自主武器——这一政策环境可能迫使 Shield AI 在国防部期望与国际人道主义法规范之间艰难周旋,尤其是当 X-BAT 逐步迈向致命交战角色时。 [CO027, CO028, CO029, CO030, CO031, CO039]

里程碑事件表
日期事件类型金额/估值/状态参与方意义
2015Shield AI 成立创立Brandon Tseng、Ryan Tseng、Andrew Reiter(三位创始人)由海豹突击队联合创始人基于实战经验缺口催生公司创立
2015首笔天使融资融资$600K天使投资人早期种子资金,用于构建 Hivemind 初始原型(Nova)
2018Nova 四旋翼无人机随美国 SOCOM 部署规模化US Special Operations Command首次将 AI 飞行员用于实战军事行动;在战场验证救生效果
2020V-BAT 投入战术无人机系统使用产品Shield AI(后收购 Martin UAV)VTOL 平台填补海军/海军陆战队 ISR 作战能力关键缺口
2021-07收购 Heron Systems(DARPA AlphaDogfight 比赛冠军)产品Heron Systems五类飞行器 AI 能力;升级 Hivemind 空战机动能力
2021-07收购 Martin UAV(V-BAT 制造商)产品Martin UAV垂直整合 VTOL 硬件与 Hivemind AI 技术栈
2021达到独角兽里程碑融资>$1B 估值多方投资者验证国防 AI 商业模式;吸引机构资本入场
2023Hivemind 在 DARPA AI 对决人类飞行员(F-16)模拟空战中胜出产品DARPA, US Air Force最高难度自主能力验证;确立 AI 飞行员在战斗机级飞行器上的公信力
2023-10完成 Series F 首期 2 亿美元融资融资$2.7B 估值Series F 投资方推动 Hivemind Enterprise 与 V-BAT 产线规模扩张
2024-04收购 Sentient Vision Systems(墨尔本)产品Sentient Vision Systems为 V-BAT 传感器套件新增 ViDAR 广域运动成像能力
2024V-BAT 投入乌克兰实战规模化乌克兰军方、美国对外军售(FMF)GPS 干扰真实战场验证;国际客户实战证明
2025-03-06完成 Series F-1 融资 2.4 亿美元,估值 53 亿美元融资$5.3BL3Harris、Hanwha、a16z、Booz Allen、Washington Harbour(投资方与合作方)估值同比翻倍;首次引入国防大型承包商战略投资
2025-03宣布与 Booz Allen Hamilton 建立合作关系合作Booz Allen Hamilton(NYSE: BAH,上市咨询 / 防务承包商)为 Hivemind Enterprise 开辟 DoD 企业级渠道;Booz Allen 迄今最大 CVC 投资
2025-05-13Gary Steele 出任 CEO;Ryan Tseng 转任总裁治理Gary Steele(前 Splunk CEO)、Ryan Tseng引入职业经理人管理层;聚焦企业软件规模化
2025-10X-BAT(VTOL 打击战斗机)正式发布产品Shield AI首款 AI 驾驶 VTOL 战斗机;瞄准协同作战飞机(CCA)角色
2026-03-26完成 Series G 融资 20 亿美元,估值 127 亿美元;收购 Aechelon Technology融资$12.7BAdvent International, JPMorganChase, Blackstone估值 12 个月内升幅 140%;CCA 项目胜出为核心驱动因素

日期来源:PR Newswire、TechCrunch、Tracxn 及 AIAA Aerospace America。部分早期轮次金额为近似值或数据不可得。根据 Defense News/prnewswire,Martin UAV 收购于 2021 年完成。

[CO001, CO018, CO021, CO023, CO025, CO026]
FO001: Shield AI 公司里程碑时间线(2015–2026)

Shield AI 自 2015 年创立至 2026 年 3 月 G 轮融资及收购 Aechelon 期间的核心产品、融资与治理里程碑。

[CO001, CO007, CO013, CO021, CO023, CO024]

1.6 重要发现

Chapter 02

02市场分析

2.1 市场边界与范围

Shield AI 在两个相互交叉的市场垂直领域展开竞争:(1)军事 AI 自主能力软件——使无人系统能够在对抗环境中无需人类干预即可感知、决策和行动的软件;(2)战术 VTOL 无人机(UAS),尤其是舰载与地面便携型平台,如 V-BAT 和 X-BAT。 该市场的采购方集中于国防领域,主要为美国国防部和盟军,需要能在 GPS 拒止和通信拒止环境下运行的自主系统。核心采购决策在于选择平台无关的自主能力软件(Hivemind Enterprise 授权模式)、战术硬件(V-BAT/X-BAT 采购)或两者组合。现状替代方案是人工驾驶或遥控平台,其操作员成本更高、认知负荷更重,且更容易受到电子战干扰。 Shield AI 可服务市场不包括:民用/商业无人机应用(农业、物流、基础设施巡检)、消费级 AI 软件、非国防卫星系统,以及大型战略无人机项目(MQ-9 "死神"或"全球鹰"等 HALE 平台,属于完全不同的预算类别)。反无人机(C-UAS)市场同样不在其中——尽管该市场在 2026 财年国防部预算中获得 $3.1B 拨款,但由独立供应商服务,与 Shield AI 产品组合无重叠。 邻近市场包括:有人-无人协同(MUM-T)软件平台、国防通信与网络(JADC2),以及最终可能将 Hivemind 作为对抗性 AI 伙伴纳入的军事训练仿真系统。 [CM021, CM011, CM031]

市场定义表
细分市场/类别纳入支出排除支出主要买家/付款方与 Shield AI 的相关性
军用 AI 自主软件GPS 拒止导航技术栈、多智能体协同、边缘 AI 推理商业 AI 软件、非国防应用DoD(AFMC、DIU、SOCOM、NavSea 等采购方)核心产品——Hivemind Enterprise
战术 VTOL 无人机硬件可舰载发射、可陆上携带的 VTOL 无人机(第 2—4 组)高空长航时平台(MQ-9、Global Hawk)、商业无人机US Navy、SOCOM、Coast Guard、盟军V-BAT 与 X-BAT 产品线
协同作战飞机软件依据 A-GRA 为无人僚机提供任务自主技术栈有人机飞控系统、商业无人机软件USAF / AFRL CCA 项目Hivemind 于 2026 年 2 月入选 YFQ-44A
盟友军用无人机系统/自主能力对外军售(FMS)及直接商业无人机非盟友出口管制市场合作伙伴军队(印度、荷兰、阿联酋、韩国、乌克兰)V-BAT 出口;Hivemind 授权潜在扩张
反无人机系统FY2026 DoD 预算 31 亿美元进攻型自主系统(Shield AI 的主要领域)USAF、US Army、US Navy(三军客户)仅为相邻市场;Shield AI 不在此领域竞争
有人驾驶平台(替代品)需要射频链路的遥控无人机;有人机N/A——此为现状替代DoD 全军种现状正被 Hivemind 自主能力主张所取代

市场边界来源:defensescoop.com FY2026 DoD 自主能力预算拆解及 Shield AI 官方产品定位。CCA 项目 A-GRA 详情来源:2026 年 2 月 13 日 Shield AI 新闻稿。

2.2 市场规模:TAM、SAM 与 SOM

军事 AI 与自主能力的市场规模估算在各分析机构间存在争议,原因在于定义标准不一——部分机构纳入硬件平台,部分仅统计软件,部分将所有无人系统不论自主程度一律打包。以下分析采用三个嵌套视角,针对 Shield AI 的实际产品覆盖进行校准。 **总可用市场(TAM):** 最广泛的相关市场为全球自主国防平台市场,Fortune Business Insights 估计 2026 年规模约为 $698 亿,以 14.0% 的年复合增长率扩张,预计 2034 年达到 $1,989 亿,涵盖全球所有自主空中、陆地和海上国防平台。采用更窄定义——仅统计全球军事无人机——则 2026 年估值区间在 $186 亿(Mordor Intelligence)至 $272 亿(Coherent Market Insights)之间,共识集中于 $190—210 亿。军事 AI 软件细分市场(Global Growth Insights)2026 年估计为 $93 亿,以 13.9% 的年复合增长率扩张,预计 2035 年达到 $299 亿。 **可服务市场(SAM):** Shield AI 的 SAM 以美国国防部 2026 财年 $134 亿专项自主能力预算为锚——这是国防部首次将自主能力单独列为独立预算类别。其中,$94 亿分配给无人及遥控空中飞行器,$12 亿专项用于"使能能力"(所有平台底层的自主能力软件)。盟国军事采购——包括韩国(韩华战略合作伙伴)、阿联酋、乌克兰、荷兰、印度和英国——带来有实质意义但难以量化的增量 SAM。仅战术 VTOL 平台的 SAM 基于 V-BAT 项目支出速率估计为每年 $20—40 亿;自主能力软件授权的 SAM 估计为每年 $10—20 亿。综合 SAM:约 $30—60 亿。 **可获取市场(SOM):** 以 2026 年预计营收约 $540M 及快速扩张的平台无关模式计算,Shield AI 目标 SAM 占比约 10—15%。若 CCA 项目达到全面量产(预计十年末装备数百架飞机),Shield AI 的 Hivemind Enterprise 授权 SOM 将大幅扩张。空军明确提出要"大量"装备无人僚机,这将带来数十亿美元规模的经常性软件授权机会。 [CM001, CM002, CM003, CM004, CM005, CM008]

总可用市场/可服务市场/可获取市场规模估算表
维度/范围2026 年估算(美元)复合年增长率(CAGR)来源方法论/注意事项
TAM——全球自主国防平台$69.8B14.0%(至 2034 年)Fortune Business Insights范围最广;涵盖全球空中、陆地及海上自主平台
TAM——全球自主军事系统$40.9B11.97%(至 2032 年)360iResearch除空中外,还涵盖地面与海上;范围略窄于 FortuneBI
TAM——仅军用无人机系统(全球,高估)$27.2B11.5%(至 2033 年)Coherent Market Insights(市场研究机构)仅限空中无人机系统;涵盖所有自主等级(遥控、半自主、全自主)
TAM——仅军用无人机系统(全球,中估)$20.7B13.8%(至 2035 年)Global Market Insights(市场研究机构)2026 年 1 月发布;部分细分市场包含民用双重用途——数据略有虚高
TAM——仅军用无人机系统(全球,低估)$18.6B7.03%(至 2031 年)Mordor Intelligence范围最窄;保守估算;对纯国防无人机系统而言最具可信度
TAM——全球军用 AI 软件$9.3B13.9%(至 2035 年)Global Growth Insights(市场研究机构)仅含软件;最贴近 Hivemind Enterprise 授权收入潜力
SAM——美国 DoD 自主能力预算(直接可触及)$13.4B FY2026N/A(年度预算)DoD / DefenseScoop年度采购预算;$9.4B 航空 + $1.2B 软件及其他;并非全部可由 Shield AI 直接承包
SAM——美国 DoD 使能自主软件$1.2B FY2026N/ADoD / DefenseScoopDoD 内最接近 Hivemind 授权 SAM 的范畴;不含盟友市场
SOM——Shield AI 可获取市场(估算)$2–5B持续扩大分析师估算(作者团队)由营收、V-BAT 项目支出及 CCA 潜力推算;不确定性较高

市场估算来自多家分析机构,方法论与范围定义各异。各行 TAM 数据不可直接横向比较。SAM/SOM 为推算值,并非公司披露数据。各 TAM 定义下的分析师估算相差 7—19 倍——造成差异的主要因素是定义范围,而非数据质量。

[CM001, CM002, CM003, CM004]
FM001: Shield AI 市场规模金字塔(2026)

Shield AI 三层市场规模视图:从最宽口径总可用市场(TAM,全部自主防务平台)、到可服务市场(SAM,国防部可触达 + 盟友),再到可获取市场(SOM,Shield AI 可获取部分),附 2026 年数值。

由于口径差异,各机构 TAM 估算相差约 7 倍。SAM 基于已公布的 FY2026 国防部预算。SOM 为作者推断,非公司披露。

[CM001, CM002, CM003]
FM002: 军用无人机系统市场估算区间(2026–2035)

2026 年及预测期末全球军用无人机系统市场的低、中、高三种情景,呈现各分析来源约 7 倍的估值差距。

由于口径差异,各估算不可直接比较。所有数据均来自商业发布的分析师报告,非政府一手数据。

[CM002, CM003, CM004]

2.3 采购方、用户与付款方细分

Shield AI 的市场涉及不同的采购方、用户和付款方角色,三者并不总是同一主体,这直接影响公司的市场进入模式。 **主要付款方和采购方:美国国防部(USAF、USN、SOCOM、陆军、海岸警卫队)。** 国防部既是主要付款方(通过拨款),也是主要采购方(通过 AFWERX、DIU、PEO Aviation、NAVSEA 和 SOCOM 等合同机构)。CCA 项目(AFMC/AFRL)是 2026 年最受关注的新增付款方。国防创新单元(DIU)通过其他交易授权(OTA)为 Shield AI 等非传统供应商快速推进合同。 **用户(操作员):** 空军战斗机飞行员和联队指挥官(CCA/Hivemind);海军舰艇指挥官(V-BAT/ViDAR 用于海上情报监视侦察);SOCOM 操作员(Nova,GPS 拒止建筑清扫);美国海岸警卫队(V-BAT 海上无人机系统服务);陆军地面部队(小型无人机集成)。用户关注点在于对抗环境下的可靠性、操作便捷性和任务效果。 **战略/盟国采购方:** 2026 财年国防部预算显示,海军申请无人系统经费 $53 亿(较 2025 财年增长 70%)。陆军申请 2026 财年小型无人机经费 $8.039 亿(较 $9,840 万大幅增长)。这些新兴采购方的预算增长直接利好 Shield AI 产品路线图。 **国际盟国采购方:** 通过合作协议和对外军售(FMS),Shield AI 已向印度、乌克兰、荷兰和亚美尼亚出售 V-BAT。韩华系统对 Shield AI 的战略投资打开了韩国国防采购厅(DAPA)市场渠道。L3Harris $100M 战略投资及后续 V-BAT 集成,扩大了盟国 VTOL 情报监视侦察市场。 **主承包商作为平台客户:** Anduril(YFQ-44A Fury CCA)、General Atomics(MQ-20 Avenger Hivemind 测试)、Northrop Grumman(Talon IQ 生态)和 Airbus(UH-72A Lakota)均作为平台客户集成 Hivemind 软件组件——这是一个有别于直接国防部采购的 B2B 授权渠道。 [CM006, CM007, CM023, CM024, CM032, CM036]

细分市场/买家图谱
买家细分预算主体使用场景采购路径估算细分规模/增速
USAF CCA 项目AFMC / AFRL PEO (Air)F-35/F-22 编队的 AI 驾驶无人僚机TMRR → EMD → LRIP → FRP (2026–2030+)约 $1–3B(全周期)—— 项目规模待 FRP 决策确定
美国海军(MEU、舰载 UAS)PEO Unmanned and Small Combatants(USC,无人及小型作战舰艇项目办公室)海上 ISR、舰载 VTOL 起降回收通过 NAVSEA 正式列装;V-BAT 配备全部 7 支 MEU约每年 $500M–1B(海军 UAS)
SOCOM / 特种作战USSOCOM 采购GPS 拒止环境下的近战侦察;Nova/HivemindOTA 合同 → 战场部署 → 标准配发小众但高价值;2018 年起已实战部署
美国海岸警卫队DHS / USCG 采购海上 UAS 服务;近岸监视竞标授奖(V-BAT 中标)约 $50–200M 项目价值
美国陆军 SUAS 项目PEO Aviation / Army Futures Command(美国陆军航空 / 未来司令部)小型班组级无人机;可消耗 ISRFY2026 SUAS 预算 $803.9M;通过 Transformation in Contact 加速推进$803.9M FY2026(广义 SUAS 类别;V-BAT Group 4 为重点)
盟国军队(FMS)通过美国国务院 FMS 的外国政府战术 VTOL、海上 ISR、边境巡逻FMS 案例 → 国家军事部署印度、荷兰、亚美尼亚、阿联酋、乌克兰客户;Hanwha 韩国渠道
主承包商(平台被许可方)承包商独立研发(IR&D)/ 政府研发合同Hivemind 集成于 Fury、MQ-20、XQ-58、Avenger符合 A-GRA 规范的 B2B 软件授权新兴模式;合同价值未披露

买方细分数据来源:defensescoop.com FY2026 预算分析、insideunmannedsystems.com 2026 年报告及 Shield AI 官方公告。细分规模估算为大约值,采用公开预算数据;实际合同授予可能存在较大差异。

[CM005, CM006, CM007, CM008]
FM003: 采购方 / 细分市场流程:Shield AI 国防部采购渠道

展示国防部预算授权如何通过多种采购渠道——正式项目记录(PoR)、OTA、对外军售(FMS)及 B2B 平台授权——转化为 Shield AI 营收的流程图。

[CM005, CM006, CM007]

2.4 增长驱动与采购催化剂

截至 2026 年,多重结构性催化剂正在加速 Shield AI 产品的需求增长。 **2026 财年国防部自主能力预算机制化。** 历史上首次,国防部将自主能力列为独立预算类别(2026 财年 $134 亿),标志着国防采购的永久性结构转变。这一举措消除了 Shield AI 早年面临的最重要采购障碍:AI 自主能力专项资金的缺失。 **乌克兰战争无人机需求信号。** 乌克兰冲突成为历史上可抛弃式无人机作战的最强有力概念验证。乌克兰 2024 年采购了 150 万架 FPV 无人机,计划 2025 年采购 450 万架,预算达 $26 亿。美国陆军的"接触中转型"计划——旨在 2026 财年末为每个班组配备小型无人机——明确借鉴了乌克兰作战战法,为战术自主系统创造了直接的需求拉动。 **CCA 项目作为平台无关软件楔入点。** 空军 CCA 项目采用开放的自主能力政府参考架构(A-GRA),为 Hivemind 在多款飞机平台的标准化集成创造了路径。Shield AI 于 2026 年 2 月入选 Anduril YFQ-44A Fury 的任务自主能力供应商,随后飞行测试中 Hivemind 在飞行途中切换至 Anduril 的 Lattice AI,验证了平台无关论点。若 CCA 项目扩展至数百架飞机,Hivemind 授权收入规模有望远超整个硬件产品线。 **全球国防预算扩张。** 2024 年全球军事支出达 $2.7 万亿(SIPRI),亚太地区连续 35 年增长。日本同比增长 21%、AUKUS 多域自主能力项目,以及中东国防现代化进程,为 Shield AI 提供了美国市场之外切实可行的近期增长机会。 **软件原生架构优势。** 国防平台从硬件中心向软件定义转型,与商业科技从本地部署向 SaaS 演进的路径相呼应。Shield AI 的 Hivemind Enterprise 平台即服务模式,无需为每款飞机重建软件栈即可快速实现多平台部署,每增加一个平台集成便带来更强的复合网络效应。 [CM009, CM010, CM011, CM012, CM013, CM018]

FM004: 国防 AI 采纳漏斗:从国防部到 Shield AI 部署

简化漏斗图,展示国防部采购意向如何从预算授权到实战部署各阶段逐步转化为 Shield AI 营收。

[CM009, CM010, CM011]

2.5 市场约束与采购障碍

在上述顺风之外,若干结构性约束抑制了近期市场的乐观预期。 **国防预算不确定性。** 特朗普政府 2026 财年预算进程波折不断,包括 DOGE 驱动的文职削减、项目取消,以及为填补缺口而依赖国会和解拨款(申请 $1,133 亿)。Inside Unmanned Systems 报告指出,部分接近研发完成的长期无人机项目已被取消,转而倾向现货采购,给依赖特定项目续期的供应商带来营收风险。 **国防部漫长的采购周期。** 尽管有 OTA 和 DIU 快速推进机制,大多数大型国防部项目仍需多年的来源选择、国会授权和漫长的作战测试。Shield AI 在 CCA 中的角色虽经飞行测试验证,但尚未进入全速生产决策阶段。V-BAT 等硬件项目面临低速初始生产/全速量产节点及政府验收测试,带来营收时间上的不确定性。 **出口管制约束。** ITAR/EAR 法规限制了 Shield AI 向特定盟友出口先进自主能力软件和 VTOL 平台,除非获得政府许可。自主武器争论在国务院和商务部层面引发了额外审查。国际扩张步伐在一定程度上受制于监管审批时间线。 **伦理与监管悬而未决。** 人权观察、国际红十字委员会以及越来越多的国际法学者认为,完全自主致命系统——包括 AI 驾驶的战斗机——可能违反武装冲突法。尽管 Shield AI 当前部署在官方国防部框架下被归类为"人在回路"或"人在路上"系统,自主武器条约谈判带来的声誉和监管风险仍是中期隐患。 **分析师估值差异作为尽调信号。** 军事 AI 和自主系统的市场规模估算 2026 年区间从约 $90 亿(仅 AI 软件)至约 $700 亿(所有自主国防平台)——差距达 8 倍。这一差异反映的是真实的定义分歧,而非数据质量问题,并表明 Shield AI 等公司的营收预测对定义范围、项目结果和预算波动高度敏感。 [CM016, CM017, CM025, CM026, CM027, CM040]

增长驱动因素与制约因素表
因素类型程度机制对 Shield AI 的影响
FY2026 美国国防部 $13.4B 自主性预算专项驱动因素首个专项自主性预算,支持多年期采购规划直接影响:所有产品线受益;$1.2B 软件专项可直接寻址
乌克兰战争无人机战术理论驱动因素可消耗自主 UAS 的实战概念验证;触发陆军 Transformation in Contact 项目间接影响:V-BAT 已在乌克兰部署;推动美军对类似系统的紧迫需求
CCA A-GRA 开放架构驱动因素平台无关的 A-GRA 标准支持 Hivemind 跨多个厂商授权直接影响:2026 年 2 月 Hivemind 获选 YFQ-44A;经常性软件授权模式
全球国防开支扩张(2024 年全球 $2.7T)驱动因素盟国预算增长推动 FMS 需求;亚太地区连续 35 年增长间接影响:Hanwha(韩国)、印度、阿联酋渠道扩张;V-BAT 出口增长
软件优先、平台无关架构驱动因素类似 SaaS 与本地部署的转型;降低每个平台的集成成本直接影响:Hivemind 的竞争护城河随每次平台集成不断扩大
国防创新单位(DIU)扩张($2B 预算)驱动因素DIU 为非传统供应商快速推进 OTA 合同;Shield AI 属于 DIU 合作型供应商间接影响:加快从演示到合同的时间;降低采购摩擦
美国国防部预算不确定性 / DOGE 压力制约因素项目取消、削减开支及预算协调不确定性带来收入时序风险实质影响:部分在途项目面临取消风险;V-BAT 项目可能受波及
美国国防部采购周期漫长制约因素即便有 OTA,FRP 决策仍需 3–7 年;FRP 前收入可见度有限实质影响:CCA 尚未达到 FRP 阶段;V-BAT FRP 决策悬而未决
出口管制 / ITAR 限制制约因素ITAR/EAR 限制在无国务院许可证情况下向盟友出口自主武器中等影响:减缓盟国拓展;韩国/印度 FMS 需逐案审批许可
自主武器监管悬而未决制约因素低至中人权观察(HRW)、国际红十字委员会(ICRC)、条约谈判风险;美国国防部 3000.09 号指令要求"适当审慎"近期低风险:美国国防部尚未限制开发;中期存在声誉风险
分析师估算分歧(总可用市场(TAM)7 倍区间)风险信号低(尽调层面)TAM 估算 8 倍区间意味着定义不确定性;预测可能与实际存在较大偏差信息性:投资逻辑的市场规模测算须考虑较大的不确定区间

增长驱动因素数据来源:FY2026 美国国防部预算(defensescoop.com)、startus-insights.com 2026 年无人机报告、 fortunebusinessinsights.com 自主防务平台报告及 insideunmannedsystems.com 2026 年采购报告。 制约因素数据来源:人权观察(HRW)、360iresearch.com 及美国国防部 3000.09 号指令相关参考资料。

[CM009, CM010, CM011, CM012, CM013]

2.6 重要发现

Chapter 03

03竞争格局

3.1 AI 优先同行:Anduril 与自主能力软件栈之争

Shield AI 战略层面最重要的竞争对手是 Anduril Industries,2017 年由 Palmer Luckey 创立,总部位于加利福尼亚州科斯塔梅萨。Anduril 是美国唯一一家具有可比软件优先自主能力架构、获风险投资支持且运营规模接近的国防科技公司。截至 2026 年 3 月,Anduril 正在 Thrive Capital 和 Andreessen Horowitz 领投的 H 轮融资中寻求 $600 亿估值,较 2025 年 6 月 G 轮的 $305 亿大幅提升。Anduril 2025 年营收约 $21 亿——约为 Shield AI 估算 $540M 的 4 倍——反映出其更早形成、更多元化的软硬件产品组合,包括 Ghost 4 无人机、Roadrunner 拦截器和 Fury 自主战斗机(YFQ-44A)。2026 年 3 月,Anduril 获得美国陆军 10 年期、$200 亿上限的自主系统合同,是公司历史上最大单笔合同,也印证了国防部对 AI 原生主承包商的强烈胃口。 技术层面,Anduril 的 Lattice AI 是一个传感器融合与指挥控制平台,作为战场神经系统运行——整合分布式传感器数据,形成统一战场态势图像,并在空中、陆地和海上域自动化威胁响应。Lattice 的核心优势是跨异构平台网络的横向集成,而 Shield AI 的 Hivemind 专注于单平台级别的纵向深度:使单架飞机或无人机能够在完全信号隔离、零人机协同的情况下导航、规划和执行任务。2026 年 3 月 YFQ-44A 双栈飞行演示——Hivemind 和 Lattice 均在同一机体上得到验证——证实了 A-GRA 模块化架构允许共存,国防部追求的是最佳组合而非单一软件栈垄断。这种竞合模式使竞争格局更为复杂,但也验证了 Shield AI 的模块化授权逻辑。 第二家 AI 优先同行是 Palantir Technologies(NYSE: PLTR),其 Gotham 平台是国防部情报社区主导的数据分析与打击 AI。Palantir 对 Shield AI 的竞争威胁是间接的——Palantir 不制造自主飞行器或 GPS 拒止边缘代理——但在 Hivemind Commander 与国防部指挥控制框架交汇的指挥层面有所重叠。Palantir 2025 年营收约 $28 亿(美国政府部门约 $11 亿),毛利率约 80%,自 2023 年起持续盈利。截至 2025 年末,Palantir 在五角大楼 AI 分析领域的垄断地位正在削弱:Shield AI、Anduril 和 Applied Intuition 的竞争平台正在特定任务领域蚕食其市场,但 Palantir 在情报与打击工作流中仍具深度壁垒。 [CP001, CP002, CP003, CP004, CP005, CP006]

