Harmattan AI
服务于 ISR、反无人机和协同空战的主权防务自主系统栈
Harmattan AI 用两年时间拿出了少见的主权防务证明点——两个国防部 Program of Record 和 Dassault Rafale F5 合作——但公开财务披露缺位,投资判断仍只能停在继续研究。
封面要素
公司概况
Harmattan AI 于 2024 年 4 月在巴黎成立,作为欧洲防务初创公司扩张速度异乎寻常。公司搭建纵向一体化自主系统栈,覆盖 Sonora 训练 ISR 无人机、Sahara 机载 SAR 载荷、Gobi 和 Gobi Tempest 反无人机拦截器、Barkhan 打击平台,以及 Kalahari 指挥控制层。成立约 18 个月内,它同时拿下法国和英国国防部正式列装项目;2026 年 1 月,Dassault Aviation 领投其 $200M Series B,估值 $1.4B。Dassault 合作把 Harmattan 推向 Rafale F5 和未来 UCAS 项目的嵌入式 AI 能力开发;Skyeton 合作则把其传感器栈延伸到面向 NATO 市场的乌克兰 Raybird 平台。
- 成立时间
- 2024-04-01
- 创始人
- Mouad M'Ghari, Martin de Gourcuff, Edouard Rosset, Marc Grelet
- 创立地点
- Paris, France
- 总部
- Paris, France
- 产品
- Harmattan 销售一套集成式防务自主系统栈:Sonora 服务于部队战备和操作员训练,Sahara 服务于全天候 SAR ISR,Gobi 和 Gobi Tempest 服务于动能反无人机,Barkhan 服务于自主精确打击,Kalahari 则在分布式自主系统之间承担任务编排和 C2。
- 客户
- 国家国防部、NATO 及盟国武装力量,以及需要主权嵌入式 AI、ISR 和反无人机能力的战略防务主承包商伙伴。
- 商业模式
- 以合同驱动的防务系统销售,同时配套自主软件、C2 集成、载荷集成、工业制造,以及与主承包商和政府买方的战略共同开发项目。
- 阶段
- Series B private company
- 融资情况
- 截至 2026 年 1 月,估计累计融资约 $242M,其中包括 Dassault Aviation 领投的 $200M Series B,投后估值 $1.4B。
执行摘要
主要优势
- 创立约 18 个月内拿下两个国家国防部 Program of Record,对一家欧洲防务创业公司来说是很强的早期客户验证。
- Dassault Aviation 合作把 Harmattan 带进 Rafale F5 和未来 UCAS 嵌入式 AI 工作流,战略护城河和退出可选性都更强。
- Harmattan 的垂直一体化栈覆盖 ISR、反无人机、打击和 C2,卖的是系统级能力,而不是单一传感器或机体。
- 关于自主架构、SAR 性能和动能拦截的产品证据足够细,按公司阶段看也有技术差异化。
主要风险
- 公司没有公开收入、利润率、烧钱速度或单位经济性,$1.4B 估值无法用财务基本面压力测试。
- 对中国稀土投入和电池化学体系的供应链依赖仍未解决,可能拖慢制造放量。
- 治理仍不透明:董事会构成、投资人权利和保护性条款都未公开披露。
- Anduril、Helsing、Shield AI、TEKEVER 以及专业 C-UAS 厂商等资本更厚的同行,可能在定价、人才和采购准入上挤压 Harmattan。
- Harmattan 的战略溢价取决于 Dassault / Rafale F5 关系中的交付零失误和范围可持续;任何一项落空都会削弱判断。
未决问题
- 当前收入、利润率结构和现金跑道未公开披露。
- Series B 投资人具体构成和优先股条款仍未公开。
- 管理层声称每月数千台产能,但缺少独立证据验证这个生产规模。
- Harmattan 在 Rafale F5 和 UCAS AI 项目中的确切范围,公开细节仍有限。
- 董事会构成和正式治理权利未公开说明。
目录
01公司概况
1.1 身份、创立与法律版图
Harmattan AI 按法国法注册为简化股份公司(SAS),登记地址为法国巴黎 75002 邮区 rue du Mail 1 号,增值税号 FR39978035392。公司由 Mouad M'Ghari 以及 Martin de Gourcuff、Edouard Rosset、Marc Grelet 等联合创始人于 2024 年 4 月创办,从第一天起就把自己定位为防务科技公司,而不是消费 AI 初创。巴黎注册地址同时也是法定住所和运营总部;截至 2026 年 6 月访问的招聘页确认,公司还在英国、瑞士、摩洛哥、阿联酋和美国设有办公室。2026 年 6 月发布的开放岗位覆盖巴黎、Lausanne、Wissous、伦敦、Arlington VA、UAE 和波兰,显示其运营版图确实横跨多国。 公司将核心战略信条概括为「协同饱和」——自主机器人部队以统一整体的方式,在完整作战谱系中大规模行动;军事优势不再只取决于单个平台性能,而转向能否大规模生产、部署并协同自主系统。该理念同时牵引产品路线图和制造战略:Harmattan 把量产准备度放在一阶目标,而不是下游愿景。公司使用条款页面最后更新于 2025 年 11 月 10 日,并引用 GDPR、CCPA/CPRA 和 CalOPPA 下的合规义务,确认其数据基础设施托管在欧盟,也显示其隐私合规姿态有意对齐防务行业企业要求。截至 2026 年 6 月,公司网站将其归入「防务科技」行业;在 Dassault Aviation 合作改变对外叙事之前,公司曾自称「下一代防务主承包商」。[CO001, CO002, CO003, CO004, CO005, CO006]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口 |
|---|---|---|---|---|
| 成立时间 | April 2024 | 2024-04 | 高 | None |
| 总部 | 法国巴黎(1 rue du Mail, 75002) | 2025-11 | 高 | None |
| 法律形式 | SAS(société par actions simplifiée),法国 | 2025-11 | 高 | 仅有 VAT 信息;完整公司章程未公开 |
| 阶段 | Series B | 2026-01 | 高 | None |
| 投后估值 | $1.4 billion | 2026-01-12 | 高 | Series B 前估值未披露 |
| 累计融资(估计) | ~$242 million | 2026-01-12 | 中 | Dassault 具体投资份额和共同投资方未披露 |
| 员工数 | >130 名员工 | 2026-01 | 中 | 准确人数和构成未披露 |
| 收入 / ARR | 未披露 | 低 | 没有公开财务指标 | |
| 法国国防部合同 | 1,000 架无人机(SORONA);首个正式列装项目 | 2025-07 | 高 | 合同金额未披露 |
| 英国国防部合同 | 最多 3,000 套系统;第二个正式列装项目 | 2025-09 | 高 | 合同金额未披露 |
| 办公地点 | 法国、英国、瑞士、摩洛哥、阿联酋、美国 | 2026-06 | 中 | 各国员工数未披露 |