竞争对手概况表
公司类别2025 年收入(估计)融资 / 估值主要产品目标细分市场相较 Shield AI 的主要不足
Anduril IndustriesAI 优先同类竞争者~$2.1B已融资 $6.3B;估值 $30.5B(2025)Lattice AI + Fury CCA + Ghost drones(软件与无人机组合)美国国防部,全域Lattice 专注于 C2/数据融合,而非 GPS 拒止边缘环境;硬件成本结构偏重
Shield AI(参考基准)AI 优先同类竞争者~$540M est.已融资 $3.52B;估值 $12.7B(2026)Hivemind + V-BAT + X-BAT美国国防部 + 盟国,空域
Kratos Defense可消耗飞机主承包商~$1B上市公司(KTOS);市值约 $3BXQ-58A Valkyrie + Prism美国空军/海军陆战队忠实僚机 CCAPrism 自主性为平台专用;可消耗设计限制重复使用
General Atomics (GA-ASI)大型 UAS 主承包商~$3B+私有公司(General Atomics 旗下)Avenger + MQ-9B SkyGuardianGroup 5 ISR/打击,盟国 FMS硬件优先;在 USAF CCA 增量 1 竞标中落败;自主软件能力有限
Palantir Technologies国防分析 / C2 AI~$2.8B上市公司(PLTR);市值约 $200BGotham + Foundry + AIP情报、目标定位、C2 分析无 GPS 拒止边缘自主能力;仅在数据/C2 层面竞争
Boeing Defense主承包商集成商约 $33B(BDS 业务段)Public (BA)MQ-25 Stingray + Palantir 合作海军自主加油机、ISR软件迭代缓慢;成本加成模式;AI 方面依赖 Palantir
L3Harris Technologies传统电子战(EW)/ISR~$23B上市公司(LHX);市值约 $45BHCSW + 自主僚机附加模块电子战、空中 ISR2024 年将 ISR 硬件出售给 Anduril;自主软件尚处起步阶段

收入数据为私营公司基于公开文件、分析师报告和新闻来源的估计。Shield AI 列示作为对比基准。估值截至最近一轮披露。

FP001: 竞争定位图
[CP007, CP022, CP038]

3.2 传统主承包商与平台集成商

第二竞争梯队由传统国防主承包商和专业无人系统公司组成,它们正在构建或收购自主能力,以与 Shield AI 和 Anduril 展开竞争。这些参与者拥有制造规模、ITAR 合规供应链和数十年积累的国防部关系资本,但受制于成本加成合同结构,限制了研发投入强度,拖慢了软件迭代周期。 Kratos Defense & Security Solutions 是可抛弃式飞机领域与 Shield AI 最为相近的竞争对手。其 XQ-58A Valkyrie——一款最初在 AFRL 的 LCAAT 项目下研发的低成本喷气动力可抛弃式自主飞机——2026 年 1 月通过 $2.315 亿合同入选美国海军陆战队 MUX TACAIR 项目,由 Northrop Grumman 牵头的团队承接,Kratos 提供飞行平台,Northrop 集成任务套件和 Prism 自主能力软件栈。关键在于,Prism 是 Northrop 自研的自主能力层——既非 Hivemind 也非 Lattice——由此形成第三款争夺 CCA 自主能力层市场的软件栈。Kratos 2025 年战术系统及卫星通信年营收约 $10 亿。其单位经济与 Shield AI 结构性不同:Valkyrie 作为可抛弃(一次性)资产设计,估计单价 $200—300 万,而 Shield AI 的 V-BAT($1M 以上)具备回收和重复使用能力。 General Atomics Aeronautical Systems(GA-ASI)是美国第五组(大型)军事无人机的主导制造商,主力产品为 MQ-9 "死神"(2023—2025 年起逐步在 USAF 退役)及其后继项目。GA-ASI 参与竞争空军 CCA 增量 1 项目未能胜出。其 Avenger 是一款更大、隐身性更强的喷气式无人机,用于情报侦察与打击任务。GA-ASI 是私人公司(General Atomics 旗下),估计年营收约 $30 亿以上,从根本上是一家以硬件为中心的组织。GA-ASI 正在其平台上集成更多机载自主能力,但其自主能力软件是自研专有的,而非平台无关。GA-ASI 和 Kratos 共同代表了与 Shield AI"软件优先"模式相对的"硬件优先"竞争原型。 Boeing Defense(MQ-25 Stingray 自主加油机、在澳大利亚的 MQ-28A Ghost Bat)、L3Harris(战术电子战和情报侦察系统、自主僚机项目)和 Northrop Grumman(RQ-4 "全球鹰"、X-47B 技术传承、B-21 突袭者集成)构成了主承包商集成商梯队。Boeing 于 2025 年 9 月与 Palantir 建立合作,在其国防和保密项目中加速 AI 应用,将 Palantir 数据分析内嵌于 Boeing 的主承包商集成角色。L3Harris 2024 年以约 $8 亿将其 WESCAM 机载情报侦察业务出售给 Anduril,这是 Anduril 获取硬件制造能力的战略举措,也印证了 AI 优先公司与传统主承包商之间的边界正通过并购日趋模糊。 [CP010, CP011, CP012, CP013, CP014, CP015]

功能 / 能力矩阵
能力Shield AIAndurilKratos + NorthropGeneral AtomicsPalantir
GPS 拒止边缘自主能力(AI 飞行员)✓ 经实战验证部分具备(传感器融合层)部分具备(Prism,已通过仿真测试)✗ 不具备✗ 不具备
平台无关软件授权✓ A-GRA 认证,15+ 平台部分具备(Lattice 运行于 Anduril 硬件 + YFQ-44A)✗ 平台专用✗ 平台专用✓ 跨平台云端/企业级
完整硬件堆栈部分具备(V-BAT、X-BAT;非主力平台)✓ Ghost, Fury, Roadrunner, IVAS✓ XQ-58A(可消耗)✓ Avenger、MQ-9B、Predator 系列✗ 无硬件
有人-无人协同(MUM-T)✓ Hivemind Commander✓ Lattice + Fury✓ Valkyrie CCA部分具备(Avenger/MQ-9 编队)部分具备(仅限 C2 层)
开放架构(A-GRA)✓ CCA 项目主技术栈✓ 传感器融合层已认证部分具备(认证中)✗ 专有部分具备(云层)
国际 / FMS 客户✓ 阿联酋、澳大利亚、印度、荷兰、亚美尼亚部分具备(英国、澳大利亚)部分具备(通过 USMC 项目 FMS)✓ 多个盟国客户✓ 多个盟国政府
实战部署记录✓ SOCOM + Ukraine✓ 边境安全 ISR有限(测试/评估阶段)✓ MQ-9 大规模部署✓ 伊拉克、阿富汗目标定位

"部分具备"表示相关能力存在,但在范围、规模或认证状态方面受到限制。评估基于截至 2026 年 5 月的公开合同授奖、产品公告及运营报告。

定价 / 产品包装对比
公司 / 产品定价 / 合同模式单位经济折扣 / 未知信息对 Shield AI 的启示
Shield AI Hivemind Enterprise(企业版)按平台软件授权 + SDK;政府定价受保密协议约束高利润率软件;硬件单独计价大规模机队授权可能享有批量折扣经常性收入模式;与硬件成本解耦
Shield AI V-BAT硬件采购;每台约 $1M+(估计)可回收 VTOL;每日约 5 架次能力具体定价未披露;IDIQ 合同相较于有人驾驶方案,单位成本相对较低
Anduril Lattice与硬件捆绑 / OTA 合同;定价受保密协议约束整体产品组合估计毛利率 40–45%固定总价模式(非成本加成)若 Lattice 将 Hivemind 商品化,Anduril 的定价权构成结构性威胁
Kratos XQ-58A Valkyrie可消耗单位成本约 $2–3M;固定价格合同模式每架次低成本;为可消耗性而设计海军陆战队首批 $231.5M OTA可消耗定价激发美国国防部对低价自主飞机的需求——对 V-BAT 可重用性溢价构成风险
General Atomics MQ-9B SkyGuardian正式采购项目;每架约 $25–35M(估计)成本加成维护;全生命周期成本高FMS 定价须经国会通报程序GA-ASI 的高定价在预算较小的项目中为 Shield AI 提供了竞争优势
Palantir Gotham/AIP企业级 SaaS;按席位 + 数据量计价;机密合同保密协议约 80% 毛利率;高度盈利政府定价保密;多年期企业许可协议(ELA)Palantir 的 SaaS 模式印证了美国国防部为软件优先 AI 订阅付费的意愿

所有定价估算均基于公开披露的合同价值、行业分析师报告及二手来源。所有产品均无官方公开定价。Shield AI 定价仅供示例参考。

FP002: 功能/能力矩阵
[CP005, CP012, CP016, CP018]

3.3 技术差异化与竞争定位

Shield AI 的主要技术差异化在于 Hivemind 在 GPS 拒止和通信拒止(COMMS-D)环境中的作战深度——这是自主能力谱系的"尖锋",飞机必须在零人机接口的情况下完全自主行动。Anduril 的 Lattice 相比之下针对网络互联、数据丰富的环境进行优化,其优势在于整合分布式传感器输入并协同多平台响应。这两家公司争夺的并非同一技术需求;它们争夺的是预算资金和合同载体,往往出现在互补架构中。2026 年 3 月的 A-GRA 演示印证了这一点:Hivemind 和 Lattice 各自"占据"同一飞机自主能力栈的不同层级。 在 GPS 拒止自主能力维度上,Shield AI 在需要该特定能力的国防系统集成商中具有无可争议的竞争优势。Hivemind 是唯一具有在 GPS/通信拒止战区(乌克兰、未公开的 SOCOM 行动)有据可查的实战部署记录的商用 AI 软件栈。Anduril 的 Ghost 4 无人机已在边境安全和情报侦察场景中部署,但无公开记录的 GPS 拒止实战部署。Kratos 的 Valkyrie 依赖 Northrop 的 Prism,后者在仿真优先环境中经过测试,但已记录的实战部署较少。 切换成本分析揭示了持久壁垒。在国防部平台上认证新的自主能力软件栈需要:(1)按 MIL-STD 和 FACE 技术标准进行软件鉴定测试;(2)为自主交战包线提供安全案例文档;(3)按国防部网络安全成熟度模型认证(CMMC)进行网络安全评估;(4)与平台项目办公室进行运营集成测试。在已认证的作战记录项目上切换自主能力栈的估计时间:18—36 个月。这为率先在特定平台上完成认证的软件栈创造了强大的在位优势——这正是 CCA A-GRA 开放架构标准既是机遇(赋能初始中标)也是风险(若竞争对手更快完成认证则赋能未来替换)的原因。 多栈分析:A-GRA 架构明确支持国防部项目在同一机体上运行多套自主能力栈——2026 年 3 月 YFQ-44A 测试已作证明。这意味着 Shield AI 和 Anduril 理论上在 CCA 平台上可直接互换。然而,国防部项目办公室的预算约束在实际中将选择限定为每个项目一套主要自主能力栈。在 CCA 中观察到的结果是:Shield AI 主导"任务自主能力"层(自主驾驶行为),Anduril 主导"传感器融合"层(战场态势感知与任务分配)。近期来看,两者是互补关系而非竞争关系。 [CP007, CP008, CP009, CP018, CP019, CP020]

FP003: 护城河/战备能力 KPI
[CP008, CP018, CP020, CP034]

3.4 护城河持久性与商品化风险

Shield AI 的竞争护城河建立在四根相互强化的支柱上:(1)GPS 拒止自主能力的作战深度——这一能力需要多年基于真实部署的强化学习积累,无法快速复制;(2)A-GRA 认证优势——已入选空军 CCA 项目任务自主能力供应商,Shield AI 的 Hivemind 成为竞争对手需要对标认证的事实参考实现;(3)平台无关授权模式——无论制造商是谁,每架搭载 Hivemind 的飞机均可带来版税收入;(4)Forge/Commander 产品层——通过集成任务规划和任务复盘工具,在中队工作流中形成应用层粘性。 主要商品化风险在硬件层,而非软件层。大疆在商业无人机硬件领域的主导,以及中国制造商在 FPV 无人机市场快速取代竞争对手的路径,表明无人机硬件一旦制造规模扩大便会快速商品化。Shield AI 的 V-BAT 和 X-BAT 所在市场,中国设计的第二组和第三组无人机受 NDAA 第 848 条限制被禁止采购,但底层硬件经济性仍将随时间推移压缩利润率。Hivemind 的软件授权模式从结构上规避了硬件商品化冲击,因为它按平台收费,硬件越便宜意味着授权机会越多,而非利润压缩。 更长期的商品化威胁来自潜在进入者:(1)大型科技公司——Microsoft、Google 和 Amazon 具备强大的 AI/ML 能力和现有的国防部云合同(JWCC),但缺乏实战部署信誉和 ITAR 合规基础设施;(2)中国自主系统——珠海紫燕和大疆子公司已展示自主蜂群能力,但出口限制和安全顾虑实际上将其排除在美军市场之外;(3)国防部自建——DARPA 和 AFWERX 资助了内部自主能力研发(ACE 项目、OFFSET 项目),但国防部 2024—2026 年的采购行为明确表明,在平台 AI 层倾向于采购而非自建。 一项不利评估:Shield AI 的护城河是真实存在的,但比 Hivemind 的营销宣传所暗示的要窄。Shield AI 公开赞颂的 A-GRA 开放架构授权,同时也是竞争风险。任何完成 A-GRA 认证的自主能力软件栈,理论上都可以在任何平台上替换 Hivemind。Anduril 的 Lattice 已通过传感器融合层的 A-GRA 认证,General Atomics 和 Northrop 也在推进各自软件栈的认证。若 A-GRA 认证逐渐成为常规操作,切换成本护城河将从"18—36 个月"压缩至"6—12 个月",从根本上改变竞争持久性方程式。 [CP020, CP021, CP027, CP028, CP029, CP033]

护城河持久性 / 竞争风险登记表
护城河主张威胁向量严重性缓解措施尽调问题
GPS 拒止实战部署可信度Anduril 或 DARPA 取得同等实战记录持续在乌克兰及 SOCOM 项目中部署;加快强化学习(RL)训练周期核实 Hivemind 在 FY2026 保密项目中的部署记录;索取实战事件报告
A-GRA CCA 任务自主认证Anduril、Northrop 或新进入者在相同平台上完成 A-GRA 认证高(结构性)通过更快迭代保持认证领先;嵌入更多 CCA 增量平台追踪 Anduril Lattice、Northrop Prism 及 DARPA 项目的 A-GRA 认证进展
平台无关授权模式美国国防部要求开源自主技术栈,或 DARPA 发布参考实现低(5 年展望期)Hivemind 的商业秘密训练数据(来自 10 年以上实战部署的强化学习积累)即使架构开源也无法复制追踪美国国防部 AI 开源政策动向(DIU、CDAO)
Hivemind 15+ 平台部署(切换成本)随着标准成熟,A-GRA 认证周期从 18–36 个月压缩至 6–12 个月以快于标准成熟的速度增加平台深度(资质数量);在竞争对手完成认证前赢得项目索取 Shield AI 新平台资质项目管道(FY2026–FY2028)
与 Anduril 的合作竞争关系(Fury + Hivemind)Anduril 开发竞争性边缘自主层并终止 Hivemind 访问权限中高A-GRA 对美国国防部的合同义务保障 CCA 平台可访问性;Shield AI 并不完全依赖 Anduril 硬件审查美国国防部 CCA 项目合同中对自主性提供商互操作义务的要求
国防主承包商合作壁垒(Boeing、L3Harris)主承包商在指挥层功能中采用 Palantir 或 Google AI 取代 Hivemind低至中Hivemind 差异化体现在平台层(而非指挥层);与 Palantir 重叠有限确认 Hivemind 与美国国防部 JADC2 框架及 CJADC2 试点项目的集成架构

严重性评级为基于公开证据的定性评估,而非 Shield AI 内部预测。"高(结构性)"指威胁内嵌于市场架构之中,无论 Shield AI 采取何种行动均需主动应对。

3.5 重要发现

Chapter 04

04财务情况

4.1 收入模式与收入来源

Shield AI 的收入来自两条相互关联的业务线:硬件系统销售和软件自主授权。硬件业务以 V-BAT(以及 2025 年 10 月发布的 X-BAT)为核心——向美国军方及盟友出售垂直起降无人机,单机均价估计约 100 万美元。软件业务 Hivemind Enterprise 包含四个模块:EdgeOS(核心自主操作系统)、Hivemind Pilot(自主飞行)、Hivemind Commander(多智能体编排)和 Hivemind Forge(机器学习模型开发环境)。Hivemind 同时向政府运营商及第三方平台集成商授权,合作方包括 Airbus、Kratos 和 L3Harris。 2024 财年(截至 2024 年 3 月),Shield AI 报告总收入约 2.67 亿美元,较 2023 年的 1.63 亿美元增长 64%。截至 2025 年 3 月的财年收入约为 3 亿美元,低于内部 4 亿美元目标约 1 亿美元,当年增速大幅放缓至 12%。管理层预计 2026 年收入将超过 5.4 亿美元,意味着在合同交付与国际扩张驱动下,增速有望回升至约 80%。 Hivemind 软件约贡献 2025 年收入的 30%,目标是在 2028 财年达到 50%。其余收入来自 V-BAT 硬件销售、服务(安装、培训、后勤支持)及运营合同。2025 年逾 50% 的收入据报道来自海外国防部,客户涵盖罗马尼亚、日本、希腊、加拿大、荷兰、乌克兰和埃及,反映出多元化但仍高度依赖政府的客户结构。 收入集中风险较高:商业模式几乎完全依赖政府采购周期,大多数合同以美元计价,采用固定价格或 IDIQ 形式。目前尚未披露任何商业或两用收入。国防部及盟国既是近期增长的天花板,也是基础保障。 [CI001, CI002, CI003, CI004, CI005, CI006]

收入来源表
收入来源产品 / 合同类型估计占比(2025)利润率概况主要合同 / 客户
V-BAT 硬件销售固定价格政府合同~50%20–35% (est.)USCG $198M IDIQ,美国国防部,盟国海军
V-BAT COCO 服务承包商持有、承包商运营(COCO)~20%30–45% (est.)USCG、日本、罗马尼亚
Hivemind 软件授权平台 / 任务集授权~30%60%+ (est.)Anduril CCA YFQ-44A、Airbus、Kratos 等竞争方
X-BAT 硬件(新兴)固定价格政府合同<1%TBD美国空军评估,2025 年 10 月公布
专业服务 / 集成时间与材料(T&M)/ 里程碑~5%低至中美国国防部,盟国国防部

收入占比百分比为基于分析师报告(Sacra、Acquinox Capital)及公司披露的估算;Shield AI 未公开具体的分部收入数据。部分报告中,COCO 服务计入硬件分部。

FI001: 收入模型桥接图
[CI001, CI002, CI003, CI007]

4.2 定价策略与市场拓展路径

Shield AI 主要通过直接政府合同而非商业渠道开展销售。其市场拓展路径依托国防创新机制——从其他交易授权(OTA)工具和国防创新部门(DIU)合同起步,再逐步迈向完整的国防采购项目。这种方式降低了早期销售阻力,但形成了政府采购典型的漫长回款周期:合同授予往往早于收入确认十二至三十六个月。 V-BAT 硬件以固定价格合同和承包商自持、承包商运营(COCO)服务安排出售,例如 2024 年 7 月美国海岸警卫队授予的 1.98 亿美元不定量不定期(IDIQ)合同。在 COCO 安排下,Shield AI 保留平台所有权,按飞行小时或持续监视可用性收费——该模式提升了经常性收入的可预测性。Hivemind Enterprise 按平台类型和任务集授权,但确切的每单位或每席位定价未公开披露。 获客成本(CAC)指标、销售周期长度(超出国防部多年采购惯例的部分)及回本周期均未披露。国防软件平台的行业类比表明 CAC 较高,但合同终身价值同样很高。Shield AI 的协同作战飞机(CCA)方案中,其 AI 技术栈被选为 Anduril YFQ-44A Fury 的核心系统,这使 Shield AI 成为嵌入式供应商——一旦 CCA 项目全面投产,Hivemind 授权收入有望规模化持续增长,形成平台锁定效应。 公司的国际扩张通过直接对接各国国防部关系推进,辅以本地集成合作伙伴。与罗马尼亚、日本、希腊等国的合同合计占 2025 年收入超过一半,但具体构成不透明,并受对外军售(FMS)监管时效影响。销售效率指标(人均收入、客户经理负责的年经常性收入)尚无公开数据。 [CI008, CI009, CI010, CI011, CI012]

定价 / 商业化模式表
产品定价模式估计单价 / 许可价格合同载体市场进入渠道
V-BAT(购买)固定单价约每架无人机 100 万美元直接面向国防部(DoD)/ 对外军售(FMS)直接政府合同
V-BAT(COCO)可用性 / 飞行小时费未披露IDIQ / 其他交易授权(OTA)直接面向 DoD、USCG
Hivemind Pilot按平台软件许可未披露OTA / 传统采购直接销售 + OEM 集成
Hivemind Commander按任务编排许可未披露OTA / 传统采购直接销售 + 协同作战飞机(CCA)项目
Hivemind Forge开发者 / 集成商许可未披露商业 + 国防部小企业创新研究(SBIR)直接面向集成商
Hivemind EdgeOS按平台运行时许可未披露嵌入 Hivemind EnterpriseOEM / 集成商

V-BAT 单价(约 100 万美元)为国防媒体行业估计值;实际定价因合同而异,不对外公开。Hivemind 许可价格未披露。COCO 可用性费用未披露。

4.3 成本结构与单位经济模型

Shield AI 的成本结构同时承载硬件制造商与软件研发公司的双重负担。V-BAT 硬件成本涵盖机身制造、推进系统、传感器载荷及供应链物流——同类平台国防硬件毛利率通常在 20%—35% 之间。Hivemind 软件毛利率据行业自主系统软件基准估计超过 60%,但 Shield AI 未披露分部毛利率。 2024 财年研发支出约占收入 22%,体现了公司对 Hivemind 国防基础模型及 X-BAT 研发项目的大力投入。这一研发强度与 IPO 前夕寻求建立平台主导地位的国防科技公司一致,但随着规模扩大,该比例需压缩以实现可接受的盈利水平。 2026 年 3 月宣布的收购 Aechelon Technology——与 G 轮融资同步披露——以未披露价格引入了仿真和合成环境能力。Aechelon 的技术为美国空军联合仿真环境(JSE)提供支持,对于用合成数据训练 Hivemind 基础模型至关重要。此次收购带来额外人员与整合成本,将在 2026—2027 年形成短期利润压力,协同效应需待此后才能显现。 营运资本动态对 Shield AI 略为有利:政府合同通常提供里程碑付款,相比商业 SaaS 可缩短现金转换周期。但 V-BAT 的 COCO 部署模式需要将资本锁定于已部署平台。资本支出数据尚未公开。截至 2026 年初,Shield AI 员工约 1,422 人,按 2025 年收入水平估算人均收入约 18.8 万美元——低于纯软件公司基准,但与这一规模的软硬件混合型企业相符。 [CI013, CI014, CI015, CI016, CI017, CI018]

单位经济学表
指标估计值 / 状态依据置信度
V-BAT 单台平均售价(ASP)~$1M多份合同报告
硬件毛利率20–35%(估计)国防硬件行业基准低(未披露)
Hivemind 软件毛利率60% 以上(估计)软件平台行业基准低(未披露)
综合毛利率未披露无公开文件None
每名员工收入(2025 年)~$188K$267M / 约 1,422 名员工
研发占收入比例~22%分析师报告(引用公司数据)
销售周期12–36 个月(典型 DoD 采购)DoD 采购规范
获客成本(CAC)未披露无公开文件None
积压订单未披露无公开文件None

大多数指标均为估计值或行业基准;Shield AI 不公开毛利率、获客成本(CAC)或积压订单。每名员工收入根据 2024 财年收入 (2.67 亿美元)/ 员工总数(约 1,422 人)计算。硬件 / 软件利润率估计基于国防行业规范。

FI002: 单位经济模型桥接图
[CI013, CI014, CI015, CI016]