数值来自 Harmattan AI 官方新闻稿、使用条款(Nov 2025)和 Defense News(Jan 2026)。收入、毛利率和烧钱速度为空,因为没有公开财务披露。员工数来自 Defense News 在 Series B 公告时引用的公司数据。累计融资为估计值,由 Series B 前 $42M(TechCrunch)和 $200M Series B 相加得出。
[CO001, CO002, CO006, CO014, CO015, CO018]Harmattan AI 的 Series B 后快照显示资本和合同验证很强,但没有公开财务指标;产能节奏和董事会治理仍是证据缺口。
累计融资根据公开分别披露的 Series B 前融资(TechCrunch 称约 $42M)和 Series B(新闻稿称 $200M)估算;没有单一来源确认合计数。
[CO015, CO033, CO034, CO036, CO037, CO042]1.2 创始人、管理层与治理
Harmattan AI 由四位核心成员共同创立。Mouad M'Ghari(CEO)毕业于 École Polytechnique 和 MIT;他是主要公开发声人,出现在所有重要新闻稿、Dassault Aviation 公告以及 2026 年 2 月 Le Grand Continent 长访谈中。Martin de Gourcuff(CTO)与 M'Ghari 共同创办公司,并在 Series B 时发布了一条值得注意的 LinkedIn 声明,反映公司的地缘政治使命叙事。Edouard Rosset 在 Tech.eu、DefenceJobs.org 等多个第三方资料中被列为联合创始人。Marc Grelet 在 DefenceJobs.org 公司资料中被识别为联合创始人。创始四人之外,Defense News 2026 年 1 月报道称公司已从 Safran、Isar Aerospace 等成熟防务公司招募高管,扩充了高层人才基础。 截至 2026 年 1 月,公司员工超过 130 人;Defense News 称其中位经验为 15 年——若数据准确,说明公司有意招聘资深防务和工程人才,而不是以初级人员为主。开放岗位页列出 Chief Engineer、System Architect、Group Leads 和 Electronic Warfare Team Lead 等高级岗位,与一个正在成长的中期工程组织相符。 治理仍是尽调缺口。没有公开披露说明董事会构成、投资人董事席位权利或治理机制。公司未作为受监管发行人公开提交文件。四位联合创始人似乎掌握高管决策,且未公开声明外部董事会监督架构;这使关键人物集中风险落在 CEO M'Ghari 和 CTO de Gourcuff 身上。Harmattan 的隐私政策和使用条款确认,所有争议适用法国法,商业争议通过 CMAP 仲裁,消费者可诉诸法院;但公司未公开单独的治理章程或公司章程。[CO008, CO009, CO010, CO011, CO012, CO033]
| 人物 | 职务 | 背景 / 资历 | 创始人-市场匹配 | 关键人物依赖 |
|---|---|---|---|---|
| Mouad M'Ghari | CEO 兼联合创始人 | 毕业于 École Polytechnique 和 MIT;主要公开发言人 | 在法国防务生态里有深厚技术和政策可信度 | 关键 — 所有外部沟通中唯一具名 CEO,也是公司门面 |
| Martin de Gourcuff(个人) | CTO 兼联合创始人 | 公司新闻稿中具名 CTO;Series B 时有 LinkedIn 露出 | 负责 AI 与系统架构栈的技术领导 | 高 — 垂直整合硬件 / AI 公司里的 CTO 岗位 |
| Edouard Rosset | 联合创始人 | Tech.eu 和 DefenceJobs.org 档案中提及;职能角色未披露 | 创始团队成员,可能负责运营或商业化(未确认) | 中 — 联合创始人身份已确认;具体运营角色不清楚 |
| Marc Grelet | 联合创始人 | DefenceJobs.org 公司档案中列出;背景未公开披露 | 创始团队覆盖(角色未披露) | 低-中 — 列为联合创始人;公开资料有限 |
| 资深招聘(未具名) | 来自 Safran 和 Isar Aerospace 的 VP / Director 级高管 | 据 Defense News Jan 2026,来自成熟防务主承包商 | 将创始团队之外的资深防务行业经验拓宽 | 中 — 人才厚度存在,但个人姓名未公开披露 |
部分列举:新闻稿中提到的联合创始人;董事会构成、投资人董事和完整高管团队未公开披露。Rosset 与 Grelet 的背景仅限于第三方档案中的联合创始人归因。Safran 与 Isar Aerospace 招聘人员在来源中未具名;列入是为了体现团队深度,并非已识别个人。
[CO008, CO009, CO010, CO011, CO035]身份、产品栈、资本结构和关键依赖共同构成一个垂直整合模型:每一层都按主权 AI 原则内部自研。
[CO001, CO003, CO004, CO015, CO021, CO027]1.3 融资历史、投资方与估值
Harmattan AI 的融资节奏异常压缩。公司先完成一轮由 Atlantic VC 领投、金额未披露的种子轮,随后完成约 $30M Series A,由 FirstMark Capital 领投,Atlantic VC 和 Tholus Capital 参投。TechCrunch 报道 Series B 前累计融资约 $42M,与 $12M 种子轮加 $30M Series A 相吻合;Motier Ventures 和 Sisyphus Ventures 也被列为早期支持者。DefenceJobs.org 资料独立列出 Dassault Aviation、FirstMark Capital、Atlantic VC、Motier Ventures 和 Tholus Capital 为 Harmattan 投资方,印证了多投资方财团结构。 2026 年 1 月 12 日,Dassault Aviation 领投 $200M Series B,将投后估值推至 $1.4B——这一数字由 Defense News、TechCrunch、Tech.eu 和 AirForce-Technology.com 报道,并在 Harmattan 自家新闻稿中确认。本轮融资与面向 Dassault 的战略合作一同宣布,Motier Ventures 继续跟投。公司称 Series B 估值较 Series A 「显著提高」,但 Series A 投后估值未公开披露。公告时,Dassault Aviation 单独投资份额的确切规模和其他 Series B 共同投资方身份均未披露。 法国总统 Emmanuel Macron 在 X 上称赞该公告,称 Harmattan 是「法国最新独角兽」,并把交易定义为「对我们的战略自主权来说是极好消息」。TechCrunch 注意到定位变化:公司此前自称「欧洲 Anduril」,Dassault 合作后则重塑为「防务科技公司」。截至 2026 年 1 月,累计融资估计约 $242M(Series B 前 $42M 加 $200M Series B),但这是基于公开报道的估计,而非经核验的股权结构表。[CO014, CO015, CO016, CO017, CO018, CO019]
| 利益相关方 | 角色 | 控制权 / 经济重要性 | 尽调问题 |
|---|---|---|---|
| Dassault Aviation | 领投方(Series B)与战略伙伴 | 领投 $200M Series B;围绕 Rafale F5 AI 和 UCAS 建立战略合作;带来航空航天系统架构专长和国际业务开发网络 | 确认投资份额、董事会席位 / 治理权、UCAS 工作是否有排他条款 |
| FirstMark Capital | Series A 领投方 | 领投约 ~$30M Series A;早期机构支持者,推定有董事会存在 | 确认董事会席位、按比例跟投权,以及 Series B 稀释后的当前持股 |
| Atlantic VC | 种子轮领投方、Series A 参与方 | 种子轮领投;继续参与 Series A;早期高信念支持者 | 确认 Series B 后持股及任何治理角色 |
| Motier Ventures | Series A 与 Series B 投资方 | 在 Series B 中追加投资(据 Defense News 引用的 LinkedIn 帖子);连续两轮出现 | 确认追加投资份额和当前经济权益 |
| Tholus Capital | Series A 参与方 | DefenceJobs.org 档案中列为 Series A 支持者 | 确认投资金额,以及是否参与 Series B |
| Sisyphus Ventures | Series A 参与方 | TechCrunch 列为早期支持者之一 | 确认投资金额和当前治理权 |
| 法国国防部(DGA) | 客户(正式列装项目 | 订购 1,000 架 SORONA 无人机;带来经常性收入潜力和主权背书 | 确认交付完成情况、复购管线和合同续约条款 |
| 英国国防部 | 客户(正式列装项目 | 为紧急作战需求订购最多 3,000 套系统 | 确认交付进度、单位经济模型和延期选项 |
部分列举:投资人来自公司新闻稿、TechCrunch、Defense News 和 DefenceJobs.org。所有投资人的投资份额、董事会席位和治理权都未披露。国防部客户行按尽调框架列入高重要性经济利益相关方。Dassault 在 Series B 中的单独投资份额未披露。
[CO014, CO015, CO016, CO017, CO019, CO021]1.4 正式列装项目、合作伙伴与规模里程碑
Harmattan AI 的商业执行记录,是其相对同融资阶段同行最突出的差异点。2025 年 7 月——成立约 15 个月后——公司拿下首个正式列装项目:来自一个 NATO 政府(法国)的数百万美元 AI 小型无人机合同。Army Recognition 将该合同描述为向法国武装部队交付 1,000 架 SORONA 四旋翼无人机,计划 2025 年底前交付,用于支持 Orion 2026 多国演习前的作战训练。该无人机重量低于 1.8 kg,续航 40 分钟,航程 2 km,搭载 EO 和可选 IR 摄像头(红外传感器由法国公司 Lynred 供应)。 2025 年 9 月,英国国防部授予 Harmattan 第二个正式列装项目,交付多达 3,000 套自主系统,用于「紧急作战需求」——这意味着公司成立不到 18 个月就拿下两个国家国防部正式列装项目。CEO M'Ghari 在 2026 年 2 月 Le Grand Continent 访谈中确认,法国国防部订单在成立后 14 个月内完成,英国国防部订单在 15 个月内完成,未具名爱沙尼亚合作伙伴订单在 7 个月内完成。2025 年 10 月 2 日,公司宣布与乌克兰 UAS 制造商 Skyeton 达成战略合作,将 Harmattan 的 Sahara SAR 载荷和嵌入式 AI 集成到 Raybird 平台(350,000+ 战斗飞行小时),目标是法国及更广泛 NATO 市场。 Dassault Aviation Series B 合作是封顶式里程碑,把 Harmattan 技术正式连接到 Rafale F5 标准(目标约 2030 年)和未来 UCAS 项目。公司参加了 UMEX 2026(阿布扎比,2026 年 1 月)、World Defense Show 2026(利雅得,2026 年 2 月)、SOF Week 2026(坦帕,2026 年 5 月),并将参加 Eurosatory 2026(巴黎,2026 年 6 月),显示其正积极开拓国际市场。公司声称产能为「每月数千套系统」,但该数字未经独立核验;DroneXL 分析还提到最高每月 10,000 架无人机的计划,同样未经核验。CEO M'Ghari 在 Le Grand Continent 访谈中强调,无刷电机磁体和电池化学体系的供应链主权仍是实质挑战,电池对中国稀土的依赖预计还会持续 3–5 年。[CO021, CO022, CO023, CO024, CO025, CO026]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2024-04 | 公司在巴黎成立 | 创立 | n/a | Mouad M'Ghari、Martin de Gourcuff、Edouard Rosset、Marc Grelet 等创始人 | 垂直整合防务科技建设启动;总部在巴黎 |
| 2024-H2 (est.) | 种子轮完成 | 融资 | 未披露(属于 Series B 前 ~$42M 总融资的一部分) | Atlantic VC(领投);其他早期天使未披露 | 获得产品开发和首批招聘的初始资本 |
| 2025-03 (est.) | Series A 完成 | 融资 | ~$30M | FirstMark Capital(领投);Atlantic VC;Tholus Capital;Motier Ventures;Sisyphus Ventures | 扩展到可外场部署产品;支持法国国防部预资格工作 |
| 2025-07 | 首个正式列装项目授予(法国) | 规模化 | 数百万 USD(具体金额未披露) | 法国武装部队 / DGA;Harmattan AI | 公司成立 15 个月后验证产品成熟度;SORONA 订单 1,000 架无人机 |
| 2025-09 | 第二个正式列装项目授予(英国国防部) | 规模化 | 数百万 GBP(具体金额未披露) | 英国国防部;Harmattan AI | 法国以外首个 NATO 盟友;最多 3,000 套自主系统用于紧急作战 |
| 2025-10-02 | Skyeton 战略合作宣布 | 合作 | n/a | Harmattan AI;Skyeton(乌克兰) | Sahara SAR 载荷与 Raybird UAS 集成;NATO 市场扩张打法 |
| 2026-01-12 | $200M Series B 完成;Dassault Aviation 合作 | 融资 | 融资 $200M;投后估值 $1.4B | Dassault Aviation(领投);Motier Ventures(追加);其他共同投资方未披露 | 法国首家防务独角兽;Rafale F5 / UCAS AI 开发授权 |