4.4 资本结构与财务状况

截至 2026 年 4 月,Shield AI 自 2015 年成立以来已累计完成 17 轮以上融资,合计约 36 亿美元。2026 年 3 月的融资事件是公司史上规模最大的一次,分为三个部分:15 亿美元 G 轮股权(由 Advent International 领投,JPMorganChase Strategic Investment Group 联合领投,Snowpoint Ventures、InnovationX、Riot Ventures、Disruptive 和 Apandion 参与投资);Blackstone 旗下基金提供的 5 亿美元固定回报优先股;以及 2.5 亿美元延迟提款信贷额度(同样来自 Blackstone,供未来使用)。本轮融资后投后估值为 127 亿美元,较一年前确立的 53 亿美元估值上升 140%。 Blackstone 优先股结构在经济上更接近准债务:Blackstone 获得固定票息,且优先股在结构上优先于普通股。加上 2.5 亿美元延迟提款额度,Blackstone 总敞口最高达 7.5 亿美元。这一结构使 Shield AI 能在股权稀释幅度不成比例的情况下获取大量资本,同时为后续收购保留灵活性。管理层表示,资金的主要用途为收购 Aechelon、持续推进 Hivemind 开发、V-BAT 和 X-BAT 规模化生产,以及国际市场拓展。 烧钱速度和现金状况未公开披露。然而,在累计融资 36 亿美元、2026 年收入运行率接近 5.4 亿美元的背景下,公司似乎拥有足够的现金跑道,可实现或超越管理层为 2028 财年设定的 10 亿美元收入里程碑。该里程碑是否与 IPO 筹备同步推进,以及市场将对盈利水平提出何种要求,仍是核心财务尽调问题。 公司未能实现 2025 年收入指引(目标 4 亿美元,实际约 3 亿美元),差距约 1 亿美元,接近 25%。这是首次公开记录的指引落空,令外界对 5.4 亿美元以上 2026 年预测的可靠性产生疑虑。多位分析师指出,在 127 亿美元估值下,Shield AI 对应预计 2026 年收入约 27—28 倍,该倍数预设了激进的执行力以及 Hivemind 高毛利软件成功转型为主要收入驱动力。 [CI019, CI020, CI021, CI022, CI023, CI024]

资本充足性表
融资轮次 / 批次金额投资方估值日期类型
A 轮$10M(估计)DIU + 早期风险投资机构未披露2016股权
B 轮$20M(估计)Andreessen Horowitz、Shield Capital未披露2019股权
C 轮$210MSnowpoint Ventures、DC VC$2.3B2021股权
D 轮$165M多家机构$2.7B2022股权
E 轮$200M多家机构$2.7B2023股权
F 轮$200M多家机构$5.3BMar 2025股权
G 轮股权$1,500MAdvent International、JPMorganChase$12.7BMar 2026股权
Blackstone 优先股$500MBlackstone 管理基金固定回报Mar 2026优先股
Blackstone 延迟提款$250M(可用)BlackstoneN/AMar 2026信贷额度

早期融资轮次金额(A 轮、B 轮)为二手来源的近似估计;早期轮次的确切数字未经公开核实。G 轮组成部分(15 亿美元股权、5 亿美元 Blackstone 优先股、2.5 亿美元延迟提款)来源于 Business Wire 和 Tech Insider。早期轮次估值推断自 Tracxn 和 Crunchbase 二手数据。

FI003: 财务估算区间
[CI020, CI021, CI024, CI025]
FI004: 资本密集度/现金流地图
[CI019, CI022, CI023, CI028]

4.5 财务信息缺口与尽调障碍

Shield AI 是一家私有公司,没有披露财务报表的义务,且对于这一规模的企业而言,其财务披露控制异常严格。这在核实完整投资逻辑所需的每一项关键指标上,都构成重大尽调障碍。最关键的缺口在于没有任何已披露的毛利率——无论是综合毛利率还是分部毛利率。缺乏毛利率数据,根本无从评估硬件—软件混合业务的真实经济性,也无法验证 Hivemind 毛利率提升的逻辑。 其他未披露指标还包括:按职能划分的运营支出(研发、销售及营销、管理费用,超出已知 22% 研发比例之外的部分)、EBITDA 或调整后 EBITDA、净利润或净亏损、现金及等价物,以及烧钱速度。收入确认政策——尤其是 COCO 安排和多年期 IDIQ 合同的确认政策——未公开记录,可能产生影响期间业绩的收入确认时间差异。 公司披露了 2024 年 2.67 亿美元的收入,但未提供地区细分、分部毛利率或积压订单数据。1.98 亿美元的海岸警卫队 IDIQ 及其他已命名合同只是实际订单的一部分;完整的积压订单数字将大幅提升核实信心。单位经济指标——具体包括 V-BAT 每飞行小时成本、Hivemind 授权毛利率和获客成本——在公开披露中完全缺失。 对于一家估值 127 亿美元、预计 2026 年收入 5.4 亿美元的企业,Shield AI 财务信息的不透明程度异常之高。参与尽调的投资者需要获取经审计的财务报表、股权结构表明细(包括期权池规模、优先股清算优先权及 Blackstone 优先股条款)、按项目划分的详细积压订单,以及分产品线的营业成本,才能据此建立具有可信度的估值模型。 [CI026, CI027, CI028, CI029, CI030]

公开财务信息缺口表
数据点已披露?尽调影响替代方案 / 代理数据
综合毛利率无法评估单位经济学或利润率扩张空间仅限行业基准
分部毛利率无法验证硬件与软件业务结构转变的投资逻辑None
各职能运营费用部分(研发约占 22%)无法建立损益或烧钱速度模型研发占比 22% 估计
净利润 / 净亏损无法评估盈利路径或现金消耗None
现金及等价物在缺乏烧钱速度背景下无法评估现金跑道None
烧钱速度无法评估资本充足性或下轮融资触发条件None
按地区划分的收入部分(国际收入占比 >50%)无法评估按战区的收入集中度部分定性信息
积压订单 / 合同总价值(TCV)无法预测收入确认时间仅限已命名合同
收入确认政策无法评估 COCO/IDIQ 收入时间None
股权结构表细节无法评估稀释、优先权或收益分配None
Aechelon 收购价格无法评估商誉 / 无形资产负担None
各项目资本支出(CapEx)无法评估资本密集度None

本表反映截至 2026 年 5 月基于公开来源的数据可用性。Shield AI 未发布经审计的财务报表,所有缺口均通过公开文件、新闻稿及媒体报道中数据缺失加以确认。

4.6 附录

Chapter 05

05产品与技术

5.1 产品矩阵概览

Shield AI 采用双层产品架构:软件自主平台(Hivemind Enterprise)与硬件产品线(V-BAT、X-BAT),辅以传感器系统 ViDAR。Hivemind Enterprise 是商业重心——一套中间件与任务 AI 技术栈,设计上可运行于从一类无人机到海军舰艇和地面车辆的任意平台。截至 2026 年初,公司宣称已在 15 种以上不同平台上完成部署,生产集成包括 V-BAT MQ-35A、MQ-20 Avenger(实时自主测试),以及通过 HII 合作的 ROMULUS 无人水面舰艇。V-BAT Group-3 VTOL 无人机是主要的硬件收入驱动力(均价约 100 万美元,续航超 13 小时,Block Upgrade 配置搭载重油发动机),并获得美国海军 MQ-35A 正式型号命名,标志着正规采购路径取得进展。2025 年 10 月发布的 X-BAT——定位为全球首款 AI 驾驶 VTOL 战斗机——展示了 Shield AI 进军 Group-5 高对抗空战领域的雄心。ViDAR 传感器产品在 GPS 拒止条件下提供远程视觉探测与测距能力,为整条产品线拓展了平台无关的感知能力。这一产品广度使 Shield AI 定位为垂直整合的国防自主化公司,而非单纯的软件 ISV。 [CE001, CE002, CE003, CE004, CE005, CE006]

产品模块 / 资产矩阵
产品 / 模块用户 / 买方状态 / 成熟度核心差异化尽调缺口
Hivemind PilotDoD 操作员、盟友部队生产阶段(15 个以上平台)GPS 拒止环境下的 SLAM/EO-IR 自主飞行尚未确认独立安全认证
Hivemind CommanderUSAF、海军、盟友操作员生产阶段(多平台演示)多智能体蜂群协调蜂群规模上限未披露
Hivemind ForgeShield AI 工程师、集成商生产阶段(AWS ECR 流水线)任务 AI 模型开发与部署尚未确认外部开发者访问权限
EdgeOS平台集成商、OEM生产阶段(插件 SDK 已部署)亚毫秒进程间通信(IPC),跨平台运行时插件生态广度未披露
V-BAT(MQ-35A)美国海军、USCG、盟友国防生产阶段(已列装,获 MQ-35A 编号)第三组垂直起降(VTOL),续航 >13 小时单位经济学 / 利润率未披露
X-BATUSAF、盟友空军开发阶段(首飞计划于 2026 年秋)第五组超声速垂直起降战斗机成本、进度及性能待定(TBD)
ViDAR平台集成商、情报监视侦察(ISR)客户已发布(产品页面上线)无 GNSS 视觉探测与测距灵敏度 / 探测距离规格未公开

成熟度等级基于公开运营证据和公司披露;"生产"表示已向付费客户实际部署。

FE001: 产品架构图
[CE008, CE009, CE010, CE016, CE025]

5.2 Hivemind 平台与 EdgeOS 架构

Hivemind Enterprise 在共享运行时层之上组织为四个功能模块。EdgeOS 是实时操作基底:采用静态 JSON 配置管理(规避 ROS 依赖),共享内存进程间通信每秒可处理数百万条消息且延迟低于 1 毫秒,插件 SDK 支持跨异构载具类型的平台无关部署。这一运行时架构允许通过编写硬件抽象插件完成新平台接入,而无需修改任务逻辑。Hivemind Pilot 通过同步定位与建图(SLAM)、视觉里程计和光电/红外传感器融合,在 GPS 拒止环境中实现自主飞行与导航——无需 GNSS 依赖。Hivemind Commander 将单机自主扩展至多智能体集群协同,支持在复杂任务中同时调度多个自主平台。Hivemind Forge 提供机器学习开发流水线:运营商和工程师在云端环境(与 AWS 合作,基于 ECR 的容器分发)中训练、评估并打包任务 AI 模型,再将其部署至边缘载具。公司的国防基础模型以国防任务数据而非通用互联网语料为训练基础,Shield AI 认为这对于在对抗环境中实现可靠性是必要的,通用大语言模型无法满足该要求。AWS 联合开发关系支持云端到边缘的工作流:在 Forge 中开发的模型经容器化后通过 ECR 推送至载具机群,支持大规模机群自主管理。与实际客户验证的集成时间线——MQ-20 Avenger 不到 12 周、ROMULUS 无人水面舰艇 6 周——是平台集成便捷性主张的主要佐证。 [CE008, CE009, CE010, CE011, CE012, CE013]

工作流 / 应用场景表
用户任务当前 / 原有工作流Shield AI 解决方案可量化收益局限性
持续 ISR(海上)有人驾驶 P-8 或承包商运营的无人机系统(UAS)V-BAT + Hivemind 实现舰载自主 ISR>13 小时续航;USCG 1.98 亿美元合同气象 / 海况包线限制
GPS 拒止飞行作战被迫停飞或依赖 GPS 的无人机系统机队Hivemind Pilot SLAM/EO-IR 导航可在电磁频谱争夺环境中运行SLAM 在无特征环境中的精度尚不明确
多域协同作战孤立的单机无人机系统作战Hivemind Commander 多智能体协调同时编排多个平台跨域(空中 + 海上)延迟未披露
平台 AI 接入多年期定制自主集成EdgeOS 插件 SDK + Hivemind 模块MQ-20 Avenger 在 12 周内完成集成需要 OEM 合作及数据权利

工作流描述基于公开集成案例研究和公司说明;量化收益为公司自报或根据集成时间线估算。

技术 / 运营架构表
层级 / 组件功能定位关键依赖技术风险
EdgeOS(运行时)实时操作系统底层;共享内存进程间通信(IPC);静态 JSON 配置硬件插件 SDK;平台 OEM 合作插件生态质量;大规模 SDK 可移植性
Hivemind Pilot自主飞行;通过 SLAM + EO/IR 实现 GPS 拒止导航EO/IR 传感器;惯性测量单元(IMU);ViDAR(可选)低特征环境中的 SLAM 漂移
Hivemind Commander多智能体协调;蜂群编排低延迟通信链路;所有智能体均运行 EdgeOS通信拒止下的蜂群韧性未有记录
Hivemind Forge + AWS机器学习(ML)模型训练、打包及 ECR 分发AWS 云基础设施;接近 FedRAMP 标准的安全措施对 AWS 的供应商锁定;ITAR 数据处理合规要求
国防基础模型用于任务 AI 微调的领域专用基础模型自研任务数据;安全计算资源训练数据质量与覆盖范围未披露
ViDAR 传感器无 GNSS 的远距离目标探测与测距光学元件供应链;与 EdgeOS 的集成性能规格未经公开验证

架构细节基于 Shield AI 技术出版物和 AWS 合作博客;内部实现细节未经独立核实。

FE002: 客户工作流/运营流程
[CE013, CE014, CE015, CE039]

5.3 硬件产品与路线图

V-BAT(MQ-35A)是 Group-3 垂直起降无人机,采用涵道风扇设计,Block Upgrade 配置搭载重油发动机,续航超 13 小时。Block Upgrade 通过新闻稿发布,新增 Hivemind 增强自主能力、SATCOM 连接和重油发动机作为渐进式能力包。均价约 100 万美元的 V-BAT 以低成本持久 ISR 和协同作战平台定位参与竞争。X-BAT 于 2025 年 10 月 21 日发布,是 Shield AI 首款 Group-5 平台:一款翼展 26 英尺(展开后达 39 英尺)、实用升限超过 50,000 英尺、航程超过 2,000 海里的超音速 AI 驾驶 VTOL 战斗机。其动力来自 GE F110-GE-129 发动机,配有轴对称矢量排气喷管(AVEN),实现推力矢量超音速 VTOL 操作。Shield AI 与 GE Aerospace 于 2025 年 10 月签署推进系统合作协议,目标是在目标单机成本约 2700 万美元的基础上,推动 X-BAT 在 2026 年秋季完成首飞。2026 年 3 月收购 Aechelon 为产品矩阵增加了合成环境与仿真能力,可在逼真虚拟环境中完成部署前训练——对于压缩 X-BAT 测试周期、降低实飞测试成本至关重要。Shield AI 的产品路线图由此从近期 V-BAT 机群扩张,延伸至中期 X-BAT 首飞,并最终指向量产级自主战斗机平台。 [CE017, CE018, CE019, CE020, CE021, CE022]

路线图 / 发布 / 开发阶段表
日期 / 阶段里程碑 / 功能状态影响来源
2025 Q1V-BAT 批量升级:Hivemind 自主能力 + SATCOM + 重油发动机已发布(PR Newswire 公告)巩固 V-BAT 竞争地位;为现有机队增加传感器融合自主能力SE011
2025 Q4X-BAT 亮相:Group-5 超音速 VTOL 战斗机,目标单机成本约 $27M已宣布(2025 年 10 月 21 日)彰显 Group-5 野心;战斗机级细分市场首款产品SE013, SE024
GE Aerospace 推进系统合作GE F110-GE-129 发动机及 AVEN 协作正式落地协议已签署验证推进路径;降低单一供应商发动机风险SE025
2026 Q1Aechelon 收购:仿真与合成环境能力已完成(2026 年 3 月)借助合成训练加速 X-BAT 与 Hivemind 测试周期SE015
2026 Q3-Q4X-BAT 首飞(目标)计划中关键技术里程碑;降低 Group-5 硬件项目风险SE007, SE024
2027+X-BAT 量产及美国国防部项目整合路线图(未确认)研发与认证周期漫长;存在资金与项目风险SE003

时间表和成本估计基于公司公告;首飞和量产日期均为目标日期,可能发生变化。

FE003: 关键依赖项地图
[CE015, CE018, CE020, CE023, CE026, CE027]

5.4 差异化优势与知识产权

Shield AI 的核心技术护城河是通过 SLAM 和光电/红外传感器融合实现 GPS 拒止环境下的自主能力——这一能力使配备 Hivemind 的平台能够在 GNSS 信号被干扰或不可用的电磁环境中可靠运行,而依赖 GNSS 的竞争对手则无法做到这一点。国防基础模型进一步巩固了这一优势:该模型以国防专属任务数据而非消费互联网语料为训练基础,Shield AI 认为通用大语言模型在对抗环境可靠性方面无法企及这种领域专属性能。Hivemind 软件与自研硬件(V-BAT、X-BAT)及传感器(ViDAR)的垂直整合构建了平台级壁垒:Shield AI 可优化硬件—软件协同设计,展示端到端任务能力,并规避纯软件供应商被商品化的风险。Hivemind 作为美国空军自主协同平台(A-GRA)项目主力自主技术栈的角色,进一步验证了其执行关键任务集成的技术可信度。第三方分析师将 Shield AI 定性为领先的纯国防自主 AI 软件公司,与将自主能力作为附加功能而非核心产品的主承包商明显区别。公司快速集成的实绩——以周计而非以年计——本身就是一种以系统集成方法论和工具链为载体的知识产权。鉴于 Shield AI 自创立起即聚焦自主飞行,其知识产权组合推断应相当丰厚,但具体专利数量和覆盖范围目前尚未公开披露。 [CE024, CE025, CE026, CE027, CE028]

FE004: 产品成熟度/能力地图
[CE001, CE026, CE028, CE030, CE037]

5.5 可信度、安全性与合规

Shield AI 在自主武器系统须遵守《国防部指令 3000.09》的领域运营——该指令要求对致命决策保留人员授权,这意味着当前配备 Hivemind 的平台在交战决策上仍保持人在回路(或人在环外),即便导航和飞行已完全自主。这一监管框架在国防部采购体系内已被充分理解,目前并不构成 ISR 和协同任务(V-BAT、海上任务)的实质性障碍。V-BAT MQ-35A 型号命名表明通过 NAVAIR 的正式军事鉴定正在推进,这是国防采购的可信度信号。Shield AI 与 AWS 的合作将类 FedRAMP 云安全实践延伸至 EdgeOS 开发环境,是联邦客户重要的合规考量。Hivemind 的 GPS 无关架构天然具备抗干扰韧性,在对抗环境中是一项安全优势。然而,尽调发现若干关键缺口:截至 2026 年初,Hivemind 或 V-BAT 尚无经第三方独立认证的适航资质(DAL A/B 等效)在审阅的公开文件中得到确认。Shield AI 未公开披露安全漏洞评估、渗透测试结果或正式安全认证时间表。X-BAT 在对抗民用和拥挤空域中的自主运行将需要超出现有无人机框架的监管批准,这一过程可能延长时间表,并需要美国联邦航空局(FAA)、国防部及国际层面的协调。这些缺口对于评估公司在 IPO 前合规准备度的机构投资者而言至关重要。 [CE029, CE030, CE031, CE032, CE033, CE034]

信任 / 质量 / 合规表
管控措施 / 认证状态适用范围缺口 / 尽调请求
DoDD 3000.09(人在回路)合规(DoD 使用要求)所有致命性自主武器系统确认交战授权架构
MQ-35A 军事资质认证进行中(已获海军编号)V-BAT NAVAIR 资质认证完整的作战使用授权(ATO)/ 生产授权时间表
接近 FedRAMP 的云安全部分(AWS GovCloud 合作)Hivemind Forge 机器学习开发环境完整 FedRAMP 授权尚未确认
适航认证(Hivemind / V-BAT)未经公开确认自主飞行软件及机体请求 DAL A/B 认证状态
安全漏洞披露未公开Hivemind 软件及 EdgeOS请求渗透测试结果;CVE 政策

合规状态仅基于公开信息;未确认认证并不意味着不合规——认证细节可能属于保密信息或尚未公开。

5.6 附录

Chapter 06

06客户情况

6.1 客户群体细分

Shield AI 的客户群分为三个主要细分,各自代表不同的采购机制和收入属性。美国国防部终端用户是最成熟的细分:美国海岸警卫队在 1.98 亿美元 IDIQ 合同下以全量产模式运营 V-BAT;美国海军授予 V-BAT MQ-35A 型号命名并正在推进 NAVAIR 鉴定;美国空军将 Hivemind 作为 A-GRA 项目主力自主技术栈,并已完成 MQ-20 Avenger 实飞自主测试。盟国国际客户细分涵盖 10 个以上国家,其中仅日本和罗马尼亚为已公开命名的部署案例,通过对外军售(FMS)或直接渠道采购 V-BAT。第三个细分——国防主承包商集成商——是最新兴但具有重要战略意义的:Huntington Ingalls Industries 在其 ROMULUS 无人水面舰艇上集成了 Hivemind,Anduril 则为其协同作战飞机 YFQ-44A 项目获得 Hivemind 软件授权。这一 B2B2G 渠道绕过直接政府采购时间线,使 Shield AI 的软件得以嵌入大型主承包商销售的平台中。在三个细分中,V-BAT 硬件驱动近期收入,而 Hivemind 软件授权是一个新兴的经常性收入组成部分,潜在单位经济效益更优。 [CU001, CU002, CU004, CU005, CU006, CU007]

客户细分表
客户细分代表性客户产品 / 服务收入重要性采购机制
美国国防部终端用户美国海岸警卫队、美国海军、美国空军V-BAT、Hivemind Enterprise(核心产品)高(已确认收入的多数)IDIQ、OTA、在册项目
盟国国际日本、罗马尼亚及 8 个以上未披露国家V-BAT中(增长中)FMS、直销
国防主承包商集成商HII(ROMULUS)、Anduril(CCA YFQ-44A)等项目Hivemind 软件许可中(新兴)软件许可、分包合同
项目级演示集成MQ-20 Avenger (USAF A-GRA)Hivemind 主技术栈低至中(演示阶段)项目集成

各细分收入占比根据公开合同及收入数据估算;Shield AI 不对外披露分部收入。"Prime Integrators" 包括将 Hivemind 嵌入平台后销售给最终政府用户的 B2B2G 渠道客户。

[CU012, CU017, CU022, CU028, CU030]
FU001: 客户旅程图
[CU012, CU018, CU027, CU030]

6.2 部署采纳轨迹

Shield AI 的增长指标表明其在各客户细分中的采纳呈现快速加速态势。2024 财年收入达 2.67 亿美元,较 2023 财年估算的约 1.63 亿美元增长 64%。管理层已指引 2026 年收入增长约 80%,隐含前瞻收入运行率接近 4.8 亿美元。V-BAT 部署已从少量试点项目扩展至 10 个以上国家的实战部署。MQ-35A 美国海军型号命名一旦转化为正式量产项目,将意味着 V-BAT 产量的重大跃升。A-GRA 项目与 MQ-20 Avenger 的集成在约 12 周内完成,表明 Hivemind 可快速接入新平台,降低了硬件和软件渠道的采纳阻力。HII ROMULUS Hivemind 集成在 6 周内完成,进一步验证了快速采纳时间线。Anduril 为 CCA YFQ-44A 获得 Hivemind 授权,是 Shield AI 软件首次已知部署于第三方主承包商自主作战飞机,确立了软件授权作为平行采纳渠道的地位。USASpending.gov 数据确认了来自多个联邦机构(美国海岸警卫队、海军、空军)的连续合同授予,以可观测的联邦采购活动(截至 2026 年 5 月)印证了收入增长轨迹。 [CU008, CU009, CU010, CU011, CU013, CU014]

客户增长 / 采用轨迹表
指标FY2022EFY2023EFY2024(已报告)FY2025EFY2026E(指引)
总收入~$93M est.~$163M est.$267M未披露~$481M(80% 指引)
同比收入增速N/A~75% est.64%未披露~80%(指引)
V-BAT 已部署国家数<5 est.~8 est.10+10+持续扩大
已确认美国国防部客户项目数234+4+持续扩大
国防主承包商软件被许可方001 (Anduril)1+持续扩大

2022 财年及 2023 财年收入数据为反推估算,未经公开文件确认。2025 财年预测未披露。2026 财年预测基于管理层指引:在 2024 财年已报告收入 $267M 基础上增长约 80%。

[CU008, CU010, CU013, CU020]
FU002: 采用/部署漏斗
[CU002, CU008, CU013, CU020]

6.3 已命名客户的验证依据

最有力的客户佐证是 2024 年 7 月美国海岸警卫队授予的 1.98 亿美元 IDIQ 合同,用于 V-BAT 量产交付和运营支持——这是 Shield AI 历史上已公开确认的单一客户承诺中规模最大的一笔。IDIQ 结构允许政府在合同履行期内发出任务订单,表明政府具有多年量产意图,降低了近期收入的可预测性风险。美国海军对 MQ-35A 的正式型号命名是第二个高可信度的佐证:正式军事型号命名仅在平台通过初始鉴定里程碑后方可授予,标志着海军的投入和采购意愿。美国空军 MQ-20 Avenger 实飞自主测试——Hivemind 在高对抗环境场景中操控飞机——代表了在已公开确认的最强能力空战平台上完成的项目级实战验证。HII 的 ROMULUS 无人水面舰艇集成在双方均发布新闻稿确认里程碑达成的情况下宣告完成,提供了 Hivemind 部署于海军水面舰艇的经第三方印证的合作方证明。Anduril 为 CCA YFQ-44A 项目获得的 Hivemind 授权是战略意义最重大的软件授权佐证,反映出一家国防主承包商对 Hivemind 自主技术栈的认可——该项目有望为美国空军大批量采购,形成大量自主平台。日本和罗马尼亚的部署虽未经独立第三方证实,但与 Shield AI 有据可查的盟国市场拓展及公司声称的 10 个以上国家部署足迹相符。 [CU001, CU003, CU005, CU006, CU007, CU009]

已命名客户证明表
客户合同 / 证据合同金额产品部署状态日期
美国海岸警卫队$198M IDIQ V-BAT 生产及支持$198MV-BAT (MQ-35A)已量产部署July 2024
美国海军MQ-35A 型号指定,NAVAIR 资质认证进行中未披露V-BAT (MQ-35A)认证进行中2024-2025
美国空军Hivemind 主技术栈用于 A-GRA;MQ-20 Avenger 实弹测试未披露Hivemind Enterprise实弹测试中2024
HII (Huntington Ingalls)ROMULUS USV 海上自主能力集成已完成未披露Hivemind Commander已集成至量产2024
Anduril IndustriesCCA YFQ-44A Hivemind 软件许可未披露Hivemind 软件许可服役开发中2024
日本(盟国)V-BAT 实战部署未披露V-BAT已部署Pre-2024
罗马尼亚(盟国)V-BAT 实战部署未披露V-BAT已部署Pre-2024