| 2026-01-20 | 参加 UMEX 2026(Abu Dhabi) | 规模化 | n/a | Harmattan AI;UAE 防务利益相关方 | 成为独角兽后首次参加中东军贸展;释放 MENA 市场开发信号 |
| 2026-02-08 | World Defense Show 2026(Riyadh) | 规模化 | n/a | Harmattan AI;KSA 防务界 | 展示公司在欧洲正式列装项目之外对海湾地区的投入 |
| 2026-05-18 | SOF Week 2026(Tampa, FL) | 规模化 | n/a | Harmattan AI;US SOCOM;盟军特种作战部队 | 释放进入美国市场和发展 SOCOM 潜在管线的信号 |
| 2026-06-15 | Eurosatory 2026(Paris Nord Villepinte)— 已安排 | 规模化 | n/a | Harmattan AI;全球防务主管部门 | 旗舰欧洲陆防展首秀;在法国利益相关方面前获得主场曝光 |
部分列举 2024 年 4 月至 2026 年 6 月已公开宣布的里程碑。种子轮和 Series A 日期根据 Series B 时披露的累计融资估计。内部 R&D 里程碑、DGA 预资格试验和未披露合作活动未反映。国防部正式列装项目合同金额未公开披露。
[CO002, CO014, CO016, CO024, CO025, CO027]Harmattan AI 的公开记录从 2024 年 4 月创立延伸到 2026 年 1 月成为法国首家防务独角兽;18 个月内拿下两项国家级国防部记录项目,是这条线的核心锚点。
种子轮和 Series A 日期根据 Series B 前累计融资总额估算。内部里程碑和未披露政府互动未反映。
[CO002, CO014, CO016, CO021, CO024, CO025]1.5 展示材料
02市场分析
2.1 市场边界、纳入和排除支出、采购语境
界定 Harmattan AI 的市场,首先要从采购而非技术出发。公司的四个任务支柱——VSHORAD(反无人机动能拦截)、持久 ISR(SAR 载荷监视)、精确打击(自主巡飞打击),以及部队战备(自主训练)——直接对应欧洲国防部和 NATO 项目线,而不是商用或民用无人机市场。因此,纳入支出包括:(1)NATO 成员国国防部用于 C-UAS 系统、自主 ISR 平台和打击无人机采购的预算;(2)Rapid Adoption Action Plan(RAAP)和 NATO Innovation Fund 下的 NATO 集体采购和互操作性项目;(3)European Defence Fund 共同出资的无人系统和 C-UAS 能力开发;(4)与 PESCO 对齐、面向无人机和反无人机互操作性的联合项目。排除支出包括:面向农业、巡检、物流或消费用途销售的民用或商用无人机硬件;民航机构在机场周界部署、但不使用国防部预算的非军事 C-UAS;以及支持但不构成无人机任务系统的通用 GNSS 或通信基础设施。 现状替代方案零散且昂贵。欧洲武装力量目前依赖传统防空、人工操作干扰系统、人类在环热成像相机,以及各国自行采购的平台拼盘。BCG 估计,欧洲已宣布的 C-UAS 平台变体数量是北美的两倍以上(35 vs. 13),导致单机成本更高、产能爬坡更慢,跨境互操作性有限。NATO 的响应——TIE 26 演习(2026 年 5 月,荷兰)和 Latvia Innovation Range TEVV 活动(2026 年 3 月)——显示机构采购正从临时性的国家采购,转向以互操作性测试作为商业系统能否获得联盟范围合同的门槛。这就是 Harmattan AI 的进入走廊:在 TIE 26、LCI-X Crucible Series 等演习中通过 NCIA 技术评估的公司,才有资格进入 Baltic Trust 26 作战演习和后续 NATO 框架协议。 对规模测算最关键的相邻市场,是民用关键国家基础设施(CNI)C-UAS——机场、核电站、能源基础设施、大型活动场馆。BCG 预计到 2029 年,该市场还需要 2,000–3,000 套系统。虽然这在技术上位于纯国防部支出之外,但 EU 正在推进监管工作,允许民用场景依法部署 C-UAS,从而打开一条军民两用采购路径;防务优先供应商可以用与军用版本相同的硬件和指挥架构切入。[CM001, CM002, CM003, CM004, CM005, CM006]
| 细分 / 类别 | 纳入支出 | 排除支出 | 买方 / 付款方 | 现状替代方案 | 相关性 |
|---|---|---|---|---|---|
| 欧洲军用 C-UAS | 国防部为军事基地、海军舰艇和陆军采购无人机探测、跟踪、识别与反制系统 | 非国防部出资的机场商用 / 民用 C-UAS;不专门针对无人机威胁的通用电子战 | 各国国防部采购办公室;NATO NSPA 框架买方;付款方是各国国防预算 | 传统防空雷达;人工操作干扰;临时短程反制手段 | Harmattan AI VSHORAD(Gobi、Gobi Tempest)的核心细分;乌克兰经验与 NATO 东翼暴露风险把短期紧迫性推到最高 |
| 自主 ISR 无人机系统 | 固定翼和 VTOL 平台,搭载持续监视载荷(SAR、EO/IR),由军队采购用于情报收集和目标指示 | 用于测绘、农业或巡检的民用 UAV;有人 ISR 飞机 | 特种作战和情报采购办公室;空军采购;付款方是国家防务或情报预算 | 有人侦察机;商业采购的 DJI 或 Parrot 平台(许多 NATO 国家受限);卫星图像订阅 | Harmattan AI Sahara SAR 载荷和 ISR 平台在这一线竞争;Eurodrone 缺口与全天候持续监视需求推动增长 |
| 可消耗打击与巡飞弹无人机 | 为规模化生产部署和作战消耗而批量采购的自主打击系统(不包括单件 MALE 平台) | 精确制导导弹和传统巡航导弹;有人打击航空;不可消耗的 MALE 平台 | 陆军、特种作战和联合火力采购;付款方是国防部资本和运营预算 | 传统精确制导弹药;既有巡飞弹供应商(以色列、美国) | Harmattan AI Barkhan 精确打击 UAS 在这里竞争;乌克兰 FPV 采用证明了向可消耗大规模系统迁移,该细分受益 |
| 部队战备与自主训练系统 | 仿真、实兵演练和 ISR 训练平台,让部队为机器人化作战做准备 | 自主系统范围之外的通用军用飞行训练;没有无人机集成的传统模拟器 | 训练司令部和作战战备部门;付款方是运营预算而非资本采购 | 传统飞行模拟器;商用无人机实飞训练 | Harmattan AI Sonora 训练平台在这里竞争;法国宣布 1,000 架 AI 无人机训练合同,反映需求已经活跃 |
| 民用 C-UAS(CNI 保护) | 由基础设施运营商或政府安全预算出资、用于机场、核电站、能源节点和大型公共活动的 C-UAS 系统 | 国防部出资的军用 C-UAS;不具备无人机反制能力的商业安保服务 | 机场运营商、能源公用事业、大型活动安保主管机关;付款方是基础设施资本预算或政府安全拨款 | 人工安保;没有反制能力的基础地理围栏;周界围栏 | 与军用市场相邻,BCG 预计到 2029 年还会增加 2,000–3,000 套系统;支持合法部署的监管框架仍在成熟 |
| 通用商用 / 消费级无人机市场 | 消费级摄影无人机、农业精准喷洒系统、物流配送试点 | 上述所有军用、安防和 C-UAS 采购 | 消费零售买家;农业运营商;物流公司 | 无特定替代方案 | 不在 Harmattan AI 范围内;IMARC 和类似面向消费者的报告覆盖这一细分,不能与防务采购 TAM 相加 |