本表仅涵盖已公开确认的客户合作。Shield AI 不对外公布完整客户名单。非美国海岸警卫队客户的合同金额在公开记录中未披露。各行按证据强度排列。

[CU001, CU003, CU006, CU007, CU009, CU031]
FU003: 客户证明矩阵
[CU001, CU004, CU005, CU006, CU031]

6.4 客户留存与黏性

Shield AI 在已确认的美国国防部客户中留存信号强劲,但受限于有限的公开数据,难以进行完整的量化评估。美国海岸警卫队关系是最直观的指标:1.98 亿美元 IDIQ 代表了从早期采购接触逐步深化的过程,表明海岸警卫队评估了 V-BAT,认为其在运营上令人满意,并在多年窗口内承诺了全量产量。截至 2026 年 5 月,在所有审阅来源中均未发现任何公开的合同取消、项目终止或客户主动退出的证据。MQ-35A 项目持续推进 NAVAIR 鉴定——这是一个多年周期的过程——表明美国海军对 V-BAT 平台保持持续投入。Hivemind 的集成架构提供了结构性留存机制:一旦 Hivemind EdgeOS 嵌入客户平台生态,替换它需要重构任务 AI 层,产生大量切换成本。V-BAT Block Upgrade——将 Hivemind AI 自主能力、SATCOM 和重油发动机作为渐进能力包叠加——为现有 V-BAT 运营商创造了自然的增购路径,深化了平台依赖。Shield AI 在 2024 和 2025 财年的公开表述中均未提及项目终止,且公司引用的多年期 IDIQ 结构进一步降低了客户流失风险。6 周完成的 ROMULUS 集成及 HII 持续的合作伙伴参与,即便没有正式 NPS 或 CSAT 评分披露,也表明合作方客户满意度良好。 [CU024, CU025, CU033, CU034, CU036, CU037]

留存 / 复购 / 满意度表
客户首次合作最近合作续约证据留存信号
美国海岸警卫队2022 est.2024 年 7 月($198M IDIQ 授标)IDIQ 结构支持多年期任务订单
美国海军2022 est.2024–2025 年 NAVAIR 资质认证MQ-35A 项目持续推进
美国空军2023 est.2024 年实弹测试行动A-GRA 项目持续推进
HII (ROMULUS)20242024 年集成里程碑集成里程碑已达成;后续合作待定
Anduril (CCA)2024 est.2024–2025 年进行中CCA 为美国国防部多年期项目

留存评估基于可观察的合同续约及项目状态;Shield AI 未公开披露 NPS、CSAT 或流失率等正式指标。留存信号评级为定性评估,仅基于可获取的公开证据。

[CU025, CU033, CU034, CU037]
FU004: 留存/重复队列
[CU024, CU033, CU037, CU039]

6.5 扩张方向与客户集中风险

Shield AI 面临重大的客户集中风险。若按年化计算,美国海岸警卫队 1.98 亿美元 IDIQ 约占 2024 财年 2.67 亿美元收入的 74%——这种单一客户依赖在国防硬件早期公司中并不罕见,但若该项目暂停或重组,将暴露出战略脆弱性。值得注意的是,已公开确认的收入中约 100% 来自美国政府客户,使公司面临国防部预算封顶、持续决议(CR)或采购优先事项调整的风险。AI Invest 的一份分析指出,Shield AI 127 亿美元的估值在很大程度上依赖于尚未在已报告收入构成中体现的软件转型叙事,因为 Hivemind 授权收入在总收入中仍占少数,而公司已按软件公司倍数定价。在扩张方面,国防主承包商授权(Anduril)和盟国部署代表了有意义的多元化方向。Anduril YFQ-44A Hivemind 授权具有重要战略意义,因为协同作战飞机项目预计将被美国空军大批量采购,实际上将单一授权安排转化为事实上的多年高量间接收入流。向更多印太和欧洲北约盟友拓展对外军售(FMS),将进一步分散收入集中度。Shield AI 2026 年 3 月 20 亿美元的融资显示投资者对这些扩张叙事抱有信心,但软件转型的执行风险仍是合理的担忧。 [CU015, CU016, CU022, CU023, CU026, CU035]

扩张与集中度风险表
风险因素当前证据严重程度缓解方向
USCG 客户集中度USCG $198M IDIQ 若按年折算约占 2024 财年 $267M 收入的 74%向盟国及主要软件被许可方分散客户
美国国防部采购依赖已确认收入约 100% 来自美国政府机构拓展盟国 FMS 及商业主承包商许可渠道
软件转型执行风险$12.7B 估值依赖 Hivemind SaaS 规模化,而该部分尚未进入收入结构证明 Hivemind ARR 增速及许可数量里程碑
国际 FMS 延迟FMS 审批周期可能使盟国 V-BAT 交付推迟 12–24 个月为核心盟友开辟直销及授权生产渠道
MQ-35A 生产采购不确定性海军 V-BAT 生产采购数量未在公开记录中确认争取美国国防部多项目、多军种平台承诺

集中度风险估算基于 $198M 美国海岸警卫队合同及 2024 财年 $267M 收入。严重程度为定性评级。软件转型风险反映分析师评论,非公司指引。缓解策略源自公司定位与市场动态推断。

[CU015, CU016, CU022, CU023, CU040]

6.6 附录

Chapter 07

07风险

7.1 风险严重程度概览

Shield AI 的风险敞口涵盖六个主要维度。按已知缓释措施后的剩余严重程度排列:(1)财务/模型风险是最高剩余风险——127 亿美元 G 轮估值隐含激进的软件转型假设,但这些假设尚未体现在已报告的收入构成中,且 2025 年收入落空(实际约 3 亿美元,指引 5 亿美元以上)已损害了近期增长的可信度。(2)X-BAT 运营/技术风险较高——截至 2026 年 5 月该项目仍在首飞前阶段,涉及 Group-5 硬件复杂度,并存在 GE F110 单一来源推进系统依赖,尚无已公开披露的替代方案。(3)监管/法律风险为中高——ITAR/EAR 管制对每一项盟国客户合同均产生重大出口许可义务,《国防部指令 3000.09》限制了 Hivemind 在进攻性致命任务中的运营范围。(4)合作伙伴/依赖风险为中等——AWS 云依赖用于 Hivemind Forge,国防部预算周期敞口是真实存在的风险,但合同 IDIQ 结构提供了部分缓释。(5)人才/执行风险为中等——对曾氏联合创始人的关键人物集中依赖,以及计划从 1,400 人扩张至 3,000 人以上的人员规模增长,构成组织层面的脆弱性。(6)运营/安全风险为中低——V-BAT 或 Hivemind 均无已知公开的安全事件或可靠性故障,但缺乏公开认证披露限制了验证。投资含义:财务模型和 X-BAT 风险是主要估值门槛因素;其余风险类别在正常国防项目执行假设下虽然重要但可管控。 [CR001, CR007, CR010, CR011, CR024, CR025]

FR001: 风险热图
[CR001, CR008, CR010, CR011, CR014, CR016]
FR002: 风险传导图
[CR007, CR008, CR011, CR015, CR025, CR038]

7.2 监管与法律风险

Shield AI 的产品组合——V-BAT、Hivemind 和 X-BAT——所处的监管环境复杂且仍在演进。根据 22 CFR 第 120—121 条(ITAR),具备自主能力的军用无人机系统可能受《美国军品清单》第 XI 类(军事电子设备)或第 XV 类(航天器及相关物品)管辖。向盟国客户出口任何受 ITAR 管制的物品,均需国务院国防贸易管制局(DDTC)批准的技术援助协议(TAA)、制造许可协议(MLA)或 DSP-5 出口许可证。AI 自主软件的两用性还可能对某些组件适用出口管理条例(EAR)。DDTC 积压审批或政策变化可能延误或阻断盟国客户收入。《国防部指令 3000.09》(武器系统自主性)带来结构性政策风险:它要求自主系统的致命决策保留人员授权,限制了 Hivemind 在进攻性任务中的运营角色。对该指令的任何政策收紧——或新的自主武器条约——都可能限制 Shield AI AI 飞行员概念在作战任务中的可寻址范围。国防采购合规要求(FAR/DFARS)和 CMMC 网络安全认证要求在合同层面增加了监管合规负担。截至 2026 年 5 月,公开记录中未发现任何针对 Shield AI 的诉讼、执法行动、ITAR 违规或知识产权纠纷。Shield AI 的合规状况——包括 ITAR 注册、商品司法管辖权确定及 CMMC 认证级别——未公开披露,构成关键尽调缺口。 [CR001, CR002, CR003, CR005, CR006, CR027]

监管 / 法律风险登记册
风险框架 / 主管机构可能性影响缓解状态
ITAR 出口管制 —— V-BAT、Hivemind、X-BAT 可能属于第 XI/XV 类管制品;向盟国出口需 TAA/MLA/DSP-5 许可证22 CFR Parts 120-121 (ITAR), DDTC高 —— 延误或阻断盟国客户收入未知;ITAR 注册及许可证组合无公开披露
DoDD 3000.09 —— 自主致命决策须经人工授权;可能限制 Hivemind 进攻性作战范围DoDD 3000.09(武器系统自主性政策)高 —— 若政策收紧,可能压缩 Hivemind 进攻性任务的总可用市场部分缓解;Hivemind 架构被描述为兼容人工监督回路
EAR 两用物项管制 —— AI 自主软件或两用组件向特定盟国转让可能需 BIS 出口许可证15 CFR Parts 730-774 (EAR), BIS中 —— 为软件出口增加额外许可合规负担未知;BIS 商品分类未披露
DFARS/FAR 合规 —— 国防合同监管要求,涵盖 CUI 处理、成本核算及 CMMC 网络安全FAR 52.212-4, DFARS 252.204-7012, CMMC中 —— 不合规可能导致合同终止或被列入禁止名单未知;CMMC 认证级别未公开披露
FAA 无人机系统授权 —— X-BAT 民用空域运营需获得 Part 135/107 或型号合格证,超出现有豁免范围FAA 14 CFR Part 107/135、NAVAIR 适航要求高(X-BAT 适用)中 —— 延迟商业或盟国空域使用;军事用途可通过适航审批推进尚未启动;截至 2026 年 5 月 X-BAT 仍处于首飞前阶段
无已知诉讼或知识产权争议 —— 截至 2026 年 5 月公开记录中未发现执法行动美国一般法律、USPTO、联邦法院低 —— 当前未发现法律责任不适用 —— 截至尽调日期记录清白

监管风险基于公开的 ITAR、EAR、DoDD 3000.09 要求及国防采购法规;Shield AI 未公开披露完整合规状况。各行按对投资逻辑预估影响严重程度排列。

[CR001, CR002, CR003, CR004, CR005, CR006]

7.3 运营与技术风险

X-BAT 项目是 Shield AI 最受关注的运营与技术风险所在。截至 2026 年 5 月,该载具尚未完成首飞,公司披露的目标为 2026 年秋季。Group-5 平台复杂度——X-BAT 是配备 GE F110 涡扇发动机、推力矢量 VTOL 系统和自主任务 AI 层的超音速 VTOL 战机——引入了多个硬件—软件集成风险向量,与 Shield AI 在小型 VTOL 平台上的现有量产经验截然不同。X-BAT 量产时间表不确定,估计需要 10 亿美元以上的研发资本才能达到量产。该项目的 FAA 民用空域授权和 NAVAIR/DOT&E 军事适航认证路径明确但耗时多年。若首飞延迟至 2026 年秋季之后,将压缩收入产生时间线,加剧投资者对资本部署效率的担忧。Hivemind GPS 拒止 SLAM 导航在低纹理或特征稀疏环境(开阔沙漠、开阔海面)下的性能,在公开来源中缺乏独立表征。Commander 模块在通信拒止条件下的集群韧性尚未在运营规模下有公开文献记录。Shield AI 未披露 Hivemind 或 EdgeOS 的任何 SOC 2、FedRAMP、DISA APL 或同等安全认证。大规模 V-BAT 和 Hivemind 赋能平台机群的 OTA 软件更新生命周期管理未有公开描述,随着机群规模扩大,这构成潜在的运营连续性风险。公开记录中未发现已知的 V-BAT 运营故障、安全事故或可靠性驱动的合同行动,这是积极信号,但部分归因于国防部项目有限的公开报道。 [CR004, CR007, CR009, CR019, CR020, CR021]

运营 / 质量 / 安全风险登记册
风险类别可能性影响证据
X-BAT 首飞拖延至 2026 年秋后 —— 项目复杂度超出 Shield AI 既往硬件经验进度 / 硬件高 —— 推迟收入产生、加大资金消耗,显露执行能力缺口公司声称 2026 年秋季目标;无独立里程碑追踪可供核实
GE F110 单一来源推进系统 —— X-BAT 无公开替代推进方案供应链 / 硬件低至中高 —— GE 任何交付延迟将直接推迟 X-BAT 进度GE 与 Shield AI 合作已确认;具体供货条款未披露
X-BAT 资本支出超支 —— 开发成本估计逾 $1B,不确定性高;实际支出未披露财务 / 硬件高 —— 超支压缩 G 轮资金跑道,可能需要额外融资分析师估算(逾 $1B);公司未披露 X-BAT 项目预算
GPS 拒绝模式下 SLAM 性能边缘案例 —— 无特征或低纹理环境可能降低导航精度技术 / 软件Unknown中 —— 在开阔海域或沙漠环境中作战效能可能受限公开资料中无独立野外测试数据;公司声称具备 GPS 拒绝能力
Hivemind/EdgeOS 无公开安全认证 —— SOC 2、FedRAMP、DISA APL 状态未披露网络安全 / 软件中 —— 可能阻碍需安全认证的敏感美国国防部项目部署公开资料中未发现认证文件
OTA 更新生命周期未明 —— 在役 V-BAT 与 Hivemind 机群如何接收软件更新未公开记录运营 / 软件Unknown中 —— 不受控的更新周期可能引入机群级可靠性或安全风险公开资料中未发现相关文件;需列为尽调问询事项
通信拒绝模式下蜂群韧性 —— Commander 模块在通信降级环境下的行为未经独立验证技术 / 软件Unknown中 —— 在通信对抗预期的争夺性环境执行任务时存在关键缺口公司声称具备通信拒绝能力;未发现第三方验证

运营与技术风险按预估严重程度排列。部分风险——GPS 拒绝模式下 SLAM 边缘案例、通信拒绝环境中的蜂群行为、OTA 更新生命周期——无法从公开资料独立核实,需直接向公司开展技术尽调。

[CR007, CR008, CR009, CR019, CR020, CR021]

7.4 合作伙伴与依赖性风险

Shield AI 的运营和产品连续性依赖少数关键外部合作伙伴和结构性依赖。GE Aerospace F110 发动机是 X-BAT 的主要推进来源,构成单一供应商风险:若 GE Aerospace 无法按时交付发动机,X-BAT 项目时间线将因无备用替代而延误。GE 与 Shield AI 已公开宣布 X-BAT 项目的推进系统合作,但尚未披露任何具约束力的供应协议条款或应急采购方案。AWS 云服务似乎支撑着 Hivemind Forge 开发环境;AWS 访问中断或费用上涨可能影响 Shield AI 的软件开发流水线,但 AWS 的可靠性记录在一定程度上缓释了这一风险。国防部预算周期依赖是一个结构性风险:Shield AI 收入的主要部分来自美国政府客户,持续决议(CR)、项目取消或预算封顶可能减少或延迟合同任务订单。2026 年 3 月宣布的 Aechelon 收购引入了整合风险——时间表、成本和所收购团队的人员保留均未公开披露。美国海岸警卫队 1.98 亿美元 IDIQ 形成的客户集中度按年化计算超过 2024 财年收入的 70%,使该单一合同的任何暂停或重组对近期收入而言都是重大事件。Anduril 获得的 Hivemind 授权取决于 Anduril 成功执行 CCA YFQ-44A 项目,形成对竞争对手项目状态的间接依赖。上述任一合作伙伴或依赖方出现问题,都可能直接波及收入、研发时间线或估值倍数。 [CR008, CR012, CR013, CR014, CR015, CR026]

合作伙伴 / 依赖项风险登记册
依赖项类型关键程度单点风险缓解措施
GE Aerospace —— 为 X-BAT 提供 F110 涡扇发动机单一来源硬件供应商关键是 —— 无已披露的替代推进供应商合作协议已宣布;具体供货条款未公开
美国国防部预算 / 持续拨款决议 —— 已确认收入约 100% 来自美国政府客户客户 / 收入关键是 —— 预算持续拨款决议或自动减支直接影响任务订单执行IDIQ 合同结构提供一定缓冲;多元化战略未披露
美国海岸警卫队 $198M IDIQ —— 单笔最大已确认收入承诺客户集中度是 —— USCG 单项若按年折算超过 2024 财年收入的 70%V-BAT 实战记录降低取消风险;仍需拓展多元化
AWS Cloud —— Hivemind Forge 开发环境依赖云端 / 平台否 —— AWS 为通用服务,存在替代方案依赖关系未完全披露;AWS 可靠性部分缓解
Aechelon Technology —— 2026 年 3 月收购;仿真 / 环境工具链整合收购 / 人员否 —— 仿真工具为辅助性,非关键路径收购近期完成;整合时间表未披露
Anduril CCA YFQ-44A 项目 —— Hivemind 软件许可收入取决于 Anduril 项目执行间接客户 / 软件否 —— Anduril 拥有独立资金来源及美国国防部关系与 Anduril 战略协同部分缓解;YFQ-44A 生产数量无合同保障
NAVAIR / FAA 监管审批 —— X-BAT 适航性及民用空域授权监管 / 审批高(X-BAT 适用)是 —— 无替代审批路径适航流程成熟;在 X-BAT 完成首飞前进度风险仍存

合作伙伴与依赖项风险按重要程度排列。美国国防部预算依赖是整个美国国防市场的结构性特征,并非 Shield AI 特有,但其量级值得重点关注。Aechelon 整合时间表尤为不确定,因 2026 年 3 月收购公告较为近期。

[CR008, CR012, CR013, CR014, CR015, CR026]
FR003: 依赖关系图
[CR008, CR012, CR013, CR014, CR026, CR029]

7.5 财务与模型风险

Shield AI 在 2026 年 3 月完成 Series G 融资,估值 127 亿美元——这一估值水平对应的是软件公司倍数,但该公司目前收入主要来自国防硬件(V-BAT 销售)。AI Invest 及其他分析机构指出,维持这一估值需要到 2028 年软件收入占比达到约 50%,而这一转型在已披露数据中尚无任何实质进展。管理层给出的 2025 年 5 亿美元以上营收指引据报未能达成,分析师估计实际 2025 年营收约为 3 亿美元——前瞻指引与实际执行之间存在明显的可信度落差。Shield AI 尚未盈利,持续烧钱,正走向 IPO;Series G 所筹的 20 亿美元须同时支撑日常运营和 X-BAT 研发所需的大额资本支出(估计须逾 10 亿美元才能投入量产)。客户集中风险进一步放大了财务模型的不确定性:一旦美国海岸警卫队合同暂停或 DoD A-GRA 计划被重组,近期营收缺口可能加速烧钱,触发被动融资事件。公司未公开披露毛利率、EBITDA 或经营现金流数据;支撑当前估值所需的软硬件收入结构转型无法从公开数据中独立核实。IDIQ 合同下的大额预付款、里程碑付款及营运资本需求带来执行层面的财务复杂性。基于公司国防行业运营背景及政府合同审计要求(DCAA),财务造假风险评估为低,但若无完整财务报表披露,则无法作出全面判断。 [CR009, CR010, CR011, CR024, CR025, CR038]

人员 / 执行风险登记册
风险类别证据影响尽调要求
Brandon Tseng(CEO)与 Ryan Tseng(COO)——联合创始人关键人集中;双领导层高度依赖关键人 / 领导层Gary Steele 接任 CEO 后,Brandon 留任核心运营职位;Ryan 任总裁兼 COO;两位联创均保持运营活跃高——一位或两位联合创始人离开可能损害战略公信力,动摇投资者信心索取正式接班计划及留任股权文件
员工规模从 1,400 人扩张至 3,000 人以上——人数翻倍带来的组织承载风险执行 / 扩张公司宣布 G 轮后将员工从约 1,400 人增至 3,000 人以上;未披露具体时间表中——快速招聘稀释文化,增加管理层级,产品交付质量控制风险上升索取招聘计划、当前离职率及关键岗位招聘进展
安全许可瓶颈——Shield AI 工程岗位的相当比例需要政府安全许可,审批周期长达 6–18 个月人才 / 运营依据国防承包商运营背景推断;Shield AI 未公开披露具体数据中——持证工程师招募延误可能拖慢 X-BAT 和 Hivemind 研发进度索取持证员工占比及平均审批周期
Aechelon 团队整合——收购后的文化与运营融合收购后整合 / 人员收购于 2026 年 3 月宣布;整合团队构成及留任激励方案未披露中——若留任方案不到位,Aechelon 核心工程师可能在收购后离职索取 Aechelon 关键工程师的留任方案结构
Gary Steele 出任 CEO——2025 年从 Splunk 空降;防务领域经验有限领导层 / 过渡Gary Steele 于 2025 年 2 月宣布出任 CEO;首次涉足防务行业,背景为企业软件低至中——Tseng 兄弟作为联创保留核心机构知识;Steele 的企业软件经验与软件化转型方向契合通过尽调访谈和顾问委员会构成评估 CEO 对防务领域的熟悉程度

人员与执行风险代表组织脆弱性,资金充裕度及管理层履历仅能部分缓解,无法全面消除。尽调问询重点集中于继任计划及安全许可人才招募渠道。

[CR016, CR017, CR018, CR035, CR036]

7.6 缓释措施与论断颠覆标准

以下是关键风险缓释措施及其监测指标。监管/法律风险方面,主要缓释措施为通过 DDTC 注册、商品管辖权认定及许可出口协议构建积极的 ITAR 合规体系;监测指标为 DDTC 或 BIS 对 Shield AI 发起任何执法行动,或任何盟友客户报告技术转让被拒。论断颠覆触发条件:国务院发起 ITAR 除名行动,或国会对自主武器出口施加限制。X-BAT 运营风险方面,缓释措施为以里程碑为节点分阶段释放资金的试飞计划;监测指标为 2026 年秋季首飞里程碑能否如期达成。论断颠覆触发条件:首飞推迟至 2027 年中期以后,且费用超支超过估算开发预算的 50%。财务/模型风险方面,缓释措施为 Series G 20 亿美元融资提供的资金跑道;监测指标为 2026 年营收增速是否达到约 80% 指引(对应约 4.8 亿美元)。论断颠覆触发条件:2026 年实际营收低于 3.5 亿美元,意味着软件转型叙事下的执行持续落空。合作伙伴依赖方面,缓释措施为 GE Aerospace 合作协议及 IDIQ 合同结构;监测指标为 GE Aerospace 任何供应中断公告,或 DoD 预算持续决议超过六个月。关键人物风险方面,缓释措施为正式接班计划及领导层保留激励权益;触发条件为两位联合创始人在 12 个月内相继离职。投资方应要求管理层披露:ITAR 合规计划文件、Hivemind 安全认证路线图、X-BAT 项目预算及里程碑计划表、2025 年实际营收与指引的对账报告,以及 Aechelon 整合时间表与成本。 [CR001, CR002, CR007, CR010, CR011, CR016]

风险缓释与终止标准一览
风险领域缓释措施监测指标逻辑破坏触发点
监管 / ITAR维持活跃的 DDTC ITAR 合规项目;为盟友客户建立出口许可证组合DDTC 或 BIS 执法行动;盟友客户转让受阻国务院 ITAR 禁止令,或盟友国家 V-BAT / Hivemind 出口许可申请被拒
DoDD 3000.09 / 自主武器政策Hivemind 采用人在回路设计;内部已记录政策合规架构DoDD 3000.09 新修订;国会自主武器立法美国新政策或条约禁止 Hivemind 在所有进攻场景中执行自主交战
X-BAT 首飞 / 进度分阶段里程碑计划、GE 推进系统合作、G 轮资本缓冲2026 年秋季首飞公告;NAVAIR 适航里程碑报告2027 年第二季度末前未完成首飞,且/或项目预算超支超过计划的 50%
财务 / 营收不达预期G 轮 20 亿美元资本跑道;多元化 IDIQ 合同管线FY2026 实际营收对比约 80% 增长指引(约 4.8 亿美元);毛利率走势FY2026 营收低于 3.5 亿美元,表明持续未达指引且软件规模化未能兑现
GE 单一来源推进系统正式合作协议;与 GE Aerospace 战略绑定GE 交付进度遵守情况;任何 GE 供应链中断公告GE Aerospace 确认无法按 X-BAT 进度交付 F110 发动机,且 90 天内未找到替代推进方案
关键人 / 联合创始人离职留任股权;正式接班规划Brandon Tseng 与 Ryan Tseng 就职状态;董事会构成变动两位联合创始人在 12 个月内相继离职,且未提前公布接班方案
营收集中 / USCG IDIQ扩大国防部项目组合;推进盟友对外军售;向主承包商授权 Hivemind新 IDIQ 合同授予;盟友对外军售宣布;Hivemind 授权协议公告USCG IDIQ 任务订单执行暂停或重组,且无同等规模替代合同

终止标准与逻辑破坏触发点均为具体可观测事件,一旦确认将从根本上动摇投资逻辑。每个触发点都关联可量化的监测指标,可按季度追踪,无需内部信息。

[CR001, CR002, CR007, CR010, CR012, CR016]
Chapter 08

08估值

8.1 投资逻辑与反向论点

Shield AI 的投资逻辑建立在五大支柱之上。其一,美国及盟友国防市场对自主 AI 的需求正处于结构性转变的初期,国防 AI 预算预计从 2024 年约 90 亿美元增至 2032 年逾 400 亿美元。Shield AI 占据平台层——Hivemind Enterprise 定位为军事自主系统的开发者操作系统,而非单点解决方案,一旦大规模部署至各类飞行器和无人机机队,将形成持久的切换成本。其二,2026 年 2 月美国空军在协同作战飞机(CCA)项目中选择 Shield AI 担任任务自主供应商,将 Hivemind 在最高采购层级获得验证,并为超过 200 架飞机的潜在项目锁定了长期收入基础。其三,Gary Steele 于 2025 年 5 月出任 CEO,大幅提升了管理层厚度;其过往履历(Proofpoint IPO、Splunk 以 280 亿美元被收购)与推动 Shield AI 走向 IPO 或战略并购的路径高度契合。其四,2026 年 3 月收购 Aechelon Technology 拓展了 Hivemind 在训练与仿真方面的优势,有望加速软件收入占比的提升。其五,公司累计融资 35.2 亿美元,在不触发近期稀释性融资压力的前提下为持续研发投入提供了充足空间。 反向逻辑同样有力。Shield AI 尚未实现 GAAP 盈利,研发和制造端烧钱持续;自由现金流转正依赖于一场迄今尚未在规模层面得到验证的软件转型。ainvest.com 的分析明确指出,127 亿美元估值依赖一场从硬件主导的 V-BAT 收入转向软件优先的高风险变革,分部毛利率至今未公开披露。公司在客户结构上高度依赖美国 DoD 及盟友军方,一旦预算重组或项目取消,将面临单一品类集中风险。管理 AI 自主软件出口的法规正在演变,可能制约国际关键市场的销售空间。 [CV011, CV012, CV013, CV014, CV015, CV022]

投资正反逻辑
类别正向逻辑(乐观)反向逻辑(悲观)
市场国防 AI 预算预计从约 90 亿美元(2024 年)增至 400 亿美元以上(2032 年);结构性顺风利好自主平台若近期在同等级冲突中 AI 自主系统的致命性指标未获验证,预算优先级可能从 AI 自主领域转移
产品 / 平台Hivemind Enterprise 是覆盖 15 个以上部署系统的开发者平台,切换成本高;Aechelon 收购带来网络效应与仿真护城河V-BAT 硬件收入仍占主导;软件授权模式尚未规模化验证;毛利率未披露
客户 / 营收美国空军 CCA 入选和海岸警卫队 V-BAT 合同构成耐久的长期营收支柱,有望多年续签国防部集中度高,单一客户风险突出;项目取消或缩减可能撤走主要营收支柱
财务FY2024 营收同比增长 64% 至 2.67 亿美元;管理层指引 FY2026 达 5.4 亿美元以上;G 轮注入 20 亿美元新资本公司尚未实现 GAAP 盈利;研发与制造烧钱速度高;盈利路径依赖尚未验证的毛利率改善
竞争格局CCA 项目击败 Anduril 和 General Atomics,印证 Hivemind 技术优势;Gary Steele 出任 CEO 释放 IPO 备战信号Anduril 资本实力更强(目标估值约 600 亿美元),自主软件差距正在收窄;出口管制阻力压缩国际可触达市场

正向逻辑基于可公开观测的证据;反向逻辑引用反向分析师观点,主要来自 ainvest.com 和 acquinox.capital。未解决事项(如毛利率)已在证据缺口中标注。

[CV011, CV012, CV013, CV014, CV015]
FV001: 投资建议逻辑
[CV011, CV028]

8.2 投资建议与融资背景

尽调结论为有条件买入,适用于后期私募阶段入场,须严格控制仓位,预计 IPO 或战略并购的流动性窗口为 18—36 个月。确信度为中等:Shield AI 的营收轨迹及平台层胜出明显突出,但以 FY2024 营收 2.67 亿美元计算,47 倍前瞻倍数相较所有国防科技上市可比公司已偏高,Palantir 除外。若管理层的 FY2026 5.4 亿美元以上指引落实,隐含前瞻倍数将降至约 23 倍,考虑到经常性软件叙事,该估值更具可辩护性。风险评级为高,主要驱动因素为尚未盈利、软件转型执行依赖性强,以及公开毛利率数据缺失。 Series G 融资于 2026 年 3 月完成,由 Advent International 和 J.P. Morgan 领投,融资 20 亿美元,投后估值 127 亿美元。据报道本轮融资约由 15 亿美元股权和 5 亿美元债务或结构融资构成。相较于 2025 年 3 月完成的 Series F(融资 2.4 亿美元,估值 53 亿美元),此次估值增幅达 140%。通过 Series G 累计融资合计约 35.2 亿美元。估值升幅之快——从 53 亿美元到 127 亿美元仅历时 12 个月——折射出 CCA 项目胜出、CEO 换届以及 2025 年地缘政治事件驱动的国防 AI 市场重新定价。入场纪律至关重要:以 127 亿美元入场的投资者应压力测试该仓位,在风险偏好收缩的公开市场中,倍数压缩 30%—40% 可能将 IPO 定价压至 80—100 亿美元区间。Series G 优先股的股权结构表组成、优先偿还堆叠及反稀释条款是关键尽调事项,目前尚未公开披露。 美联邦支出数据库确认美国政府为 Shield AI 的主要合同方,Palantir 的 10-K 年度报告提供了先例——国防 AI 软件公司在实现盈利并获得政府全范围合同授予后,IPO 后可维持高收入倍数。 [CV001, CV002, CV003, CV004, CV005, CV006]

投资建议摘要
维度评估
投资建议有条件买入——以 127 亿美元投后估值进入上市前后期阶段,严格控制仓位(≤组合的 3%)
确信度中——营收增长轨迹与 CCA 项目胜出已确认;盈利时间表与毛利率走势未披露
风险评级高——尚未盈利;国防部客户集中度高;软件化转型执行风险显著;FY2024 营收倍数高达 47 倍
估值立场溢价有据但偏高——FY2024 营收倍数 47 倍;若指引兑现,FY2026E 前瞻倍数约 23 倍;悲观情景下约 10 倍
目标收益与持有期18–36 个月内实现 1.5–2.5 倍回报,路径为以 140–240 亿美元估值 IPO 或以 150–250 亿美元估值被战略收购;退出触发点:IPO 或并购事件

评估以 2026 年 5 月为时点,基于 G 轮公开信息及分析师观点。确信度与风险评级为定性判断,获取非公开财务数据后可能发生重大变化。

[CV001, CV003, CV004]
FV004: 投资关键指标
[CV001, CV006, CV007]

8.3 乐观/基准/悲观估值情景

三个概率加权情景界定了 Shield AI 投资者在未来 24—36 个月内的收益区间。 乐观情景(概率约 20%):Shield AI FY2026 营收超过 7 亿美元,FY2027 实现 EBITDA 转正,赢得更多 CCA 增量席位并在国际市场扩展 Hivemind Enterprise 授权,以 240—350 亿美元估值完成 IPO 或战略出售。软件收入占比超过总营收的 55%,毛利率接近 60% 以上,市场给予与高增长企业软件相符的 30—45 倍 NTM 收入倍数。ipo.club 的分析将此上行情景 IPO 估值约定在 240 亿美元。 基准情景(概率约 55%):Shield AI FY2026 营收达到 5.4 亿美元以上(与管理层指引吻合),Hivemind Enterprise 软件订单持续实现两位数同比增长,FY2027 年底前软件贡献利润转正。2026—2027 年窗口 IPO 定价在 140—180 亿美元,对应 FY2026E 营收 26—33 倍——高于 Kratos 但明显低于 Palantir 当前倍数。以 127 亿美元入场的投资者在 18—24 个月内可实现 1.1—1.4 倍回报,仓位纪律因此至关重要。 悲观情景(概率约 25%):营收增速低于 30% 同比,软硬件收入结构未能实质改善,DoD 预算限制或 CCA 项目重新划定范围压缩订单可见性,倍数收缩将估值压至 50—80 亿美元——低于 Series G 入场价。这一情景是反向分析师识别的核心风险,也印证了 Shield AI 在毛利率尚未经证实、尚未盈利背景下的现实处境。ainvest.com 的分析专门指出,这一下行风险是当前估值中最被低估的隐含风险。 [CV031, CV032, CV033, CV034, CV035]

乐观 / 基准 / 悲观情景
情景核心假设FY2026E 营收隐含估值营收倍数概率
乐观CCA 快速规模化;Hivemind 国际扩张;软件收入占比超 55%;毛利率超 60%;以溢价 IPO 或被收购$700M+$24–35B34–50x FY2026E~20%
基准管理层指引兑现(5.4 亿美元以上);软件占比提升;2026–2027 年窗口期以 26–33 倍 FY2026E 倍数 IPO 或并购$540M+$14–18B26–33x FY2026E~55%
悲观增长减速(同比 < 30%);软件化转型停滞;估值倍数压缩;CCA 缩减或被竞争对手替代<$400M$5–8B12–20x FY2026E~25%

营收与估值数据均为推演性估算,来源于公开指引、可比倍数及分析师观点。概率为主观判断,非财务预测。前瞻营收倍数假设市场环境与 2026 年第一季度一致。

[CV032, CV033, CV034]
FV002: 估值敏感性
[CV008, CV017]
FV003: 估值 / 回报区间
[CV031, CV032, CV033]

8.4 可比公司估值分析

Shield AI 的可比集需横跨上市国防大型承包商、高增长国防软件公司及私营国防 AI 同类公司——因为没有任何一家上市公司能完整捕捉 Hivemind 将 AI 软件授权与硬件集成结合起来的商业模式。分析共设五个参考锚点。 Anduril Industries 是最接近的私营同类竞争对手:双方在 CCA 和自主系统上直接竞争,Anduril 同样向硬件加软件集成方向延伸,据报 2026 年初正目标以 600 亿美元估值完成新一轮融资。Anduril 此前经 Acquinox 披露的 305 亿美元估值已隐含对收入的大幅溢价。若 Anduril 估值达到 600 亿美元,Shield AI 127 亿美元的估值相对其最直接的战略竞争对手已形成明显折价,尽管两者增长轨迹相近。 Palantir Technologies 是最主要的公开软件可比公司。2026 年初市值约 1,100 亿美元,对应约 38—50 倍过去十二个月营收,印证了国防 AI 软件公司在实现盈利后可维持极高倍数。但 Palantir 已实现盈利且拥有多元化商业收入基础,Shield AI 两者均尚不具备。 Kratos Defense & Security Solutions 是一家以硬件为主的无人机制造商,市值约 40 亿美元,对应 3—4 倍收入,揭示了硬件偏重国防业务所承受的倍数折价——这正是软件转型失败后 Shield AI 的下行情景。 L3Harris Technologies 是一家全规模国防大型承包商,市值约 200 亿美元,对应约 1 倍收入,代表高度商品化国防硬件业务的估值下限。 Axon Enterprise 是一家拥有经常性软件收入的高增长政府技术公司,市值约 300 亿美元,对应约 15 倍收入,为政府市场中的软件属性公司提供了参考可比。 上述可比集仅为示例;完整枚举还应包括 CACI International、Booz Allen Hamilton 及国防 AI 并购中的买入倍数。 [CV016, CV017, CV018, CV019, CV020, CV021]

可比估值对照表
公司阶段FY2024 约收约估值EV / 营收对 Shield AI 的主要参照意义
Anduril Industries私有公司(上市前)~$1.0B(估)~$60B(目标)~60x最直接的自主系统同行;CCA 竞争对手;硬件整合程度更高
Palantir Technologies (PLTR)上市公司(纽交所)~$2.9B~$110B 市值~38–50x TTM防务 AI 软件先例;验证政府 AI 盈利后的高倍数估值
Kratos Defense & Security (KTOS)上市公司(纳斯达克)~$1.1B~$4B 市值~3–4x以硬件为主的无人机可比标的;软件化转型失败时的悲观情景参照
L3Harris Technologies (LHX)上市公司(纽交所)~$21B~$20B 市值~1x大型国防主承包商;商品化硬件营收模式的估值底部基准
Axon Enterprise(AXON,上市可比公司)上市公司(纳斯达克)~$2.0B~$30B 市值~15x高增长政府科技可比标的;验证政府市场经常性软件收入的估值溢价

营收与估值数据均为近似值,来源于 2026 年第一季度的公开文件、分析师估算及新闻报道。私有公司数据(Anduril)来自二级来源估算。EV/Rev 倍数为推演性参考;Palantir 10-K 年度报告作为公开市场防务软件先例。标注"~"的单元格为估算值。

[CV016, CV018, CV019, CV020, CV024]

8.5 退出准备、论断颠覆触发条件与最终尽调清单

Shield AI 最可能的退出路径是在 2026—2027 年窗口 IPO。Gary Steele 于 2025 年 5 月出任 CEO,被普遍解读为 IPO 准备动作;他有 Proofpoint IPO 经验,并主导了 Splunk 被收购的全程。公司 127 亿美元私募估值设定了较高的 IPO 定价门槛,需要持续的营收动能和利润率改善才能撑住。若 2026 年以 140—180 亿美元估值上市,S-1 须展现清晰的营收加速、软件毛利率超过 50%,以及 FY2027—28 实现 GAAP 盈利的可信路径。ipo.club 分析师预测,乐观情景下 IPO 估值可达约 240 亿美元。 战略并购是第二条可行退出路径。Boeing、L3Harris、Northrop Grumman 和 RTX(Raytheon Technologies)是合理的潜在买家,均会将 Hivemind 视为自身自主系统业务的软件平台层。主要承包商以 150—250 亿美元完成收购,意味着较 Series G 入场估值实现 1.2—2.0 倍回报,对后期投资者而言属于合理收益区间。鉴于当前市场环境,SPAC 或直接上市路径概率较低。 论断颠覆监控应聚焦五个触发条件:(1)CCA 项目落败于 Anduril 或 General Atomics 等竞争对手;(2)FY2026 同比营收增速低于 30%,表明增长减速;(3)FY2026 全年毛利率仍低于 40%,确认软件转型失败;(4)CEO Gary Steele 在 12 个月内离职;(5)Hivemind 国际部署遭受出口管制执法行动。 任何投资承诺前的最终尽调须包括:(1)分部毛利率明细(软件授权、V-BAT 硬件与政府服务的拆分);(2)股权结构表及优先股堆叠分析,明确 Series G 清算优先权及反稀释条款;(3)软件年经常性收入(ARR)及净收入留存率指标;(4)按客户与项目拆分的管道与订单构成;(5)Hivemind 国际部署的 ITAR/EAR 出口管制合规状况。 [CV029, CV034, CV035, CV036, CV037, CV038]

逻辑破坏与终止触发点
触发点可观测阈值建议行动严重程度
CCA 项目失标或缩减Shield AI 被移出 CCA 任务自主提供商,或项目平台数量缩减超 50%退出或全面重新评估;CCA 是核心长期营收支柱严重
营收增速下滑FY2026 营收低于 4.5 亿美元(低于指引超 17%),或同比增速跌破 25%减仓 50%;重新评估 IPO 时间表与基准估值
软件毛利率停滞任何 S-1 招股说明书或投资者披露确认 FY2026 软件部门毛利率低于 40%退出:硬软件组合不支持溢价估值;将标的重新归入硬件同行可比严重
CEO 离职Gary Steele 在任职 18 个月内离职(2027 年 5 月前)暂停新增投资;重新评估战略方向与 IPO 时间表
出口管制执法针对 Hivemind 国际部署的 ITAR / EAR 执法行动或禁令重新评估总可用市场(TAM);乐观和基准情景均预设了国际收入

触发点定义为可观测事件或指标,一旦确认,需启动全面重新评估或退出持仓。严重程度:严重 = 立即考虑退出;高 = 减仓;中 = 加强监测。

[CV036, CV037]
最终尽调要求清单
尽调事项理由主要信息来源
各营收板块毛利率(软件 vs. 硬件 vs. 服务)投资逻辑依赖软件化转型;未披露毛利率是承保不确定性中最大的一块经审计财务报告或管理层数据室
股权结构表与优先股堆叠分析G 轮融资含股权与债务;优先股的清算优先权和反稀释条款影响普通股价值公司提供的股权结构表、股份购买协议(SPA)及公司章程
软件年经常性收入(ARR)与净收入留存率(NRR)ARR 和 NRR 是软件平台估值的核心指标;缺失这两项数据,无法与 SaaS 同行进行比较内部 BI 或投资者报告包
按客户和项目划分的管线与积压订单国防部客户集中度风险要求细化至项目层级的积压分析,以压力测试悲观情景业务拓展管线报告或 CRM 数据提取
ITAR / EAR 出口管制合规状态Hivemind 已部署于乌克兰及其他冲突地区;出口管制违规可能导致执法行动或国际收入受限法律顾问备忘录及监管文件(SAM.gov、DDTC 记录)

以下为最终确认投资承诺前所需的最低限度非公开数据。多数条目为上市前后期尽调的标准要求,但鉴于毛利率与股权结构均未披露,其重要性尤为突出。

[CV038, CV039]