边界逻辑按采购预算划分:纳入支出必须来自国防部、NATO 项目、EU 防务基金或安全主管机关拨款;排除支出是民用商业或消费用途。CNI 行与核心市场相邻且在增长,但仍受监管闸门约束。
[CM001, CM002, CM003, CM004, CM005, CM006]NATO 制度化的“测试到采用”管线形成结构化采购漏斗;公司每通过一个阶段,买方风险就降低一步,也更接近框架合同。
漏斗数值是序数权重,表示资质管线相对收窄程度,并非绝对公司数量或转化率。NCIA 的实际通过率数据并未公开。
[CM007, CM008, CM009, CM037, CM038, CM039]2.2 TAM、SAM、SOM——相互竞争的视角与保留的矛盾
多个市场测算机构覆盖的边界重叠却不一致;正确输出不该是一个混合后的单一数字,而是一组带注释的视角。矛盾真实存在,并且有分析意义。 Grand View Research 将欧洲无人机市场估为 2024 年 USD 17.05 billion,按 13.4% CAGR 增至 2030 年 USD 36.34 billion。Expert Market Research 估计欧洲 UAV 2025 年为 USD 9.91 billion,到 2035 年达 USD 18.29 billion,CAGR 为 6.32%。MarketsandMarkets 预计 2025 年 USD 5.00 billion,到 2030 年 USD 7.98 billion,CAGR 9.8%。IMARC Group 只给出 2025 年 USD 275.6 million,到 2034 年 USD 600.2 million,CAGR 9.03%。同一基准年差异超过 50 倍,因为 Grand View 和 Expert Market Research 把防务硬件纳入主导类别,而 IMARC 只捕捉较小的商用 / 消费细分。这个差距不是噪音;它反映的是:把欧洲无人机定义为防务采购市场,还是商业设备市场。 更贴近 Harmattan AI 采购现实的是防务专门切片。Market Research Future 将 2024 年欧洲军用无人机收入单列为 USD 2.15 billion,到 2035 年增至 USD 4.65 billion,CAGR 7.26%;德国是最大国家市场,英国是增长最快的细分。BCG C-UAS 白皮书(2026 年 5 月)给出最可执行的采购视角:欧洲国防部已承诺 C-UAS 支出从 2025 年 USD 600 million 增至 2029 年 USD 800 million,但若计入未分配的防务项目和 CNI 保护,总可寻址 C-UAS 需求预计到 2029 年达到 USD 5–7 billion,年增长率 58%+。BCG 这条轨迹由系统数量经济学驱动——欧洲需要 4,000–5,000 套系统——而不是总收入外推。 在全球层面,StartUs Insights 将军用无人机市场定为 2030 年 USD 22.81 billion(CAGR 7.6%)和 2033 年 USD 98.24 billion(自 2026 年起 CAGR 8.9%)。这些数字大体符合美国、欧洲和亚太采购轨迹之和,但会抹平有意义的区域和细分差异。仅美国 FY2026 国防预算就申请了 USD 9.4 billion 用于无人和远程操作飞行器、USD 3.1 billion 用于 C-UAS,显示自主与反自主通常作为配对预算项同步放大。 Harmattan AI 的 SAM 位于三者交汇处:(a)欧洲国防部和 NATO C-UAS 采购,(b)欧洲自主 ISR 无人机项目,(c)可消耗型打击无人机订单。仅 C-UAS 一项,BCG 自下而上的系统数量估计就把 2029 年机会锚定在数十亿美元级别,还未计入 ISR 和打击相邻市场。由于 Harmattan AI 未披露管线或中标率,SOM 无法用公开数据量化。[CM010, CM011, CM012, CM013, CM014, CM015]
| 发布方 | 年份 | 地区 | 数值 | CAGR | 方法 | 可信度 | 局限 |
|---|---|---|---|---|---|---|---|
| BCG(2026 年 5 月) | 2025 | 欧洲(仅 C-UAS) | 国防部已承诺 $600M | 到 2029 年年化 58%+(总需求) | 自下而上:系统数量 × 单价;国防部已承诺预算 + 未分配预算 + CNI 上行空间分层计算 | 高 | 咨询公司估算;方法论已披露,但底层数据表在付费墙后;与采购公告最一致。 |
| BCG(2026 年 5 月) | 2029 | 欧洲(C-UAS,防务 + CNI) | $5–7B 总需求 | 年化 58%+ | 自下而上:系统数量(4,000–5,000 套)× 价格层级;需求远高于已宣布项目 | 中 | 预测远超已承诺预算;CNI 上行空间仍具猜测性,取决于监管是否完成。 |
| Market Research Future 研究机构 | 2024 | 欧洲(仅军用无人机) | $2.15B | 7.26% (2025–2035) | 自研市场模型;最后更新于 2026 年 4 月;仅军用平台 | 中 | 未单独纳入 C-UAS;CAGR 低于 BCG 的 C-UAS 轨迹,因为仅平台口径排除了软件和服务。 |
| Market Research Future 研究机构 | 2035 | 欧洲(仅军用无人机) | $4.65B | 7.26% | 按相同 CAGR 外推至期末 | 低 | 长期外推;2026 年 4 月版本的模型未反映 2026 年 Q1 再武装提速。 |
| MarketsandMarkets | 2025 | 欧洲(全部 UAV) | $5.00B | 9.8% (2025–2030) | 二级研究 + 专家访谈;全部应用 | 中 | 口径扩至商用 UAV;公开摘录未拆出军用专项占比。 |
| MarketsandMarkets | 2030 | 欧洲(全部 UAV) | $7.98B | 9.8% | 沿用上方模型 | 中 | 同样的范围限制;CAGR 低于 BCG 的 C-UAS 轨迹,因为商用业务拖低增速。 |
| Expert Market Research 研究机构 | 2025 | 欧洲(全部 UAV) | $9.91B | 6.32% (2026–2035) | 自研模型;包含军用和商用;纳入 2026 年 Q1 更新 | 中 | 与 MarketsandMarkets 同年数字冲突($5.00B vs. $9.91B);很可能是俄罗斯是否纳入或防务系统边界定义不同。 |
| Grand View Research | 2024 | 欧洲(全部无人机) | $17.05B | 13.4% (2025–2030) | 收入模型;硬件占主导;防务是主要细分 | 中 | 发布口径最宽;在同一地区约为 IMARC 的 50×,说明 IMARC 可能只捕捉消费细分。 |
| IMARC Group | 2025 | 欧洲(无人机) | $275.6M | 9.03% (2026–2034) | 聚焦商用 / 消费细分的市场模型 | 中 | 比 Grand View 和 MarketsandMarkets 低一个数量级;几乎肯定排除了军用硬件;不适合作为防务市场规模输入。 |