免责声明

本报告为基于公开证据的尽职调查快照,不构成投资建议。重要的财务、法律、技术及合同事实仍属非公开信息,在作出任何投资决策前,应直接向管理层及原始文件进行核实。

证据索引

结论
编号陈述可信度来源
CO001 Shield AI was founded in 2015 by Brandon Tseng, Ryan Tseng, and Andrew Reiter in San Diego, California. SO001, SO004
CO002 Brandon Tseng is a co-founder of Shield AI, a former US Navy SEAL with combat deployments in Afghanistan and the Pacific, and currently serves as President and Chief Growth Officer. SO001, SO004, SO005
CO003 Ryan Tseng co-founded Shield AI and served as CEO from 2015 until May 2025, growing the company to over 900 employees and raising more than $1B in funding before transitioning to President and Chief Strategy Officer. SO004, SO005, SO008
CO004 Andrew Reiter is the third co-founder of Shield AI and served as an AI technologist building the foundational algorithms for Hivemind. SO001, SO002
CO005 Shield AI is headquartered in San Diego, California. SO001, SO004
CO006 Shield AI has offices in San Diego (HQ), Dallas TX, Washington D.C., Abu Dhabi UAE, Kyiv Ukraine, and Melbourne Australia. SO004, SO020
CO007 Gary Steele was appointed Shield AI's CEO effective May 13, 2025, succeeding co-founder Ryan Tseng, and also joined Shield AI's Board of Directors. SO004, SO005, SO009
CO008 Gary Steele was previously CEO of Splunk, which he led to 58% revenue growth before Cisco acquired it for $28 billion in 2024. SO005, SO009, SO020
CO009 Gary Steele founded Proofpoint, served as its CEO from founding through IPO, and grew it to an enterprise value of $12 billion. SO004, SO020
CO010 Shield AI raised $240 million in its Series F-1 round announced March 6, 2025, at a post-money valuation of $5.3 billion. SO003, SO004, SO012
CO011 The Series F-1 round was led by strategic investors L3Harris Technologies and Hanwha Aerospace, with participation from Andreessen Horowitz, Booz Allen Hamilton, and Washington Harbour Partners. SO003, SO019, SO022
CO012 Andreessen Horowitz participated in Shield AI's Series F-1 round as an existing investor. SO003, SO019
CO013 Shield AI raised $2 billion in its Series G round on March 26, 2026, at a post-money valuation of $12.7 billion—a 140% increase from the $5.3B valuation set one year earlier. SO006, SO007, SO021
CO014 The Series G was co-led by Advent International and JPMorganChase's Security and Resiliency Initiative. SO006, SO007
CO015 Blackstone provided $500 million in fixed-return preferred equity as part of Shield AI's Series G financing package. SO006, SO021
CO016 Shield AI secured a $250 million credit facility as part of the Series G financing, in addition to equity raised. SO006
CO017 Shield AI has raised approximately $3.52 billion in total funding across 17 rounds since its founding in 2015. SO007, SO006
CO018 Shield AI's first external funding was a $600K angel round in 2015. SO007
CO019 Shield AI's flagship product is Hivemind Enterprise, an AI autonomy platform composed of four components: EdgeOS (runtime), Pilot (autonomy catalog), Commander (C2 toolkit), and Forge (autonomy factory). SO004, SO011, SO013
CO020 Hivemind operates in GPS-denied and communications-denied environments, enabling autonomous mission execution without reliance on remote human control. SO001, SO013, SO016
CO021 Shield AI acquired Martin UAV, the manufacturer of V-BAT, in July 2021 to integrate Hivemind autonomy into the V-BAT VTOL platform. SO016, SO018
CO022 The V-BAT is a VTOL drone with up to 11 hours of endurance, 25-pound payload capacity, and a takeoff/landing footprint as small as 4.6 m × 4.6 m. SO014, SO016
CO023 Shield AI acquired Heron Systems in July 2021; Heron Systems won the DARPA AlphaDogfight Trials by defeating all competing AI teams and human F-16 pilots. SO016, SO018
CO024 Hivemind flew the X-62A VISTA (a modified F-16) autonomously in a DARPA Air Combat Evolution program test that included a simulated dogfight against a human F-16 pilot in 2023. SO008, SO011, SO005
CO025 Shield AI acquired Sentient Vision Systems, a Melbourne-based provider of wide-area motion imaging (ViDAR) software, in April 2024. SO007, SO004
CO026 Shield AI acquired Aechelon Technology, a maker of high-fidelity military flight simulation software, in March 2026 as part of the Series G financing. SO006, SO007, SO011
CO027 Shield AI has approximately 1,422 employees as of March 2026, up from about 900 at the time of the May 2025 CEO transition. SO007, SO004
CO028 V-BAT drones have been operationally deployed in Ukraine since at least 2024, providing long-range reconnaissance and targeting in GPS- and communications-jammed environments. SO008, SO010
CO029 V-BAT was competitively selected for the US Coast Guard Maritime UAS Services program. SO014
CO030 V-BAT has been operated by all seven US Marine Expeditionary Units (MEUs) and deployed on nearly every class of US Navy ship. SO014, SO017
CO031 X-BAT, the world's first AI-piloted VTOL strike fighter jet, was unveiled by Shield AI in October 2025; it weighs approximately 23,000 pounds and is designed for collaborative combat aircraft (CCA) roles. SO008, SO013
CO032 Hivemind has been deployed on more than 15 distinct platforms, including the MQ-20 Avenger, X-62A VISTA, BQM-177A target drone, and Kratos MQM-178 Firejet. SO011, SO013
CO033 Ryan Tseng previously founded WiPower, a wireless power startup, before co-founding Shield AI, and served as WiPower's CEO and CTO. SO009
CO034 Shield AI has a global office presence in six locations: San Diego CA, Dallas TX, Washington DC, Abu Dhabi UAE, Kyiv Ukraine, and Melbourne Australia. SO004, SO022
CO035 Shield AI's stated mission is 'to protect service members and civilians with intelligent, autonomous systems.' SO001, SO004
CO036 Booz Allen Hamilton partnered with Shield AI in March 2025 to deliver AI-enabled autonomous solutions for the DoD, and made its largest-ever corporate venture capital investment in Shield AI's F-1 round. SO022, SO020
CO037 Nathan Michael serves as Shield AI's Chief Technology Officer. SO023, SO002
CO038 Kingsley Afemikhe serves as Shield AI's Chief Financial Officer. SO023
CO039 Shield AI's Nova quadcopter was the first drone with an onboard AI pilot deployed in combat with US Special Operations Command in 2018, used for building-clearance operations to find IEDs and barricaded shooters. SO001, SO008
CO040 Shield AI reached unicorn status (valuation over $1 billion) in 2021, five years after founding. SO007
CO041 Human Rights Watch warned in March 2026 that the Pentagon's direction to accelerate fully autonomous weapons—after rejecting Anthropic's ethical red lines—signals inadequate safeguards on autonomous weapons development. SO015
CO042 International humanitarian law scholars and advocates argue that fully autonomous weapons systems like those Shield AI develops may violate the principle of distinction between combatants and civilians required under the laws of armed conflict. SO015
CO043 Shield AI's exact revenue, ARR, gross margin, burn rate, and profitability are not publicly disclosed as the company is private and has filed no public securities offerings. SO006, SO007
CO044 The V-BAT's ViDAR sensor system—derived from the Sentient Vision Systems acquisition—provides a coverage rate of 3,140 NM² per hour, more than 2.5 times the nearest competitor. SO014
CO045 Shield AI's Series G valuation of $12.7B was driven primarily by Hivemind's selection as the autonomy software provider for the US Air Force Collaborative Combat Aircraft drone prototype program. SO006, SO021
CO046 Shield AI's Hivemind was selected to power Anduril's Fury autonomous fighter jet as the AI autonomy layer for the US Air Force CCA program, demonstrating cross-company software interoperability in defense. SO006
CM001 The global autonomous defense platforms market was estimated at $69.8 billion in 2026, growing at a 14.0% CAGR to $198.9 billion by 2034, with North America holding a 37.5% share. SM015
CM002 The global military UAS/drone market was estimated at $19.25 billion in 2026, growing to $29.57 billion by 2030 at a CAGR of 11.3% (Business Research Company), and at $20.7 billion in 2026, growing to $66.5 billion by 2035 at a CAGR of 13.8% (Global Market Insights). SM002, SM008
CM003 Analyst estimates for the global military UAS market in 2026 range from $18.55 billion (Mordor Intelligence, CAGR 7.03%) to $27.2 billion (Coherent Market Insights, CAGR 11.5%), and to $98.24 billion by 2033 at an 8.9% CAGR in one broader estimate—a 7× spread driven primarily by definitional scope differences. SM009, SM012, SM007
CM004 The global military AI software market was valued at $9.28 billion in 2026 with a CAGR of 13.9% to $29.93 billion by 2035, according to Global Growth Insights. Stratistics MRC puts military AI systems at $8.5 billion in 2026 with a 15.1% CAGR to $26.3 billion by 2034. SM019, SM020
CM005 The DoD's FY2026 budget requests $13.4 billion for autonomy and autonomous systems—the first year autonomy was designated as its own dedicated budget line—broken down as $9.4B for aerial UAS, $1.7B for maritime autonomy, $734M for underwater, $210M for ground vehicles, and $1.2B for enabling autonomy software. SM001, SM010, SM023
CM006 The US Navy's FY2026 unmanned systems budget request is $5.3 billion—a 70% increase over FY2025—including procurement of MQ-25 carrier-borne drone tankers and additional maritime drone assets. SM001, SM006
CM007 The US Army's FY2026 budget request includes $803.9 million for small UAS programs ($747.9M procurement + $56M R&DTE), with a stated goal to embed drone capabilities across formations and ensure every squad has a UAS by end of FY2026. SM006, SM007
CM008 North America dominates the global military drone market with an estimated 41.3% share in 2026 (Coherent Market Insights), or 33.9% growth share during 2025–2030 (Technavio), making the US DoD the single most important buyer globally. SM009, SM016
CM009 Shield AI was selected in February 2026 as the mission autonomy provider for the US Air Force Collaborative Combat Aircraft (CCA) program, with Hivemind software integrated on Anduril's YFQ-44A Fury aircraft following a competitive TMRR evaluation. SM003, SM004
CM010 On February 26, 2026, Anduril's YFQ-44A CCA flew successfully with Hivemind activated, completing all required test points over the Mojave Desert and then switching to Anduril's Lattice AI mid-flight—validating the A-GRA platform-agnostic architecture. SM005, SM003
CM011 The Air Force CCA program uses the Autonomy Government Reference Architecture (A-GRA), an open standard that makes autonomy software swappable across CCA platforms—providing a structural advantage to platform-agnostic software vendors like Shield AI over platform-specific software developers. SM003, SM004, SM005
CM012 Ukraine procured approximately 1.5 million FPV drones in 2024 and planned 4.5 million for 2025 with a ~$2.6 billion budget, demonstrating attritable drone warfare at scale and directly informing the US Army's Transformation in Contact doctrine. SM007, SM006
CM013 The Defense Innovation Unit's FY2026 budget was increased to $2 billion, up from $1.3 billion the prior year, expanding its capacity to fast-track OTA contracts with non-traditional defense vendors like Shield AI. SM023, SM014
CM014 Air Force selected a second batch of 9 vendors for CCA Round 2 (2026), representing a 'broad spectrum' of concepts to be narrowed for prototyping—creating additional integration licensing opportunities for autonomy software providers like Shield AI. SM022
CM015 Shield AI's Hivemind has demonstrated A-GRA-compliant autonomy across platforms including the MQ-20 Avenger (GA-ASI), Northrop Grumman Talon IQ, BQM-177 test aircraft, and the UH-72A Lakota helicopter, establishing platform diversity as a competitive moat. SM003, SM004
CM016 The Trump administration's FY2026 budget process has canceled several drone programs that were near development completion, with the Inside Unmanned Systems report noting a 'move-fast' procurement ethos that prizes off-the-shelf capability over long-running development programs. SM006
CM017 DoD Directive 3000.09 (reaffirmed under 360iResearch) requires that autonomous and semi-autonomous weapon systems be used with 'appropriate care and in accordance with the law of war,' setting a legal-regulatory floor that limits fully autonomous lethal decision-making without human oversight. SM021
CM018 Global military spending reached approximately $2.7 trillion in 2024 (SIPRI), with Asia-Pacific posting military expenditure growth for the 35th consecutive year, Japan posting a 21% YoY increase, and the Middle East recording a 15% increase in 2024. SM021, SM007
CM019 The autonomous military systems market is expected to reach $40.86 billion in 2026 (CAGR 11.97% to $80.76B by 2032), with the 360iResearch report noting that DoD Directive 3000.09 and DARPA's focus on trustworthy AI are creating a compliance layer that advantages vendors with validated, explainable autonomy. SM021
CM020 The autonomous defense platforms market's mission-level autonomy segment is expected to grow at a 21.6% CAGR, the fastest sub-segment, over the 2026–2034 forecast period (Fortune Business Insights)—directly aligned with Hivemind's core product positioning as mission-level AI for unmanned platforms. SM015
CM021 ITAR/EAR export control regulations limit Shield AI's ability to sell advanced autonomous weapons software and tactical UAS to allied militaries without State Department licenses—slowing the pace of international expansion relative to domestic program growth. SM021, SM006
CM022 Shield AI's estimated 2026 revenue of ~$540M is approximately 10–15% of its estimated SAM of $3–6B and less than 6% of the US DoD's dedicated autonomy budget of $13.4B in FY2026, indicating substantial remaining market penetration opportunity. SM001, SM019
CM023 Hanwha Systems' strategic investment in Shield AI's Series F provides a channel into the Korean defense acquisition program (DAPA), and Hanwha is one of the world's top 20 defense contractors—expanding Shield AI's ally-market access beyond FMS channels. SM021, SM003
CM024 Fixed-wing drone platforms represent approximately 51% of the military drone market in 2026 by revenue, while VTOL/hybrid platforms hold the fastest CAGR of approximately 14.8% (GMInsights), directly relevant to Shield AI's V-BAT and X-BAT hybrid VTOL positioning. SM002, SM015
CM025 The DoD FY2026 budget request matches the prior year's $848 billion total and may be supplemented by $113.3 billion in requested reconciliation funding—but that reconciliation funding is not guaranteed, creating material budget uncertainty for autonomy program continuity. SM006, SM011
CM026 L3Harris introduced AMORPHOUS in February 2025 as an open-architecture platform intended to operate thousands of autonomous assets, representing a direct competitive threat to Hivemind's position as the leading platform-agnostic military autonomy stack. SM021
CM027 The Air Force's CCA Round 2 expanded to 9 vendors (2026) for drone wingman prototyping—suggesting that while Shield AI has a first-mover autonomy software advantage, the CCA program is deliberately diversified to prevent single-vendor dependency. SM022
CM028 The INSS National Defense University analysis of the FY2026 RDTE budget identifies autonomous and AI systems as the top strategic innovation priority, with a specific call for government-industry partnerships to accelerate disruptive technology deployment. SM013
CM029 Over 47% of global defense procurement contracts in 2024 included AI components for combat readiness enhancement, per Global Growth Insights—indicating that AI integration is becoming a baseline requirement rather than a differentiator in defense procurement. SM019
CM030 The Technavio military drone market analysis projects $12 billion in incremental growth from 2025 to 2030 at a 13.6% CAGR, with North America accounting for 33.9% of growth—consistent with Technavio's separate research on defense robotics. SM016
CM031 The market for autonomous military systems is being shaped by a 'software-led disruptors' layer: L3Harris (AMORPHOUS), Shield AI (Hivemind), and Palantir (AI Platform) are positioning against traditional defense primes, with the competitive boundary shifting from airframe ownership to software interoperability and sustainment. SM021
CM032 The HigherGov SUAS budget tracker shows the Army SUAS program as one of the fastest-growing budget lines in DoD, with Shield AI's V-BAT (Group 4) positioned as a higher-payload complement to smaller Group 1–2 SUAS platforms being fielded through Transformation in Contact. SM018, SM006
CM033 The MarketsandMarkets autonomous navigation market was valued at $6.0 billion in 2023, projected to grow to $12.9 billion by 2028 at a 16.4% CAGR—an earlier and narrower estimate for the same segment that contextualizes the market's growth trajectory over the prior decade. SM024
CM034 CDO Magazine reports the Pentagon FY2026 autonomy line specifically calls out 'enabling capabilities—the autonomy software, the things that underlie all these systems, working and operating together as a central brain' as $1.2 billion of the $13.4B total. SM010, SM001
CM035 The Air Force has accelerated the CCA initiative by using A-GRA across vendor platforms (confirmed February 2026), creating a modular, open procurement architecture that favors software-native autonomy providers with demonstrated multi-platform integration. SM015, SM003
CM036 Shield AI's V-BAT has been exported to Ukraine, India, the Netherlands, and Armenia, confirming foreign military sales traction—though contract values and units shipped are not publicly disclosed. SM003, SM006
CM037 Breaking Defense reports that the Air Force tapped four companies (Beehive, Honeywell, Pratt & Whitney, GE/Kratos) to mature engine designs for future drone wingmen—indicating that the CCA program is expanding its industrial base and creating more integration touchpoints for Hivemind software. SM022
CM038 Shield AI's estimated 2026 revenue of ~$540M against a $69.8B autonomous defense platform TAM implies a sub-1% TAM penetration—however, using the narrower $1.2B enabling-software SAM, Shield AI's estimated revenue represents ~45% of the software-specific SAM, indicating high concentration in its core niche. SM001, SM015
CM039 The INSS NDU analysis notes that the DoD FY2026 RDTE budget emphasizes 'leveraging disruptive technologies' and cross-domain integration as a top priority—directly validating the government's intent to accelerate spend on the autonomy software layer where Hivemind Enterprise operates. SM013
CM040 The Autonomous Defense Platforms market growth in subsea segment (CAGR 18.6%) and Joint/Integrated commands segment (CAGR 20.4%) indicates that the broader autonomous defense addressable market will continue growing faster than overall defense spending, providing headroom for Shield AI's long-term expansion. SM015
CP001 Anduril Industries was valued at $30.5 billion in its June 2025 Series G funding round and is targeting approximately $60 billion in a Series H round as of March 2026, with Thrive Capital and Andreessen Horowitz as lead investors. SP003, SP009, SP013, SP017
CP002 Anduril's 2025 revenue reached approximately $2.1 billion (4× Shield AI's estimated $540M), with 2026 revenue projected at approximately $4.3 billion, driven by the $20B Army contract and Arsenal-1 factory scaling. SP003, SP004
CP003 Anduril has raised over $6.3 billion in total funding across Series A through G, with an expected Series H of $4–8 billion that would bring total funding above $10 billion. SP003, SP009, SP017
CP004 In March 2026, the US Army awarded Anduril a 10-year contract with a $20 billion ceiling covering approximately 120 autonomous systems procurement programs, the largest single contract in Anduril's history. SP003, SP013
CP005 Anduril's Lattice AI is a sensor fusion and command-and-control platform that integrates data from distributed sensors into a common operating picture and automates threat response across multi-domain environments, in contrast to Hivemind's per-platform edge autonomy focus. SP001, SP003, SP012
CP006 Anduril builds a vertically integrated hardware-plus-software stack including Ghost 4 drones, Fury (YFQ-44A) combat aircraft, Roadrunner interceptors, Anvil counter-UAS, and integrated IVAS sensors, giving it a broader product portfolio than Shield AI's software-first model. SP003, SP021
CP007 Shield AI and Anduril are the two US AI-first defense companies with commercially available autonomy stacks and significant DoD deployment track records; they are the closest direct peers in the defense AI sector. SP003, SP012, SP024
CP008 In March 2026, Shield AI's Hivemind and Anduril's Lattice were both validated on the same YFQ-44A Fury airframe in a single flight test, demonstrating A-GRA open-architecture modularity and the co-opetition relationship between the two companies. SP001, SP002, SP010
CP009 A-GRA (Autonomy Government Reference Architecture) is the US Air Force's open-architecture standard for CCA autonomy stacks; it enables autonomy software providers to certify on any compliant platform, and Shield AI is the first to achieve mission autonomy certification under A-GRA for the CCA program. SP001, SP002, SP015
CP010 Kratos Defense's XQ-58A Valkyrie is a low-cost attritable autonomous aircraft originally developed under AFRL's LCAAT program, designed to fly as an autonomous loyal wingman for manned fighters at an estimated unit cost of $2–3 million. SP005, SP006, SP007
CP011 In January 2026, the US Marine Corps awarded a $231.5 million OTA contract to a Northrop Grumman-led team (with Kratos providing the XQ-58A Valkyrie airframe) for the MUX TACAIR program, the Marines' first loyal wingman/CCA program. SP005, SP006, SP007, SP025
CP012 Northrop Grumman is integrating its own Prism autonomy software stack onto the Valkyrie for the MUX TACAIR program, creating a third competing autonomy stack (alongside Hivemind and Lattice) for the DoD CCA autonomy layer market. SP006, SP007, SP020
CP013 General Atomics Aeronautical Systems competes in the Group 5 (large, high-altitude) UAS market with the Avenger jet UAS and MQ-9B SkyGuardian; GA-ASI is estimated to generate approximately $3+ billion in annual revenue from DoD UAS programs. SP016, SP022
CP014 General Atomics competed against Anduril for the US Air Force CCA Increment 1 program and was not selected, establishing Anduril's YFQ-44A Fury as the primary Air Force CCA platform and limiting GA-ASI's CCA market access in the near term. SP004, SP013
CP015 In September 2025, Boeing Defense partnered with Palantir to integrate Palantir's AI and data analytics capabilities across Boeing's classified defense programs, positioning Palantir as Boeing's AI layer rather than building internal AI capability. SP008, SP014
CP016 Palantir's Gotham platform is the dominant AI analytics and battlefield intelligence tool for the US intelligence community and DoD targeting workflows, with an estimated US government revenue of approximately $1.1 billion in 2025. SP018, SP019, SP024
CP017 As of late 2025, Palantir's near-monopoly in Pentagon AI analytics is eroding as Shield AI, Anduril, Applied Intuition, and other specialists displace Palantir in specific mission segments including platform-level autonomy and ISR data processing. SP024, SP012
CP018 Shield AI's Hivemind is deployed on more than 15 distinct aircraft and drone platforms as of early 2026, including V-BAT, X-BAT, Anduril's YFQ-44A, and multiple classified platforms, making it the most broadly deployed platform-agnostic autonomy stack in the US military. SP015, SP010
CP019 Shield AI's CCA mission autonomy contract—Hivemind as the primary AI pilot on the Air Force's Increment 1 CCA platform—is a major competitive differentiator because competitors offering hardware-only solutions (GA-ASI, Kratos) or C2-only AI (Palantir) cannot fulfill this specific mission autonomy function. SP001, SP009, SP015
CP020 Shield AI's GPS-denied autonomy moat is built on over 10 years of reinforcement learning training from real-world combat deployments; this accumulated training data cannot be replicated quickly by any competitor and represents the durable non-architectural competitive asset. SP011, SP012, SP015
CP021 Switching an autonomy stack on a DoD program of record from Hivemind to a competing stack requires re-qualification testing under MIL-STD, FACE, and CMMC standards plus operational integration testing, creating an estimated 18–36 month switching cost barrier. SP011, SP015
CP022 Anduril's Lattice AI and Shield AI's Hivemind serve different layers of the CCA autonomy stack: Lattice manages sensor fusion and battlefield coordination (network layer), while Hivemind handles per-platform mission execution in signal-denied environments (edge layer); these are currently complementary, not winner-take-all competing. SP001, SP005, SP012
CP023 Applied Intuition is an autonomy testing and simulation platform for DoD vehicles; it competes with Shield AI indirectly in the autonomy development toolchain but does not offer deployed edge autonomy software for operational platforms. SP024, SP012
CP024 Skydio serves primarily the commercial and government inspection/mapping drone market and does not offer GPS-denied combat autonomy; it is not a direct competitor to Shield AI's military autonomy segment despite shared UAS technology. SP012
CP025 L3Harris Technologies develops its own autonomous system capabilities for ISR and electronic warfare, including the Autonomous Wing Platform (AWP), and competes with Shield AI at the EW-layer of autonomous systems but not at the pure autonomy software level. SP023
CP026 Northrop Grumman is integrating AI autonomy capabilities into its existing platforms (B-21 Raider, E-2D Advanced Hawkeye, X-47B heritage research) through internal R&D and the Prism software stack; it competes at the platform-integration layer rather than as a pure autonomy software vendor. SP006, SP007
CP027 Legacy defense prime contractors (Northrop, Raytheon, Boeing, L3Harris) achieve 8–10% gross margins under cost-plus contracting structures, versus Anduril's estimated 40–45% gross margins and Shield AI's software-weighted margins that are expected to be meaningfully higher than prime averages. SP003, SP011
CP028 The DoD's A-GRA open architecture mandate prevents any single autonomy stack from locking out competitors at the platform level; all compliant platforms must be certifiable with multiple autonomy stacks, creating both an opportunity (easy entry) and a risk (easy displacement). SP001, SP015, SP011
CP029 Shield AI's Hivemind is sold via a per-platform software licensing model, decoupling revenue from hardware unit economics and creating a recurring royalty stream; this contrasts with traditional cost-plus and firm-fixed-price hardware contracts used by defense primes. SP015, SP011
CP030 Boeing's MQ-25 Stingray is an autonomous carrier-based aerial refueling drone; it competes in a distinct autonomy segment (autonomous tanker, not combat CCA) and is a potential Hivemind licensing customer rather than a direct competitor. SP008, SP014
CP031 General Atomics is estimated to generate approximately $3+ billion in annual revenue, primarily from DoD UAS programs (MQ-9 Reaper fleet sustainment, SkyGuardian FMS sales, Avenger development); it is privately held and does not disclose financials. SP016, SP022
CP032 In 2024, Anduril acquired L3Harris's airborne ISR (WESCAM) business for approximately $800 million, expanding Anduril's hardware manufacturing capability and signaling that the line between AI-first companies and legacy primes is increasingly porous through M&A. SP004, SP013
CP033 Both Anduril (estimated $800M+ annual loss despite high revenue growth) and Palantir (reached GAAP profitability in 2023) demonstrate that defense AI companies are investing heavily in R&D ahead of profitability; Shield AI is not yet profitable and faces similar burn dynamics. SP003, SP024
CP034 Shield AI's combat deployment track record—Hivemind used in SOCOM operations and Ukraine—provides a 'trust through use' credibility moat that competitors building in simulation cannot replicate without matching operational experience. SP011, SP012, SP015
CP035 No commercially available pure-play substitute for GPS-denied edge autonomy at the tactical level exists in the US market as of May 2026; DoD internal build programs (DARPA ACE, AFWERX) have not produced operationally deployable alternatives. SP012, SP015
CP036 Big Tech companies (Microsoft, Google, Amazon) have large-scale AI capabilities and DoD cloud contracts (JWCC), but lack combat deployment credibility, ITAR infrastructure, and mission-specific training data for GPS-denied autonomy, making near-term entry into Shield AI's core segment unlikely. SP019, SP024
CP037 A-GRA certification for a new autonomy stack currently requires 18–36 months of qualification testing; if standards mature and streamline to 6–12 months, Shield AI's first-mover moat could compress significantly as Anduril and Northrop accelerate parallel certification efforts. SP011, SP015
CP038 In the CCA ecosystem, Shield AI's Hivemind owns the mission autonomy (platform AI pilot) layer and Anduril's Lattice owns the sensor fusion (battlefield awareness) layer; current market evidence suggests these layers are complementary rather than competitive in the near term. SP001, SP002, SP012
CP039 Hardware commoditization in the drone market (DJI dominance in commercial, FPV swarm economics in Ukraine) demonstrates that drone platforms commoditize rapidly at scale; Shield AI's software licensing model is structurally insulated from this dynamic because lower hardware costs increase licensing volume rather than compressing Hivemind margins. SP011, SP012
CP040 Shield AI and Anduril operate in a co-opetition dynamic: Anduril is Shield AI's closest AI-first competitor for defense autonomy budget dollars and relationships, yet also its most important hardware partner (Hivemind on Fury) and potential future licensing customer. SP001, SP002, SP010