| StartUs Insights | 2030 | 全球(军用无人机) | $22.81B | 7.6%(自 2026 年起) | 创业公司生态和采购数据扫描;全球 | 低 | 区域颗粒度被折叠;按支出计算,欧洲大约占全球市场 20–25%。 |
| StartUs Insights | 2033 | 全球(军用无人机) | $98.24B | 8.9% (2026–2033) | 同一模型外推;全球口径 | 低 | 长期外推;七年预测误差带很宽;仅用于展示全球规模背景。 |
各行不能相加,也不能求平均。不同发布方采用不同边界(仅 C-UAS、军用无人机、全部 UAV、含消费级的全部无人机);直接比较前必须按范围归一化。BCG 是 C-UAS 的主要规模锚点;MRFR 用于欧洲军用无人机;MarketsandMarkets 用于更宽的欧洲 UAV 口径。可信度评级反映方法透明度,以及与 Harmattan AI 采购市场的匹配度。
[CM001, CM002, CM010, CM011, CM012, CM013]从宽口径欧洲 UAV 到防务采购 SAM 和 C-UAS 子细分的嵌套市场层级,构成 Harmattan AI 短期可直接切入的需求边界。
各发布方估计不可相加,且口径不兼容;金字塔层级是概念嵌套,不是算术拆分。BCG 数字是 C-UAS 规模测算最贴近采购的锚点。
[CM001, CM002, CM013, CM014, CM010]不同发布方估计差出一个数量级,源于口径差异而非预测错误;真正相关的防务采购锚点,是 BCG 的 C-UAS 系统数量轨迹和 MRFR 的军用无人机细分。
BCG 中值按 2029 年 $5B 与 $7B 两端平均计算;此处显示为期末年化值,不是 2025 年数字。BCG 的低 / 高列分别代表 2025 年已承诺国防部基数(低)和 2029 年高情景(高)。其他行显示基准年至预测年的区间;单位为 US$B 年市场收入。
[CM001, CM002, CM010, CM011, CM012, CM013]2.3 买方、用户、付款方与 NATO/欧洲防务采购路径
欧洲防务无人机和 C-UAS 采购的买方结构具有机构化、多阶段特征。在国家层面,国防部采购办公室是正式购买实体;作战司令部(陆军、空军、特种作战部队)提出需求并影响采购,NATO 能力目标和互操作标准则限制各国规格自由度。付款方通常是国家国防预算,并由 European Defence Fund 和 PESCO 联合项目提供 EU 共同出资补充。用户是一线操作者——旅级指挥官、特种作战单位、海军舰员或边境部队——他们在 TIE 26、LCI-X Crucible Series 等演习中的反馈会直接影响采购决策。 各细分市场采购周期不同。C-UAS 系统处于快速采购模式:乌克兰经验、NATO 波罗的海侧翼暴露,以及 EU 机场附近无人机事件叠加,形成了立即采购的政治紧迫性。各国政府正在正常多年资本项目之外批准紧急采购;德国批准 Helsing 和 Stark 的 EUR 536 million 采购就是这种快速购买动态的例子。相比之下,持久 ISR 和大型 MALE 无人机平台仍处于更长开发周期,Eurodrone(法国、德国、西班牙、意大利参与)只是因 2026 年 Q1 冲突经验才加速。 三个买方细分直接对应 Harmattan AI 产品组合。第一,陆军部队——旅、营和快速反应单位——是 VSHORAD(Gobi、Gobi Tempest)和便携式 C-UAS 的主要买方。预算权在各国国防部陆军采购办公室;采用触发点是战场紧迫性和 NATO 目标合规。第二,特种作战部队和情报机构购买具备 SAR 能力的紧凑 ISR 平台(Sahara 载荷);采购通过保密专项项目推进,周期更短,对单价容忍度更高。第三,NATO 联合和多国演习框架——包括 Latvia Innovation Range TEVV 和 LCI-X Crucible Series——相当于准采购事件,通过技术和作战评估的系统,会进入框架协议资格。对 Harmattan AI 这样的规模化阶段初创而言,NATO 测试活动本质上是客户资格路径,而不只是演示。 民用基础设施保护正在形成第四类买方。EU 监管框架逐步成熟,使机场、核电站和能源节点能够依法部署 C-UAS;Harmattan AI 供给军方的同一套硬件,也可能通过该渠道触达公用事业运营商和机场管理机构。该渠道仍处早期:BCG 预计 2029 年前需要 2,000–3,000 套 CNI C-UAS 系统,但也指出监管框架尚未最终落定。JRC(European Commission)已经发布关键基础设施保护的五阶段 C-UAS 方法论,显示防务、联邦警察、地方安保和空域管理方之间的利益相关者整合很复杂。[CM003, CM004, CM005, CM008, CM009, CM018]
| 细分 | 买方 | 使用者 | 付款方 | 采购流程 | 预算所有者 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 陆军 C-UAS | 各国国防部陆军采购局 | 旅 / 营级指挥官;防空排 | 国家防务资本预算 | 需求定义 → NATO 演习验证(TIE、LCI-X)→ 国家招标或框架下单 → 交付 | 陆军参谋长;国防部采购秘书处 | 乌克兰战场无人机造成的伤亡;NATO 能力目标达标;波罗的海侧翼暴露风险 |
| 特种作战 ISR | 特种作战司令部;情报采购办公室 | SOF 小队;ISR 操作员 | 机密专项预算 | 通过专项直接采购;周期短于标准采购 | SOF 指挥官;情报主管 | 全天候持续监视需求;城市侦察;GPS 受限环境作业 |
| 海军 / 舰载 C-UAS | 海军采购局 | 舰长;海军防空艇员 | 海军资本装备预算 | 平台采购周期长;截至 2026 年中公开项目很少 | 海军参谋长;国防部海军分支 | 巡飞弹和无人机对舰艇构成威胁;海上设施保护 |
| NATO 集体 / 多边 | NCIA;NSPA;NATO ACT;多国项目办公室 | 跨国盟军作战单位 | NATO 共同出资预算;联盟成员国出资 | 技术评估(TIE)→ 作战演习(Baltic Trust)→ 框架协议 → 各国框架下单 | NATO SG / SACEUR;各国代表团 | 联盟互操作要求;TIE 26 和 LCI-X 评估闸门 |
| 民用基础设施(机场、核电、能源) | 机场运营商安保;核设施主管机关;能源公用事业 | 安保经理;现场警卫 | 基础设施运营商资本预算或政府安全补助 | 合法部署 C-UAS 的监管批准 → 采购招标 | 场址安保总监;基础设施 CFO | EU 监管框架成熟;无人机闯入事件;责任风险 |
| 部队战备 / 训练 | 国家训练司令部;作战战备部门 | 自主系统操作员、飞行员和维护人员 | 运营预算;训练拨款 | 训练项目定义 → 平台采购 → 部署到训练靶场 | 训练司令部主管;国防部作战战备预算负责人 | 快速熟悉自主系统;向机器人优先部队结构转型 |
采购流程根据 NATO 演习文件、BCG 分析和 EP 政策文件重构。具体合同时间线和预算归属会因国家和项目密级而异。
[CM003, CM004, CM005, CM007, CM008, CM009]防务无人机采购由机构买方多阶段推进;NATO 演习资质路径在国家追加订单前,先起到采纳漏斗作用。
采购周期估计基于 NATO 文件和 BCG 分析;各国和保密级别不同,实际时间线会变化。海军和 CNI 行反映 Harmattan AI 参与的公开证据有限。
[CM003, CM004, CM005, CM007, CM008, CM009]2.4 增长驱动、采用约束与结构性逆风
首要需求驱动是地缘政治紧迫性,具体来自俄罗斯对 NATO 东翼国家的混合战争威胁,以及乌克兰冲突暴露出的持续无人机威胁。European Parliament 2026 年 1 月关于无人机和新型战争系统的主动倡议报告明确要求,在 EU 防务规划各层级快速整合无人机和反无人机能力,并建设强健、自主的欧洲无人机产业。European Parliament Think Tank 数据确认,到 2024 年 5 月,无人机已超过所有其他武器,成为乌克兰冲突中的首要伤亡原因。这为紧急采购支出提供了政治正当性,否则这类支出会面临漫长预算审批周期。 第二个驱动是 NATO 标准化和测试基础设施。Rapid Adoption Action Plan(RAAP)下设立五个 NATO Innovation Ranges——包括 Latvia UAS/C-UAS 靶场和 Cyprus/Estonia 网络靶场——为经过测试、具备互操作性的系统创造了机构需求。TIE 26 演习框架(40 家公司、300+ 参与者、60+ 系统测试)起到认证事件作用:展示技术和作战互操作性的系统,才有资格进入 NATO 作战演习,并随后进入联盟框架合同。这个结构化采用管线,降低了有能力跨过技术门槛的公司的商业化阻力。 欧洲主权再武装支出是第三股顺风。欧洲航空航天和防务行业 2024 年营业额达 EUR 325.7 billion,研发投入 EUR 25.2 billion;Eurostat 指出 2022 年防务支出为 GDP 的 1.3%,且存在上行压力。EU 到 2030 年的防务准备路线图包含欧洲无人机防御倡议,目标是部署具备互操作性的 C-UAS。仅德国就批准了 EUR 536 million 无人机供应合同(Helsing 和 Stark)。SIPRI arms-transfer database(2026 年 3 月更新 1950–2025 数据)提供了长期背景:欧洲国家历史上一直是主要常规武器净进口方,但现在具备建设本土供应链的结构性动机。 主要采用约束来自产业碎片化、采购周期长度和技术转移风险。BCG 清楚指出重复建设问题:欧洲已宣布 35 种 C-UAS 平台变体,北美为 13 种,造成囚徒困境——每个国家项目都太小,无法达到制造规模,但政治压力又让整合进展缓慢。这意味着 Harmattan AI 这类欧洲初创面临悖论:总需求很大,但单个合同金额仍可能中等,因为没有任何单一国防部能承诺足以支撑高效生产的数量,多国标准化也仍在形成。BCG 经验曲线分析估计,每次产量翻倍可带来 10–20% 单位成本下降,但欧洲订单碎片化阻止多数供应商沿曲线下行。 技术信任和保密约束是第二个逆风。军方 C-UAS 采购办公室在授予生产合同前,需要大量评估、国家安全审查,有时还要求本土制造。欧洲偏好本土供应商,为 Harmattan AI 这样总部在法国、背靠 Dassault Aviation 的公司创造机会;但也意味着 Anduril 的 USD 642 million USMC 正式列装项目等美国平台,可能设定能力标杆,却因主权要求不易被欧洲国防部采购。EP 报告明确警告不要依赖非 EU 供应商,并主张建设强健、自给自足的欧洲无人机产业。这既是竞争护城河,也是市场约束:它抬高 Harmattan AI 的欧洲战略价值,同时放慢任何需要长期产业基础评估的采购决策。 最后,技术快速迭代既是机会也是约束。Shahed 无人机在不到两年里完成重大升级——新战斗部、诱饵、推进系统、AI 瞄准。反制系统必须以同样速度迭代。这有利于软件优先的自主平台,而不是纯硬件传统厂商,正好符合 Harmattan AI 的定位;但它也带来持续高研发强度,并压缩任何单一采购合同的生命周期价值。[CM006, CM019, CM020, CM021, CM022, CM023]
| 驱动因素 / 约束 | 方向 | 时点 | 含义 | 尽调问题 |
|---|---|---|---|---|
| 乌克兰战争无人机经验与 NATO 东翼威胁 | 利好 | 当前 | 把政治紧迫性转成常规预算周期之外的紧急采购;德国 EUR 536M 合同可作为其他国家模板 | 确认 Harmattan AI 是否进入任何由乌克兰经验采购轮次催生的活跃国防部招标流程 |
| NATO 制度化测试管线(TIE、LCI-X、TEVV) | 利好 | 当前 | 从创新到框架合同形成结构化资质路径;NCIA 评估实际上就是采购闸门 | 判断 Harmattan AI 系统是否已进入 TIE 26 评估,或是否排期参加拉脱维亚靶场即将开展的 TEVV 活动 |
| EU 无人机防务倡议与欧洲防务基金 | 利好 | 2026–2030 | 把可互操作 C-UAS 和自主系统的共同出资集中起来;主权供应采购对各国预算的依赖下降 | 跟踪 Harmattan AI 是否参与任何 EDF 或 PESCO 共同出资项目 |
| 主权供应偏好(非 EU 供应商风险) | 利好(对欧盟本土供应商) | 当前 | EP 报告明确把依赖非 EU 供应商列为战略风险;为法国 / 欧洲公司创造偏好采购环境 | 验证 Harmattan AI 平台的出口分类,以及是否已取得法国 DGA 批准 |
| Dassault Aviation 战略合作 | 利好 | 当前 | EUR 200M B 轮融资和 Dassault 关系带来工业可信度,并可能打开 Rafale F5 集成路径;增强国防部供应商资质 | 确认 Dassault 商业关系在财务投资之外的范围;评估 Kalahari 是否集成 MBDA 或 Thales |
| 欧洲产业碎片化(35 种平台变体 vs. 北美 13 种) | 阻力 | 当前 | 国防部订单分散,任何单一供应商都难以拉起生产规模;单位成本保持高位;各国项目之间存在预算竞争 | 要求提供平均合同规模,并判断在当前 NATO 标准化协议下是否可行多国框架订单 |
| 采购周期长且存在密级要求 | 阻力 | 当前 | 3–5 年资本采购时间线与创业公司资本需求错配;密级审查增加延迟;还需要技术风险评估 | 向 Harmattan AI 询问从首次国防部接触到签署合同的典型时间线;识别是否已有项目进入最终批准阶段 |
| 无人机威胁快速演化(对手迭代周期短) | 双向 | 当前且持续 | 迫使软件 / 硬件持续更新;有利于 AI 软件优先模式,但需要持续 R&D 投入,并可能频繁重新认证平台 | 了解 Harmattan AI 的软件更新和 OTA 能力;确认在 NCIA 标准下重新认证的成本和周期 |
| GPS 韧性与 EW 生存性要求 | 阻力 | 当前 | 2026 年 Q1 冲突数据把不依赖 GPS 的导航列为设计要求;所有无人机供应商的开发复杂度和成本上升 | 确认 Harmattan AI 平台是否通过 GNSS 干扰韧性测试,以及 Sahara SAR 是否能在 GPS 受限环境中运行 |
| 美国关税对欧洲 C-UAS 供应链的影响 | 阻力(中等) | 当前 | 2025 年美国关税影响欧洲零部件进口;美国来源子系统成本上升;MarketsandMarkets C-UAS 报告提示关税会影响欧洲定价 | 识别 Harmattan AI 供应链对受美国关税影响零部件的暴露;评估欧洲替代品是否已获资质 |