CI001 Shield AI reported approximately $267 million in total revenue for fiscal year 2024, representing 64% year-over-year growth from $163 million in fiscal year 2023. SI003, SI004, SI008
CI002 Shield AI's revenue for the fiscal year ending March 2025 was approximately $300 million, falling short of an internal guidance target of approximately $400 million by roughly $100 million. SI003, SI009
CI003 Hivemind software revenue accounted for approximately 30% of Shield AI's total revenue for fiscal year 2025, with hardware (V-BAT) and services comprising the remaining 70%. SI003, SI004
CI004 Shield AI's revenue is almost entirely derived from government contracts — U.S. DoD and allied-nation defense ministries — with no disclosed commercial or dual-use revenue from non-government customers. SI001, SI002, SI023
CI005 Approximately 50% or more of Shield AI's fiscal year 2025 revenue was derived from international defense ministries, including Romania, Japan, Greece, Canada, the Netherlands, Ukraine, and Egypt. SI002, SI004
CI006 Hivemind autonomy software is deployed on more than 15 different platform types globally as of April 2026, and was selected as the autonomy stack for Anduril's YFQ-44A Fury CCA in February 2026. SI002, SI020, SI021
CI007 Shield AI management has targeted Hivemind software growing from approximately 30% to 50% of total revenue by fiscal year 2028, representing a deliberate hardware-to-software mix shift to improve blended gross margins. SI003, SI004
CI008 Shield AI's primary go-to-market channels are direct government contracting using Other Transaction Authority (OTA) vehicles, Defense Innovation Unit (DIU) contracts, IDIQ awards, and Foreign Military Sales (FMS) for international customers. SI011, SI014, SI019
CI009 The U.S. Coast Guard awarded Shield AI a $198 million indefinite-delivery/indefinite-quantity (IDIQ) contract in July 2024 to provide V-BAT maritime unmanned aircraft system services under a contractor-owned, contractor-operated model. SI014, SI019, SI011
CI010 V-BAT drones are sold at an estimated unit price of approximately $1 million, making them accessible for smaller defense budgets and allied-nation procurement compared to larger fixed-wing UAS platforms. SI011, SI012, SI022
CI011 The contractor-owned, contractor-operated model for V-BAT deployments creates a recurring services revenue stream tied to platform availability and flight hours, improving revenue predictability relative to one-time hardware sales. SI011, SI013, SI019
CI012 Shield AI's selection as the Hivemind autonomy stack for the Anduril YFQ-44A Fury CCA program is expected to create embedded, recurring per-platform licensing revenue if the CCA program scales to full production. SI002, SI021
CI013 Defense hardware platforms comparable to V-BAT carry gross margins typically in the 20–35% range, while autonomous systems software platforms typically achieve gross margins of 60% or higher at scale; Shield AI has not disclosed margins in either segment. SI009, SI010
CI014 Shield AI's hardware business is capital-intensive: the COCO model requires the company to own deployed platforms, tying working capital to field assets and creating balance sheet exposure not typical for pure-software defense companies. SI011, SI013
CI015 Shield AI spent approximately 22% of revenue on research and development in fiscal year 2024, a high R&D intensity reflecting pre-IPO investment in the Hivemind Foundation Model for Defense, X-BAT development, and the Aechelon integration. SI009, SI004
CI016 With approximately 1,422 employees and $267 million in fiscal year 2024 revenue, Shield AI's revenue per employee is approximately $188,000 — below pure-software defense benchmarks (~$300K+) but consistent with a hardware-software hybrid at its current scale. SI023, SI024, SI003
CI017 R&D investment at Shield AI is partially subsidized by government contract funding through cost-plus-fixed-fee and SBIR/STTR vehicles, meaning reported commercial R&D expense may understate total AI development investment. SI004, SI009
CI018 The Aechelon Technology acquisition — announced March 26, 2026 — adds high-fidelity simulation and synthetic environment capabilities supporting the U.S. Air Force Joint Simulation Environment, at an undisclosed acquisition price. SI017, SI025, SI006
CI019 Shield AI's March 2026 financing event totaled $2 billion, composed of $1.5 billion in Series G equity led by Advent International and co-led by JPMorganChase Strategic Investment Group, plus $500 million in fixed-return preferred equity from Blackstone-managed funds. SI001, SI025, SI005
CI020 The Blackstone preferred equity investment of $500 million functions economically as quasi-debt: Blackstone receives a fixed coupon return and holds structural seniority over common equity, limiting dilution for existing shareholders while providing Shield AI access to non-dilutive growth capital. SI001, SI006, SI008
CI021 Blackstone also committed a $250 million delayed-draw credit facility as part of the March 2026 financing, providing Shield AI with an additional capital reserve available for future acquisitions or operational needs, for total Blackstone exposure of up to $750 million. SI001, SI025, SI006
CI022 Shield AI's post-money valuation following the March 2026 Series G is $12.7 billion, a 140% increase from the $5.3 billion valuation set in March 2025 at the time of the Series F. SI008, SI007, SI025
CI023 Shield AI has raised approximately $3.6 billion in cumulative equity and debt financing across more than 17 rounds since its founding in 2015, making it one of the most heavily capitalized pre-IPO defense AI companies in the United States. SI007, SI016, SI001
CI024 Shield AI missed its fiscal year 2025 internal revenue guidance by approximately $100 million, reporting approximately $300 million against a target of approximately $400 million — a ~25% guidance miss that introduces credibility risk around the $540 million+ 2026 projection. SI003, SI009
CI025 Shield AI management has projected revenue exceeding $540 million for fiscal year 2026, representing approximately 80% year-over-year growth from fiscal year 2025, targeting $1 billion in annual revenue by fiscal year 2028. SI001, SI002, SI004
CI026 Shield AI has not disclosed gross margins, operating income, net income, EBITDA, or cash and equivalents in any public communication as of May 2026, making it impossible to assess profitability trajectory without access to audited financials. SI009, SI010, SI003
CI027 Revenue recognition policies for Shield AI's COCO deployments and multi-year IDIQ contracts are not publicly documented, creating uncertainty about whether reported annual revenue figures reflect actual cash generation or contract milestone recognition timing. SI009, SI010
CI028 Shield AI has not disclosed capital expenditure figures; however, the COCO platform ownership model and Aechelon acquisition create material working capital and investment requirements that distinguish the company's capital intensity from pure software defense companies. SI010, SI009
CI029 Analysts at IPO Club and AI Invest have noted that Shield AI's $12.7 billion valuation requires execution on at least three simultaneous strategic bets: continued 80%+ revenue growth, a successful hardware-to-software revenue mix shift, and integration of the Aechelon acquisition without margin deterioration. SI009, SI010
CI030 At $12.7 billion valuation on projected $540 million 2026 revenue, Shield AI trades at approximately 23–28x forward revenue — a multiple reflecting tech-sector sentiment rather than traditional defense contractor valuation norms (typically 3–5x revenue). SI009, SI004, SI010
CI031 Shield AI's V-BAT is competing for up to $800 million in U.S. Navy intelligence, surveillance, and reconnaissance contracts as of April 2026, representing a significant potential revenue uplift if awarded. SI012
CI032 The primary use of Series G proceeds as disclosed by management is: funding the Aechelon acquisition, scaling V-BAT and X-BAT production, expanding Hivemind software capabilities and the Foundation Model for Defense, and accelerating international market entry. SI001, SI017, SI025
CI033 Shield AI's historical funding progression — from early-stage DIU contracts in 2016 through a $5.3B valuation in March 2025 and $12.7B in March 2026 — reflects compounding investor confidence in the defense autonomy platform market, though the 140% single-year valuation jump is exceptional even by defense-tech standards. SI007, SI016, SI022
CI034 Series G equity investors beyond Advent International and JPMorganChase include Snowpoint Ventures, InnovationX, Riot Ventures, Disruptive, and Apandion, reflecting broad institutional backing across specialist defense-tech and generalist growth funds. SI001, SI025, SI006
CI035 Shield AI's cost structure as of fiscal year 2024 involves significant hardware supply chain and manufacturing costs (for V-BAT production), heavy AI/ML compute costs for Hivemind training and inference, and substantial G&A overhead to support DoD contracting compliance and international operations. SI023, SI004
CI036 Shield AI is pursuing an IPO candidacy at an estimated target valuation of $24 billion or higher, according to analyst projections, contingent on achieving $800M–$1B in annual revenue and demonstrating a credible path to EBITDA-positive operations. SI009
CI037 The company's total headcount grew to approximately 1,422 employees by early 2026 from fewer than 500 in 2022, with active hiring across AI research, embedded systems engineering, defense contracting, and international business development. SI024, SI023
CI038 Revenue concentration risk for Shield AI is substantial: the business is almost exclusively dependent on the U.S. federal defense budget and allied-nation defense procurement, with the U.S. government and its allies together representing over 95% of revenue. SI004, SI009
CI039 Shield AI has generated revenue across at least 13 countries as of fiscal year 2025, a geographic diversification strategy that mitigates single-country concentration risk but introduces foreign military sales regulatory complexity and foreign exchange exposure. SI002, SI004
CI040 The Aechelon acquisition adds a DoD-embedded simulation software business with an existing contract base supporting the Joint Simulation Environment, representing a strategic rather than purely financial acquisition that expands Shield AI's addressable market into the DoD synthetic training market. SI017, SI018, SI025
CE001 Shield AI has deployed Hivemind on over 15 distinct platform types including fixed-wing, rotary-wing, maritime surface vessels, and ground vehicles as of early 2026. SE001, SE021
CE002 The X-BAT, announced October 21, 2025, is described by Shield AI as the world's first AI-piloted VTOL fighter jet with Group-5 UAS classification. SE013, SE024
CE003 ViDAR (Visual Detection and Ranging) is Shield AI's proprietary sensor product for long-range target detection and ranging that does not require GNSS. SE014
CE004 The V-BAT MQ-35A is a Group-3 VTOL UAS with greater than 13 hours of flight endurance and a heavy-fuel engine on the Block Upgrade configuration. SE004, SE011
CE005 Shield AI's product portfolio spans autonomous software (Hivemind Enterprise modules), hardware platforms (V-BAT, X-BAT), and sensor systems (ViDAR). SE001, SE015
CE006 Shield AI completed integration of Hivemind autonomous capabilities on the MQ-20 Avenger drone in less than 12 weeks, demonstrating live autonomy testing with USAF. SE006
CE007 HII and Shield AI completed the first maritime Hivemind integration on the ROMULUS unmanned surface vessel in approximately 6 weeks. SE005
CE008 Hivemind Enterprise comprises four primary modules: Hivemind Pilot (autonomous flight), Hivemind Commander (multi-agent coordination), Hivemind Forge (ML development), and the Foundation Model for Defense. SE016, SE017, SE018
CE009 EdgeOS uses static JSON-based configuration management rather than ROS, enabling deterministic platform-agnostic deployment across heterogeneous vehicle types. SE001, SE021
CE010 EdgeOS employs shared-memory interprocess communication capable of handling millions of messages per second at sub-millisecond latency, supporting real-time autonomous control loops. SE001, SE021
CE011 Hivemind Pilot enables autonomous flight and navigation in GPS-denied environments using simultaneous localization and mapping (SLAM) and visual odometry with EO/IR sensor fusion. SE018, SE001
CE012 Hivemind Commander provides multi-agent swarm orchestration, enabling simultaneous coordination of multiple autonomous platforms during complex missions. SE017
CE013 Hivemind Forge provides an ML model development pipeline where engineers train, evaluate, and package mission AI models for deployment to edge vehicles. SE016
CE014 EdgeOS is platform-agnostic via a plugin SDK that enables hardware abstraction, allowing new platforms to be onboarded by writing hardware interface plugins without modifying mission logic. SE001, SE002
CE015 Shield AI and AWS collaborate to deliver cloud-based EdgeOS development with edge deployment via AWS Elastic Container Registry, enabling fleet-scale autonomy management. SE002
CE016 EdgeOS integrates EO/IR sensor fusion for obstacle avoidance and navigation, allowing autonomous operation without GNSS in denied electromagnetic environments. SE001, SE021
CE017 The X-BAT has a wingspan of 26 feet (extending to 39 feet), a service ceiling exceeding 50,000 feet, and a range of over 2,000 nautical miles. SE013, SE024
CE018 The X-BAT is powered by a GE F110-GE-129 engine equipped with an axisymmetric vectoring exhaust nozzle (AVEN), enabling supersonic VTOL flight capability. SE025, SE024
CE019 X-BAT first flight is targeted for fall 2026 with an approximate production unit cost target of $27 million per aircraft. SE007, SE024
CE020 GE Aerospace and Shield AI formalized a propulsion collaboration agreement for the X-BAT vehicle program in October 2025. SE025
CE021 The V-BAT Block Upgrade announced in 2025 adds Hivemind-powered advanced autonomy, satellite communications, and a heavy-fuel engine as incremental capability packages. SE011
CE022 X-BAT uses VTOL capability for launch and recovery from ship decks or unprepared ground without a runway, enabling carrier and expeditionary operations. SE013, SE007
CE023 Shield AI's acquisition of Aechelon in March 2026 added simulation and synthetic environment capabilities to the product stack, supporting pre-deployment training and reducing live-test costs. SE015
CE024 GPS-denied autonomy via SLAM and EO/IR sensor fusion is Shield AI's core IP differentiator, enabling operations in GPS-jammed environments where GNSS-dependent competitors cannot reliably function. SE001, SE003
CE025 Shield AI's Foundation Model for Defense is a domain-specific AI model trained on defense mission data rather than general internet corpora, providing operational reliability that commercial LLMs lack for contested-environment missions. SE021, SE003
CE026 Hivemind is the prime autonomy stack for the US Air Force's Autonomous Collaborative Platform (A-GRA) program, indicating mission-critical program-level integration. SE022, SE006
CE027 Vertical integration of Hivemind software with proprietary hardware (V-BAT, X-BAT) and sensors (ViDAR) creates a combined platform-level competitive barrier that pure-software autonomy vendors cannot replicate. SE003, SE012
CE028 Third-party defense tech analysts characterize Shield AI as the leading pure-play autonomous AI software company for defense, distinguishing it from large prime contractors that offer autonomy as a feature. SE003
CE029 DoD Directive 3000.09 requires human authorization for lethal decisions by autonomous weapons systems; Hivemind-equipped platforms today operate with a human in or on the loop for engagement decisions. SE003, SE012
CE030 Hivemind is described by Shield AI as a non-jamming, GPS-independent system suitable for contested electromagnetic environments, a key trust attribute for military operators. SE001, SE021
CE031 Independent reporters and analysts note that X-BAT's autonomous operation in contested and congested airspace will require regulatory approval beyond existing UAS frameworks, a process involving FAA, DoD, and potentially foreign regulators. SE009, SE022
CE032 Shield AI does not publicly disclose security vulnerability assessments, CVE disclosures, or penetration testing results for Hivemind or EdgeOS in any reviewed public documentation as of May 2026. SE012
CE033 The MQ-35A US Navy designation for V-BAT signals progress through NAVAIR military qualification, representing formal procurement pathway progress and a trust signal for defense customers. SE004, SE011
CE034 Shield AI's AWS collaboration for EdgeOS development extends FedRAMP-adjacent cloud security practices to the ML development environment, relevant for federal customer data-handling requirements. SE002
CE035 No independent third-party airworthiness certification (equivalent to DAL A/B) for Hivemind software or V-BAT airframe has been publicly confirmed in any sources reviewed as of May 2026. SE012
CE036 Hivemind's role as prime autonomy stack for the USAF A-GRA program demonstrates mission-critical integration at the program level, validating product readiness for high-stakes operational deployment. SE022, SE006
CE037 EdgeOS's plugin SDK architecture is designed to support rapid platform integration, enabling Shield AI to onboard new vehicle types without modifying core mission logic or requiring full re-architecture. SE001, SE014
CE038 X-BAT design prioritizes reduced acoustic and radar signature characteristics suitable for denied-access environments, per company marketing descriptions. SE013, SE008
CE039 Shield AI's product suite supports DoD multi-domain operations concepts by integrating autonomous air, sea, and ground platforms under a single software framework. SE003, SE022
CE040 V-BAT's estimated ~$1M average selling price positions it as an affordable Group-3 persistent ISR asset relative to full manned systems or larger UAS platforms in the same endurance class. SE004, SE012
CU001 The US Coast Guard awarded Shield AI a $198M IDIQ contract for V-BAT production delivery and operations support in July 2024. SU001, SU002, SU003
CU002 Shield AI reports V-BAT has been deployed with allied forces in Japan, Romania, and more than 10 countries total as of 2025. SU007, SU013
CU003 The US Coast Guard is operating V-BAT in full production (not a pilot program) for maritime patrol missions. SU001, SU003
CU004 The US Navy has awarded V-BAT the formal MQ-35A military type designation, indicating NAVAIR qualification progress. SU006, SU013
CU005 The US Air Force has deployed Hivemind as the prime autonomy stack for the A-GRA program and conducted live MQ-20 Avenger test flights. SU010, SU013
CU006 Huntington Ingalls Industries completed integration of Hivemind on the ROMULUS unmanned surface vessel for maritime autonomy. SU009
CU007 Anduril Industries has licensed Hivemind software for its Collaborative Combat Aircraft YFQ-44A program, per analyst and news reporting. SU013, SU004
CU008 Shield AI reported FY2024 total revenue of $267M, representing 64% year-over-year growth from an estimated FY2023 base of approximately $163M. SU004, SU005
CU009 Live MQ-20 Avenger test flights demonstrated Hivemind operating the aircraft in contested-environment air scenarios. SU010, SU013
CU010 Shield AI management has guided for approximately 80% revenue growth in 2026, implying roughly $481M in forward revenue. SU004, SU005
CU011 USASpending.gov confirms multiple sequential contract awards to Shield AI from at least three distinct federal agencies. SU003, SU001, SU002
CU012 Shield AI serves customers across US DoD agencies, allied international forces, and commercial defense prime integrators. SU007, SU013
CU013 V-BAT unit deliveries are estimated in the several-hundred range based on production scale references in company and analyst materials. SU006, SU013
CU014 The USCG $198M IDIQ represents the largest single publicly confirmed customer commitment in Shield AI's history and anchors near-term revenue. SU001, SU002, SU003
CU015 The Anduril Hivemind license appears to be the first publicly known pure-software licensing arrangement with a defense prime, establishing Shield AI's B2B2G channel. SU013, SU004
CU016 Hivemind software licensing is less capital-intensive than V-BAT hardware sales and may generate recurring revenue with superior unit economics. SU004, SU013
CU017 Romania and Japan are the only publicly named allied-nation V-BAT deployments; 8+ additional countries are claimed but not independently confirmed. SU013, SU014
CU018 V-BAT has been used in real operational scenarios by US Navy and US Coast Guard personnel in maritime patrol and surveillance roles. SU001, SU006
CU019 The HII ROMULUS Hivemind integration was completed within 6 weeks, confirming rapid platform onboarding speed. SU009
CU020 Shield AI raised $2B at a $12.7B valuation in March 2026, primarily to scale customer delivery capacity and product development. SU004, SU005
CU021 Shield AI does not publicly disclose a complete customer list, limiting independent verification of the claimed 10+ country deployment footprint. SU007, SU016
CU022 The USCG IDIQ, if annualized, would account for approximately 74% of the reported $267M FY2024 revenue — an elevated single-customer concentration ratio. SU001, SU017
CU023 Analysts note Shield AI's $12.7B valuation depends substantially on a software-pivot narrative not yet reflected in reported revenue mix. SU017
CU024 No public evidence of customer cancellations, program terminations, or customer-initiated disengagements has been found through May 2026. SU003, SU016
CU025 The USCG's $198M IDIQ represents a repeat and deepening engagement following earlier procurement, indicating incumbent advantage and structural retention. SU001, SU003
CU026 The percentage of Shield AI revenue attributable to international customers versus US government agencies is not publicly disclosed. SU007, SU016
CU027 V-BAT's track record in USCG operations serves as a reference-customer proof point to accelerate allied-nation adoption decisions. SU001, SU014
CU028 Shield AI serves both direct hardware buyers (V-BAT) and software licensees (Hivemind), creating two distinct procurement pathways with different switching costs. SU006, SU013
CU029 GE Aerospace's formal propulsion collaboration for X-BAT represents a Tier-1 prime partner engagement that validates Shield AI's platform ambition. SU024
CU030 Defense prime integrators purchasing Hivemind software (HII, Anduril) represent a B2B2G indirect channel allowing Shield AI to reach end-user governments through prime partners. SU009, SU013
CU031 The USAF A-GRA program integration with MQ-20 Avenger demonstrates Air Force customer proof at the program level, with Hivemind serving as the prime autonomy stack. SU010, SU013
CU032 USASpending.gov data shows Shield AI has received contract awards from at least three distinct federal agencies — USCG, Navy, and Air Force. SU003, SU006
CU033 Shield AI reports no government program terminations in FY2024 or FY2025 public commentary, consistent with a stable customer retention profile. SU004, SU021
CU034 V-BAT Block Upgrade adds Hivemind autonomous AI, SATCOM connectivity, and heavy-fuel engine as incremental capability packages for existing V-BAT customers. SU006, SU008
CU035 Hivemind software licensing to third-party platforms extends Shield AI's customer reach without requiring direct hardware sales, enabling lower-friction customer acquisition. SU013, SU007
CU036 The MQ-35A designation is a milestone in a multi-year qualification process; full programmatic production buy quantities are not yet confirmed in public records. SU006, SU013
CU037 Shield AI has achieved multi-year customer retention with the US Coast Guard and Navy based on sequential contract awards and ongoing qualification activity. SU001, SU003
CU038 International allied-nation V-BAT expansion supports future Foreign Military Sales pathways, enabling US government to facilitate and finance allied purchases. SU014, SU017
CU039 No publicly disclosed customer satisfaction scores, NPS metrics, CSAT data, or formal retention rates have been found for any Shield AI product line.
CU040 The concentration of approximately 100% of publicly confirmed Shield AI revenue with US DoD customers creates significant dependency risk if procurement priorities shift. SU017, SU013
CR001 Shield AI products — V-BAT, Hivemind, and X-BAT — are likely subject to ITAR controls under 22 CFR Part 121 Category XI (Military Electronics) or Category XV (Spacecraft and Related Articles), requiring export licenses for allied customer sales. SR001, SR002
CR002 DoDD 3000.09 (Autonomy in Weapon Systems) requires human authorization for lethal decisions by autonomous systems, structurally constraining Hivemind's operational scope in offensive combat missions. SR026, SR005
CR003 No ITAR violations, regulatory enforcement actions, or legal sanctions against Shield AI have been found in publicly available records as of May 2026. SR012
CR004 X-BAT requires NAVAIR airworthiness certification for military use and FAA UAS authorization for civil airspace operations, both multi-year processes that introduce schedule risk. SR013, SR014
CR005 Shield AI, as a defense contractor, is subject to FAR and DFARS compliance requirements including Controlled Unclassified Information (CUI) handling and CMMC cybersecurity standards. SR006, SR007
CR006 Shield AI's CMMC certification level and compliance posture under 32 CFR Part 170 are not publicly disclosed, limiting independent verification of cybersecurity compliance status. SR007
CR007 Shield AI announced a first-flight target of fall 2026 for X-BAT; as of May 2026, the vehicle has not yet flown and the program is in pre-flight development. SR022, SR013
CR008 GE Aerospace is the sole publicly identified propulsion provider for X-BAT, using the F110 turbofan engine; no alternative propulsion vendor has been disclosed. SR030, SR013
CR009 X-BAT development is estimated to require $1B+ in capital expenditure to reach production, based on analyst estimates; Shield AI has not disclosed an official program budget. SR008, SR009
CR010 Shield AI's 2025 revenue guidance of $500M+ was reportedly missed, with analyst estimates placing actual 2025 revenue at approximately $300M — a meaningful shortfall that undermines near-term growth credibility. SR008, SR009
CR011 The $12.7B Series G valuation implies approximately 50% software revenue mix by 2028, a transition not yet begun to materialize in reported revenue figures. SR008
CR012 The US Coast Guard $198M IDIQ, if annualized, represents approximately 74% of reported FY2024 revenue of $267M — an elevated single-customer concentration ratio. SR023, SR024
CR013 AWS cloud services appear to underpin the Hivemind Forge development environment based on Shield AI's official technical documentation, representing a cloud platform dependency. SR017
CR014 The Aechelon Technology acquisition announced in March 2026 introduces post-acquisition integration risk; integration timeline, cost, and key personnel retention have not been publicly disclosed. SR010, SR011
CR015 DoD budget continuing resolutions, sequestration events, or program-level cancellations represent a structural revenue risk for Shield AI given that approximately 100% of confirmed revenue derives from US government customers. SR009
CR016 Brandon Tseng (CEO) and Ryan Tseng (COO/President) represent key-man concentration; both co-founders are operationally active and their departure would reduce strategic credibility and government relationship access. SR015, SR016
CR017 Shield AI announced intent to grow from approximately 1,400 employees to 3,000+ following the March 2026 Series G close, representing a 2x+ headcount expansion with associated organizational capacity risk. SR015, SR016
CR018 A significant portion of Shield AI engineering roles require US security clearances, which can take 6-18 months to process, representing a hiring bottleneck that could slow X-BAT and Hivemind development programs. SR009
CR019 Hivemind's GPS-denied SLAM navigation performance in low-texture or featureless environments (e.g., open ocean, uniform desert terrain) has not been independently characterized in public sources; company claims GPS-denied capability without published edge-case characterization. SR017
CR020 No SOC 2, FedRAMP, DISA APL, NIAP Common Criteria, or equivalent security certification for Hivemind or EdgeOS has been found in public disclosures as of May 2026. SR017
CR021 Shield AI has not publicly described OTA (over-the-air) software update procedures, version management, or lifecycle support protocols for its fielded V-BAT and Hivemind-enabled fleets. SR017
CR022 Comms-denied swarm resilience for Hivemind Commander module — specifically its behavior when inter-vehicle communications are degraded or jammed — has not been publicly documented or independently validated at operational scale. SR009
CR023 X-BAT's Group-5 platform complexity — supersonic-capable VTOL with thrust-vectoring and an autonomous AI pilot — exceeds the hardware complexity of Shield AI's prior V-BAT production track record, introducing elevated integration risk. SR019, SR020
CR024 Shield AI remains pre-profit and is burning cash in pursuit of a likely IPO; the $2B Series G at $12.7B valuation represents late-stage private capital that must be deployed efficiently before the next capital event. SR010
CR025 The March 2026 $2B Series G at $12.7B valuation implies aggressive revenue growth and software mix assumptions that are not yet validated by Shield AI's publicly reported financial performance. SR010, SR008
CR026 The DoD A-GRA program (Air-Ground Reconnaissance Autonomy) represents a meaningful revenue and programmatic concentration risk if the program is restructured, descoped, or has its autonomy requirements modified. SR009
CR027 No IP disputes, patent infringement claims, or litigation involving Shield AI have been found in public legal records, court filings, or USPTO records as of May 2026. SR012
CR028 22 CFR Part 121 Category XI covers military electronics including autonomous weapons systems guidance and control electronics, which would encompass Hivemind's mission autonomy hardware components. SR001, SR002
CR029 Export of ITAR-controlled defense articles to allied nations requires a DSP-5 permanent export license, Technical Assistance Agreement, or Manufacturing License Agreement approved by DDTC; this process creates timeline risk for allied customer transactions. SR001, SR004
CR030 The Export Administration Regulations (EAR) may apply to certain Shield AI AI software components with dual-use commercial and military applications, adding BIS licensing complexity to the export compliance stack. SR003
CR031 DoDD 3000.09 (signed January 2023) explicitly requires that autonomous and semi-autonomous weapons systems be designed to allow commanders and operators to exercise appropriate levels of human judgment over the use of force. SR026, SR005
CR032 The regulatory framework governing autonomous and semi-autonomous weapons is evolving rapidly at both US domestic policy and international treaty levels, creating forward-looking regulatory uncertainty for companies like Shield AI. SR027, SR028
CR033 Human rights organizations including Human Rights Watch are actively lobbying for restrictions on fully autonomous weapons, increasing the reputational and regulatory pressure environment around AI-piloted combat systems. SR027
CR034 A US government continuing resolution (CR) that restricts DoD spending to prior-year authorization levels would delay new task order issuance under existing IDIQ contracts and slow new program awards to Shield AI. SR009
CR035 The Aechelon Technology acquisition closed or was announced in March 2026; integration timeline, team retention packages, and technology consolidation plan are not publicly disclosed as of May 2026. SR011, SR010
CR036 Post-acquisition integration of the Aechelon team could divert management attention and engineering resources from core V-BAT production and X-BAT development programs during a critical period. SR011
CR037 X-BAT production deployment will require NAVAIR airworthiness certification, USAF/Navy operational test and evaluation (OT&E), and potentially FAA authorization for non-restricted airspace — a multi-regulatory approval pathway with at minimum a 3-5 year timeline. SR019, SR020, SR022
CR038 Analysts note that Shield AI's $12.7B valuation rests on a software-pivot narrative that is not yet reflected in reported revenue mix, with hardware (V-BAT) still dominating recognized revenue as of the available public data. SR008, SR029
CR039 Series G proceeds must fund both ongoing operations (cash burn) and X-BAT development capex estimated at $1B+, plus the Aechelon integration costs — creating capital deployment pressure that may require additional fundraising before an IPO. SR010, SR011
CR040 Shield AI's 2025 revenue miss — if confirmed at approximately $300M vs $500M+ guidance — creates investor credibility risk heading into an expected IPO, potentially compressing the achievable public market valuation multiple. SR008, SR009
CV001 Shield AI closed its Series G round in March 2026, raising $2 billion at a post-money valuation of $12.7 billion. SV001, SV012, SV014
CV002 The Series G was led by Advent International and J.P. Morgan, according to multiple news reports. SV001, SV011
CV003 The $12.7 billion Series G valuation represents a 140% increase from the $5.3 billion Series F valuation completed in March 2025. SV001, SV013, SV016
CV004 Total cumulative funding raised by Shield AI through the Series G close is approximately $3.52 billion. SV001, SV011, SV002
CV005 The Series G round included approximately $1.5 billion in equity and $500 million in debt or structured financing, per tech-insider reporting. SV002, SV003
CV006 Shield AI reported FY2024 revenue of $267 million, representing 64% year-over-year growth from approximately $163 million in FY2023. SV001, SV009
CV007 Management projects FY2026 revenue exceeding $540 million, representing approximately 102% growth over FY2024. SV009, SV008
CV008 At the $12.7 billion Series G post-money valuation, Shield AI trades at approximately 47x its FY2024 revenue of $267 million. SV001, SV004
CV009 Shield AI reported FY2023 revenue of approximately $163 million, based on back-calculation from the 64% growth rate to $267 million in FY2024. SV011, SV004
CV010 If the FY2026 guidance of $540 million is met, the implied forward revenue multiple drops from 47x to approximately 23x. SV009, SV006
CV011 Shield AI's investment thesis centers on Hivemind Enterprise as the 'operating system' for military autonomy, offering durable switching costs across 15+ deployed platforms. SV004, SV001
CV012 Defense AI spending is projected to grow from approximately $9 billion in 2024 to over $40 billion by 2032, creating a large structural tailwind for Shield AI's addressable market. SV004, SV020
CV013 The primary anti-thesis is Shield AI's ongoing software pivot risk: hardware revenue from V-BAT still dominates and the software licensing model is unproven at scale, with gross margins not publicly disclosed. SV007, SV004
CV014 Shield AI is substantially dependent on the US DoD and allied militaries as customers, creating single-category concentration risk. SV007, SV004
CV015 Shield AI has not achieved GAAP profitability and requires continued cash investment in R&D and manufacturing; the path to profitability depends on a margin improvement that has not yet been demonstrated. SV007, SV005
CV016 Anduril Industries is targeting approximately $60 billion in a new funding round as of early 2026, making it Shield AI's closest private peer by revenue and autonomy-platform positioning. SV018, SV019, SV021
CV017 Palantir Technologies trades at approximately 38–50x trailing twelve-month revenue on a market capitalization of approximately $110 billion, providing the key defense-AI software public-market precedent. SV005, SV025
CV018 Kratos Defense & Security Solutions trades at approximately 3–4x revenue on a market capitalization of approximately $4 billion, representing the hardware-centric downside comparable for Shield AI. SV005, SV017
CV019 L3Harris Technologies trades at approximately 1x revenue on a market capitalization of approximately $20 billion, representing the floor benchmark for highly commoditized large defense businesses. SV005, SV017
CV020 Axon Enterprise trades at approximately 15x revenue on a market capitalization of approximately $30 billion, serving as a high-growth government-technology software comparable. SV005, SV020
CV021 Acquinox Capital analysis rates Shield AI's $12.7 billion valuation as reflecting an early premium for AI-autonomy software that is justified by defense-spending tailwinds and the platform moat. SV004, SV001
CV022 Ainvest.com characterizes Shield AI's $12.7 billion valuation as hinging on a risky software pivot, warning that the hardware revenue base does not support the current premium multiple. SV007
CV023 IPO Club analysts estimate Shield AI's post-IPO valuation could reach approximately $24 billion if it achieves software-like gross margins and wins additional CCA program slots. SV006, SV005
CV024 Anduril's previously reported valuation of approximately $30.5 billion per Acquinox analysis provides a near-term baseline; the $60 billion target reflects a further step-up driven by the $20 billion US Army contract win in March 2026. SV017, SV019
CV025 Shield AI's gross margin breakdown by revenue segment (software vs. hardware vs. services) is not publicly disclosed, which is the single largest analytical gap in underwriting the current valuation. SV007, SV004
CV026 Shield AI's Series F in March 2025 raised $240 million at a $5.3 billion post-money valuation, from investors including Snowpoint Ventures and IVP, according to Bloomberg and TechCrunch. SV013, SV015, SV002
CV027 Federal spending databases (USASpending.gov) confirm the US government is Shield AI's primary contracting counterparty, with contract awards visible across DoD, USAF, USMC, and USCG programs. SV023, SV022
CV028 Shield AI was selected by the US Air Force as the mission autonomy software provider for the Collaborative Combat Aircraft program in February 2026, a highly significant contract win that directly drove the Series G valuation re-rating. SV001, SV016, SV004
CV029 Export control regulations (ITAR/EAR) governing AI autonomy software are evolving and could constrain Shield AI's international deployments, representing a tail risk to the bull case that assumes significant international revenue. SV007, SV013
CV030 Shield AI acquired Aechelon Technology in March 2026 to strengthen its simulation and high-fidelity training pipeline for Hivemind, which underpins the software platform defensibility thesis. SV002, SV003
CV031 In the bull scenario, Shield AI achieves $700 million or more in FY2026 revenue, reaches positive EBITDA, wins additional CCA increment slots, and executes an IPO or strategic transaction at $24–35 billion. SV006, SV004
CV032 In the base scenario, Shield AI delivers $540 million or more in FY2026 revenue, progresses toward contribution-margin break-even, and achieves an IPO or M&A exit in the 2026–2027 window at approximately $14–18 billion. SV009, SV005
CV033 In the bear scenario, revenue growth decelerates below 30% year-over-year, the software-pivot fails to improve the revenue mix, and multiple compression drives valuation to $5–8 billion below the Series G entry price. SV007, SV004
CV034 The expected IPO window for Shield AI is the 2026–2027 calendar year, based on CEO Gary Steele's background, the Series G investor base, and current market conditions. SV006, SV004, SV001
CV035 Strategic acquisition paths include Boeing, L3Harris, Northrop Grumman, and RTX as potential acquirers that would value Hivemind as a software platform layer for autonomous systems at potential acquisition multiples of $15–25 billion. SV004, SV005
CV036 Loss of the CCA program (full removal or >50% platform-count reduction) is the primary thesis-break trigger and would warrant an immediate re-underwrite or exit. SV004, SV007
CV037 Failure to demonstrate software gross margins above 40% through FY2026 would be a kill trigger, confirming the hardware-heavy revenue model and invalidating the software-platform premium multiple. SV007, SV005
CV038 Final diligence requires disclosure of gross margin by revenue segment (software licensing, hardware, and government services) as the minimum non-public financial data item. SV007, SV004
CV039 Cap-table and preference-stack analysis — including Series G liquidation preferences and any anti-dilution provisions — is a critical diligence item not publicly disclosed. SV007, SV011
CV040 The Palantir 10-K filing confirms that defense AI software companies can sustain 30–50x revenue multiples post-profitability and post government-wide contract wins, providing a public-market precedent for Shield AI's valuation. SV022, SV017
来源
编号出版方标题引文
SO001 Shield AI About Shield AI What does the military of 2030 look like, and what role does autonomy play? The answer: AI pilots powering every military asset: aircraft, drones, ships, satellites, and submarines.
SO002 Wikipedia Shield AI
SO003 TechCrunch Shield AI raises $240M at a $5.3B valuation to commercialize its AI drone tech Shield AI is now the second-largest defense tech startup in the U.S. by valuation, after Anduril.
SO004 Shield AI / PR Newswire Shield AI Appoints Gary Steele as CEO, Ryan Tseng Named President Shield AI has achieved extraordinary growth, expanding to more than 900 teammates, raising over $1 billion in funding, and becoming the fourth multi-billion-dollar defense technology company.
SO005 Breaking Defense Shield AI names former Splunk CEO as new chief executive Shield AI has grown from a small San Diego-based start up to one of the Pentagon's so-called 'defense tech unicorns.'
SO006 TechCrunch Defense startup Shield AI lands $12.7B valuation, up 140%, after US Air Force deal The new round comes after Shield raised $240 million at a $5.3 billion valuation in March 2025. That means its value leaped 140% in one year.
SO007 Tracxn Shield AI – 2026 Company Profile & Team Shield AI has raised a total funding of $3.52B over 17 rounds. Its latest funding round was a Series G round on Mar 26, 2026 for $2B.
SO008 Aerospace America (AIAA) Shield AI Co-Founder Shares 10-Year Journey Reimagining Air Power In 2023, Shield AI became the first company to fly a combat aircraft with an AI pilot, pitting it against an F-16 operated by a U.S. Air Force weapon school pilot in a first-of-its-kind dogfight.
SO009 GovCon Wire Shield AI Appoints Gary Steele CEO, Ryan Tseng President
SO010 The Defense Post Shield AI's V-BAT Drone to Compete for Up to $800M in US Navy ISR Contracts
SO011 Shield AI Shield AI – Home Shield AI to acquire Aechelon, raise $2B at $12.7B valuation.
SO012 Bloomberg Drone Defense Startup Shield AI Lands $5.3 Billion Valuation
SO013 Shield AI X-BAT – World's First AI-Piloted VTOL Fighter Jet Hivemind unlocks AI-powered mission autonomy for X-BAT. Resilient execution in denied, degraded, or disconnected conditions — no GNSS, no comms, no human input required.
SO014 Shield AI V-BAT – Unmanned ISR and Targeting Aircraft V-BAT was competitively selected for the U.S. Coast Guard Maritime Unmanned Aircraft System Services program.
SO015 Human Rights Watch US Military's Dangerous Slide Toward Fully Autonomous Killing The United States Department of Defense's decision on February 27 to reject the artificial intelligence company Anthropic's ethical red lines for AI for military use is a clear sign that the Pentagon is unlikely to uphold meaningful safeguards on weapons' development.
SO016 Shield AI / PR Newswire Shield AI Signs Definitive Agreement to Acquire Martin UAV Shield AI will integrate its combat-proven autonomy software, Hivemind, into the V-BAT, reinforcing Shield AI's leadership position in defense-focused edge autonomy.
SO017 Wikipedia Shield AI MQ-35 V-BAT
SO018 Defense News Here's how Shield AI wants to boost V-Bat's capability on a contested battlefield
SO019 Washington Harbour Shield AI raises $240M at $5.3B valuation to scale Hivemind Enterprise The funds will enable Shield AI to expand the deployment of Hivemind Enterprise to OEMs, governments, and companies to empower the larger robotics and drone industrial base to build autonomy products.
SO020 International Finance Business Leader of the Week: Shield AI accelerates with new CEO Gary Steele
SO021 Tech Insider Shield AI's $1.5B Series G at $12.7B: Inside the CCA Win (2026)
SO022 Shield AI Booz Allen and Shield AI Partnership Brings New Autonomous Airborne Solutions to U.S. Military Booz Allen recently made its largest corporate venture capital investment to date in Shield AI's F-1 strategic funding round.
SO023 Craft.co Shield AI CEO and Key Executive Team
SO024 IPO Club Shield AI Stock vs. Anduril Shield AI: Autonomy at the Edge. Shield AI is obsessed with one problem. What happens when humans are cut off? No GPS. No comms. No joystick. Autonomy is not a feature. It is the product.
SO025 Shield AI Shield AI – Press Releases
SM001 DefenseScoop Billions for new uncrewed systems and drone-killing tech included in Pentagon's 2026 budget plan
SM002 Global Market Insights Military Drones Market Size, Share & Growth Report, 2026–2035
SM003 Shield AI Shield AI selected as mission autonomy provider for the U.S. Air Force Collaborative Combat Aircraft program
SM004 PR Newswire Shield AI Selected as Mission Autonomy Provider for the US Air Force Collaborative Combat Aircraft Program
SM005 The Aviationist YFQ-44A CCA Tests Shield AI's Hivemind and Anduril's Lattice AIs During Single Flight
SM006 Inside Unmanned Systems Report: What Unmanned Systems is America's Military Buying in 2026?
SM007 StartUs Insights Military Drones Report 2026: Key Insights
SM008 The Business Research Company Global Military Drones Market Report 2026
SM009 Coherent Market Insights Military Drone Market Share & Opportunities 2026–2033
SM010 CDO Magazine Pentagon Seeks $13.4bn for AI and Autonomy FY 2026 Budget Request
SM011 Defense One Defense department budget request goes hard on AI, autonomy
SM012 Mordor Intelligence UAV Market Size, Share & Industry Trends, 2031
SM013 INSS / National Defense University Strategic Innovation in the DoD FY 2026 RDTE Budget: Leveraging Disruptive Technologies
SM014 GrantedAI Pentagon AI Budget Hits $13 Billion: Beyond DARPA, Who Else Is Funding?
SM015 Fortune Business Insights Autonomous Defense Platforms Market Size, Share [2026–2034]
SM016 Technavio Military Drone Market Growth Analysis — Size and Forecast 2026–2030
SM017 The AI Insider Pentagon Elevates AI as Cornerstone of FY2026 Defense Tech Strategy
SM018 HigherGov SMALL UNMANNED AERIAL SYSTEMS Budget — FY2026
SM019 Global Growth Insights AI in Military Market Size Insights Report 2026–2035
SM020 Stratistics MRC Military AI Systems Market: CAGR, size, share, trends, growth, value
SM021 360iResearch Autonomous Military Systems Market Size & Share 2026–2032
SM022 Breaking Defense Collaborative Combat Aircraft Coverage — Tag Page
SM023 labla.org Pentagon's $13.4B AI Budget: Every Dollar in 2026
SM024 MarketsandMarkets Autonomous Navigation and Weapons Market Research Reports
SM025 MarketsandMarkets Military AI Systems Market Size, Share & Global Forecast to 2033
SP001 Anduril Industries YFQ-44A Flies with Mission Autonomy Software from Anduril and Shield AI The YFQ-44A successfully flew with both Hivemind and Lattice AI software stacks installed, demonstrating A-GRA open-architecture modularity.
SP002 Aerospace Global News Anduril YFQ-44 Fury flown by Shield AI Hivemind in new CCA test
SP003 Acquinox Capital Anduril Industries: Investor Insights — Rewriting the Defence Industrial Base at a $30.5B Valuation Anduril's commercial-inspired model enables 40–45% gross margins versus 8–10% for legacy primes.
SP004 Kavout Is Anduril Industries the Next Defense Titan?
SP005 The War Zone (TWZ) USMC XQ-58 Valkyrie Development Makes Leap Forward With New Contract
SP006 The Aviationist Northrop Grumman, Kratos to Develop XQ-58 Valkyrie as CCA for the USMC
SP007 DefenseScoop Marine Corps picks Northrop Grumman, Kratos team to build CCA drones
SP008 Boeing Boeing Defense, Space & Security Partners with Palantir to Accelerate AI Adoption Across Defense, Classified Programs The partnership enables Boeing to embed Palantir's AI capabilities across its defense production and classified program workflows.
SP009 UBOS Tech Anduril Targets $60B Valuation in New Funding Round
SP010 Hot Paths Shield AI Hivemind Flew Anduril Drone Wingman for First Time
SP011 Jared Watkins (independent researcher) Shield AI — Research Notes
SP012 RABBT Research The AI Defense Vanguard: Anduril, Palantir, and Shield AI — A Battle for the Pentagon's Future
SP013 Parsers.vc Defence AI Soars: Anduril Eyes $60B Valuation as Pentagon Boosts Cyber Spending
SP014 National Defense Magazine Boeing Partners with Palantir to Improve Defense Production
SP015 Shield AI Hivemind EdgeOS: A Game-Changer for Autonomous Robotics
SP016 General Atomics Aeronautical Systems (GA-ASI) Avenger Unmanned Aircraft System
SP017 Startup Wired Anduril Nears $60B Valuation After Major VC Investment
SP018 Palantir Technologies Palantir Gotham — Defense AI Platform
SP019 AI News Hub Palantir's AI Revolution: How One Company is Transforming Military Defense
SP020 Sea Power Magazine Kratos Receives $34.8M Contract for Valkyrie Mission System Integration
SP021 Anduril Industries Transforming U.S. Defense Capabilities with Advanced Technology
SP022 General Atomics Aeronautical Systems (GA-ASI) GA-ASI Unmanned Aircraft Portfolio
SP023 L3Harris Technologies L3Harris Autonomous Systems Capabilities
SP024 Benzinga Palantir's Monopoly Is Breaking: It's No Longer Pentagon's Only Favorite Palantir's near-monopoly in Pentagon AI is fracturing as Shield AI, Anduril, Applied Intuition, and others carve out specialized defense AI market segments.
SP025 Defence Industry EU Kratos secures contract expansion to advance Valkyrie unmanned aerial systems for U.S. Marine Corps
SI001 Tech Insider Shield AI Raises $2B at $12.7B Valuation [2026]
SI002 Tech Insider Shield AI's $1.5B Series G at $12.7B: Inside the CCA Win [2026]
SI003 Sacra Shield AI at $267M/year growing 64% YoY
SI004 Acquinox Capital Shield AI: Why the $12 Billion Defence Tech Opportunity May be Taking Off
SI005 Tech Life Blog Shield AI Just Raised $2 Billion and Doubled Its Valuation in a Year
SI006 Idlen Shield AI: $2B Series G, $12.7B Valuation, Aechelon Acquired
SI007 PM Insights Shield AI Valuation | PM Insights
SI008 TechCrunch Defense startup Shield AI lands $12.7B valuation, up 140%, after US Air Force deal
SI009 IPO Club Shield AI's Path to a $24 Billion Valuation
SI010 AI Invest Shield AI's $12.7B Valuation Hinges on a Risky Software Pivot
SI011 Inside Unmanned Systems Shield AI Wins $198 Million Contract to Provide U.S. Coast Guard with V-BAT Platform
SI012 The Defense Post Shield AI's V-BAT Drone to Compete for Up to $800M in US Navy ISR Contracts
SI013 GovConWire Shield AI to Deliver VTOL Drone System to Coast Guard Under $198M Award
SI014 Shield AI Shield AI's V-BAT Selected for $198 Million Contract to Provide U.S. Coast Guard Maritime UAS Services
SI015 USASpending.gov Shield AI Recipient Profile — Federal Contract Awards
SI016 AI Certs Shield AI rockets to $5.3B valuation
SI017 Aechelon Technology Shield AI to Acquire Software Simulation Company Aechelon and Raise $2B at $12.7B Valuation
SI018 Doolpa Shield AI $2B Series G Funding, Aechelon Acquisition (March 2026)
SI019 PR Newswire Shield AI's V-BAT Selected for $198 Million Contract — PR Newswire
SI020 Shield AI Hivemind EdgeOS: A Game-Changer for Autonomous Robotics
SI021 Shield AI Hivemind Enterprise — Shield AI
SI022 Shield AI V-BAT Unmanned Aircraft System — Shield AI
SI023 Shield AI About Shield AI
SI024 Shield AI Careers at Shield AI
SI025 Business Wire Shield AI to Acquire Software Simulation Company Aechelon and Raise $2B at $12.7B Valuation
SE001 Shield AI Hivemind EdgeOS: A Game-Changer for Autonomous Robotics
SE002 Amazon Web Services Shield AI and AWS Collaborate to Deliver Mission Autonomy at Fleet Scale
SE003 Defense Tech Signals Shield AI: Autonomy Stack Deep Dive
SE004 Milivox Media Shield AI V-BAT: Autonomous VTOL Teaming
SE005 Complete AI Training HII and Shield AI Complete First Maritime Hivemind Integration
SE006 Defence Blog AI Takes Control of Avenger Drone in Live Tests
SE007 Defense Feeds Shield AI Launches X-BAT
SE008 Defence Agenda Shield AI X-BAT VTOL Stealth Jet
SE009 Ragex Defense Shield AI X-BAT VTOL Fighter
SE010 The Defense News Shield AI Unveils the World's First AI-Piloted VTOL Fighter Jet — the X-BAT
SE011 PR Newswire / Shield AI Shield AI Unveils V-BAT Block Upgrade: Hivemind Autonomy, SATCOM, and Heavy-Fuel Engine
SE012 Jared Watkins Research Shield AI — Aerial Drones Research Profile
SE013 Shield AI X-BAT Product Page
SE014 Shield AI ViDAR Product Page
SE015 Shield AI Shield AI Press Releases
SE016 Shield AI Hivemind Forge Product Page
SE017 Shield AI Hivemind Commander Product Page
SE018 Shield AI Hivemind Pilot Product Page
SE019 Shield AI X-BAT Announcement Page
SE020 Shield AI X-BAT First Reveal Page
SE021 Shield AI Hivemind EdgeOS Product Page
SE022 The Aviationist Shield AI Unveils X-BAT
SE023 Defense News Shield AI Unveils X-BAT Autonomous Vertical Takeoff Fighter Jet
SE024 PR Newswire / Shield AI Shield AI Unveils X-BAT: An AI-Piloted VTOL Fighter Jet for Contested Environments
SE025 GE Aerospace GE Aerospace and Shield AI to Collaborate on Propulsion for X-BAT Vehicle Program
SU001 PR Newswire / Shield AI Shield AI Awarded $198 Million Contract with US Coast Guard
SU002 Defense One Shield AI Wins $198 Million Contract for Coast Guard V-BAT Drones
SU003 USASpending.gov Shield AI Federal Contract Award History
SU004 Defense One Shield AI Raises $2B at $12.7B Valuation
SU005 DC Velocity Shield AI Raises $2B and Acquires Aechelon
SU006 Shield AI V-BAT Product Page
SU007 Shield AI About Shield AI
SU008 Shield AI Shield AI Blog
SU009 Complete AI Training HII and Shield AI Complete First Maritime Hivemind Integration
SU010 Defence Blog AI Takes Control of Avenger Drone in Live Tests
SU011 Breaking Defense Shield AI Wins $198M USCG Contract for V-BAT
SU012 C4ISRNET Shield AI Wins $198M USCG V-BAT Contract
SU013 Defense Tech Signals Shield AI Autonomy Stack Deep Dive
SU014 Milivox Media Shield AI V-BAT Autonomous VTOL Teaming
SU015 Nextgov / FCW Shield AI Wins $198 Million Coast Guard Contract
SU016 Sacra Research Shield AI Company Profile and Revenue Analysis
SU017 AI Invest Shield AI $12.7B Valuation Hinges on Risky Software Pivot
SU018 Shield AI X-BAT Announcement
SU019 Shield AI Hivemind Pilot Product Page
SU020 Shield AI Hivemind Commander Product Page
SU021 Shield AI Shield AI Press Releases
SU022 IR-IA Shield AI Receives $198 Million Contract to Produce V-BAT ISR Drones for US Coast Guard
SU023 PR Newswire / Shield AI Shield AI Unveils X-BAT
SU024 GE Aerospace GE Aerospace and Shield AI Collaborate on X-BAT Propulsion
SU025 Defence Agenda Shield AI X-BAT VTOL Stealth Jet
SR001 eCFR 22 CFR Part 120 — ITAR Purpose and Definitions
SR002 eCFR 22 CFR Part 121 — The United States Munitions List
SR003 Bureau of Industry and Security Export Administration Regulations (EAR)
SR004 DDTC Directorate of Defense Trade Controls — ITAR Compliance Portal
SR005 Cornell Legal Information Institute 22 CFR Part 120 — ITAR Definitions (LII)
SR006 SAM.gov Shield AI Opportunities — SAM.gov
SR007 Acquisition.gov FAR 52.212-4 Contract Terms and Conditions — Commercial Products
SR008 AI Invest Shield AI $12.7B Valuation Hinges on Risky Software Pivot
SR009 Defense Tech Signals Shield AI Autonomy Stack Deep Dive
SR010 Defense One Shield AI Raises $2B at $12.7B Valuation
SR011 DC Velocity Shield AI Raises $2B and Plans Aechelon Acquisition
SR012 Jared Watkins Research Shield AI Research Profile
SR013 Defence Agenda Shield AI X-BAT VTOL Stealth Jet
SR014 Defense Feeds Shield AI Launches X-BAT
SR015 Shield AI About Shield AI
SR016 Shield AI Shield AI Press Releases
SR017 Shield AI Hivemind EdgeOS Technical Overview
SR018 Milivox Media Shield AI V-BAT Autonomous VTOL Teaming
SR019 The Aviationist Shield AI Unveils X-BAT
SR020 Defense News Shield AI Unveils X-BAT VTOL Fighter Jet
SR021 Ragex Defense Shield AI X-BAT VTOL Fighter
SR022 Shield AI X-BAT Product Page
SR023 PR Newswire / Shield AI Shield AI Awarded $198M USCG Contract
SR024 USASpending.gov Shield AI Federal Contract Award History
SR025 Defense One Shield AI Wins $198M Coast Guard V-BAT Contract
SR026 Office of the Under Secretary of Defense for Policy DoD Directive 3000.09 — Autonomy in Weapon Systems
SR027 Human Rights Watch US Military's Dangerous Slide Toward Fully Autonomous Killing
SR028 Government Accountability Office GAO — Artificial Intelligence: DOD Should Improve Strategies, Inventory, and Collaboration
SR029 C4ISRNET Shield AI Raises $2 Billion at $12.7 Billion Valuation
SR030 GE Aerospace GE Aerospace and Shield AI Collaborate on Propulsion for X-BAT Vehicle Program
SV001 TechCrunch Defense Startup Shield AI Lands $12.7B Valuation, Up 140%, After US Air Force Deal
SV002 Tech Insider Shield AI Closes $1.5B Series G at $12.7B Valuation with CCA and Aechelon
SV003 Tech Insider Shield AI Raises $2 Billion Series G — Defense AI and Autonomous Warfare 2026
SV004 Acquinox Capital Shield AI — Why the $12 Billion Defence-Tech Opportunity May Be Taking Off
SV005 IPO Club Shield AI Stock vs. Anduril: Pre-IPO Comparison
SV006 IPO Club Shield AI $24 Billion Valuation — IPO Upside Estimate
SV007 AI Invest Shield AI $12.7B Valuation Hinges on Risky Software Pivot
SV008 Dnyuz / New York Times Defense Startup Shield AI Is Projecting More Than $540 Million in Revenue This Year
SV009 Yahoo Finance Defense Startup Shield AI Projecting $540M+ Revenue in 2026
SV010 AI2 Work Shield AI's $12.7B Valuation Signals Defense Autonomy Has Arrived
SV011 Crunchbase Shield AI Funding Rounds — Crunchbase
SV012 CNBC Shield AI Raises $2 Billion at $12 Billion Valuation
SV013 Bloomberg Drone Defense Startup Shield AI Lands $5.3 Billion Valuation
SV014 Bloomberg Shield AI Raises $2 Billion at $12.7 Billion Valuation
SV015 TechCrunch Shield AI Raises $240 Million at a $5.3 Billion Valuation to Commercialize AI Drone Tech
SV016 Breaking Defense Shield AI Raises $2B Series G at $12.7B Valuation
SV017 Acquinox Capital Anduril Industries — Rewriting the Defence Industrial Base at a $30.5B Valuation
SV018 Ubos Tech Anduril Targets $60B Valuation in New Funding Round
SV019 Parsers VC Defence AI Soars — Anduril Eyes $60B
SV020 Kavout Is Anduril Industries the Next Defense Titan?
SV021 Startup Wired Anduril Nears $60B Valuation After Major VC Investment
SV022 SEC EDGAR Palantir Technologies Inc — 10-K Annual Report Filings (CIK: PLTR)
SV023 USASpending.gov Shield AI Federal Contract Award History
SV024 SAM.gov Shield AI Opportunities — SAM.gov
SV025 Yahoo Finance Palantir Technologies (PLTR) — Stock Quote and Financials
SV026 Defense One Shield AI Raises $2 Billion; Valuation Hits $12.7 Billion
SV027 Sacra Shield AI Research — Sacra
SV028 PitchBook Shield AI Valuation 2026 — PitchBook
SV029 Defense News Shield AI Achieves $12.7 Billion Valuation
SV030 Axios Shield AI Fundraising and Valuation — 2026