时点和方向综合 BCG、EP EPRS、NATO 和 MarketsandMarkets 来源;在最终判断前,全部都需要内部经营数据和合同状态确认。
[CM006, CM018, CM019, CM021, CM022, CM023]2.5 Harmattan AI 在既定市场中的可触达机会
Harmattan AI 站在 NATO 欧洲三大增长最快采购细分的交汇点——C-UAS、自主 ISR 和可消耗型打击;其产品架构(由 Kalahari C2 编排 Gobi VSHORAD、Sahara SAR、Barkhan 打击和 Sonora 训练)与 NATO 在 TIE 26 和 LCI-X 测试的多层互操作需求相呼应。公司注册地在法国,又有 Dassault Aviation 背书;在 EP 报告明确将依赖非 EU 无人机供应商列为战略风险的时点,这使其成为面向法国和更广泛 NATO 防务客户的欧洲主权供应商。 近期可触达切片是经过 NATO 测试的 C-UAS 和 ISR 采购管线。BCG 自下而上估计 NATO 欧洲需要 1,500–2,000 套军用 C-UAS 系统;已确认的 BCG 口径显示,2025 年国防部 C-UAS 支出承诺为 USD 600 million,并将在 2029 年增至 USD 800 million。这在计算潜在 USD 5–7 billion 总需求轨迹前,已经给出了有来源支撑的采购底线。Harmattan AI 不需要拿下很大市场份额,也能获得可观收入:到 2029 年,即便只拿下 USD 5 billion C-UAS 市场的 1%,也对应 USD 50 million 年合同额,而防务合同通常跨多年,并附带后续维护义务。 中期机会来自从测试转向规模化生产,而 NATO 创新架构的设计目的正是加速这一转换。LCI-X Crucible Series、Baltic Trust 26 和 TIE 26 周期不是一次性演示;它们是从实验走向联盟层面作战采用的、反复运行的制度化路径。通过 NCIA 技术评估的公司,会获得某种采购预资格;随着联盟范围标准化压缩合规供应商数量,这种资格价值会上升。Harmattan AI 参加 Eurosatory 2026、SOF Week 2026、World Defense Show 2026 和 UMEX 2026,显示它正面向欧洲和出口军事买家主动追逐这条资格路径。 投资尽调的关键开放问题包括:Harmattan AI 是否已经完成或进入任何 NATO Innovation Range 技术评估;2026 年具名国防部合同管线(已披露法国训练订单之外)是什么样;Kalahari C2 层是否按 NATO 互操作标准(STANAG、Link 16 兼容数据链)设计,以满足联盟级认证要求。这些具体采购闸门问题,无法仅凭公开数据中的规模测算解决。[CM001, CM002, CM003, CM004, CM007, CM009]
2.6 展示材料
03竞争格局
3.1 竞争格局概览
Harmattan AI 进入的是一个竞争激烈、资本高度充足的欧洲防务自主市场,可分为五类竞争者。直接欧洲同行——Helsing(德国)和 TEKEVER(葡萄牙、英国、法国)——在主权资质、制造规模和已验证作战部署上竞争。美国防务科技主承包商——Anduril 和 Shield AI——带来压倒性的资本规模、先进自主系统组合和已验证的 DoD 合同工具,但受 ITAR 约束,EU 主权定位有限。欧洲 ISR-sUAS 专家——主要是 Quantum Systems(德国)——瞄准同一情报收集和拒止环境利基,但不提供打击或 C2。MyDefence(丹麦)等 C-UAS 专家覆盖反无人机拦截细分,与 Harmattan 的 Gobi 和 Gobi Tempest 拦截器重叠。Delair(法国)等商用相邻无人机公司在军用 ISR 边缘竞争,但不是主要军事系统集成商。没有单一竞争者复制 Harmattan 在单一纵向一体化栈内同时集成 SAR ISR、自主打击、动能 C-UAS 和 AI 驱动 C2 的全谱野心;该栈还由一家 Tier-1 OEM 与 NATO 标准战斗机项目的合作锚定。[CP036, CP037, CP038, CP001, CP008, CP013]
| 竞品 | 类别 | 估计累计融资 | 总部 | 主要细分 | 关键差异化 | 相比 Harmattan 的关键局限 |
|---|---|---|---|---|---|---|
| Harmattan AI | 直接可比 / 欧盟防务主承包商 | $275M(€250M 估计) | 法国巴黎 | 法国 / 英国 / NATO 主权多域 | ISR + 打击 + C-UAS + C2 全组合;Dassault Rafale F5 / UCAS 集成 | 仍处早期;独立实战验证有限 |
| Helsing | 直接可比 / 欧洲打击无人机 | 累计约 $791M | 德国慕尼黑 | NATO 打击无人机;乌克兰冲突 | Resilience Factory 可扩展 HX-2 生产;Altra 多无人机平台 | 专注打击 / 自杀式无人机;无 C-UAS 硬杀伤;无 SAR;依赖 EU 资金 |
| Anduril | 美国防务主承包商 | 累计 $11B+ | 美国加州 Costa Mesa | 美国 DoD + 盟军 | Lattice OS;2025 年收入 $2.2B;多域宽度;估值 $61B | ITAR / 主权限制削弱 EU 主权叙事;以美国为中心 |
| Shield AI | 美国自主平台 | 估计累计 $2B+ | 美国加州 San Diego | 美国 DoD / 海事 / Collaborative Combat Aircraft | Hivemind AI 飞行员(26 类载具);V-BAT 海事;EdgeOS 中间件 | Reuters 报道安全事件;V-BAT 之外硬件有限 |
| TEKEVER | 相邻 / ISR 即服务 | £400M OVERMATCH 项目 | 里斯本 / 伦敦 / Cahors | 海上巡逻;边境监视;英国 / 法国国防部 | 已盈利;IaaS 订阅模式;EMSA 框架;NATO IF 支持 | 能力窄(ISR / 海事);无打击;无硬杀伤 C-UAS |
| Quantum Systems | 相邻 / ISR sUAS | 累计 €310M+ | 德国慕尼黑 | 防务 ISR + 商用军民两用 | Vector AI 经实战验证;GNSS 受限环境作业;Hensoldt + Airbus D&S 投资 | 仅 ISR;无 C2 平台;不是全谱系主承包商 |
| Delair | 相邻 / 企业无人机 | 未披露(VC 支持) | 法国图卢兹 | 企业测绘 + 防务 ISR 相邻 | 开放载荷架构;企业分析;军民两用 | 以商业为主;无自主打击或硬杀伤 C-UAS |
| MyDefence | C-UAS 专家 | 未披露 | 丹麦哥本哈根 | 军用 / 关键基础设施 C-UAS | 5,000+ 次部署;可穿戴 / 便携 Wingman 和 Pitbull;美国生产 | 仅 RF / 软杀伤;无 ISR;无打击;无 C2 层 |
累计融资来自截至 2026 年 6 月已知的最新轮次;不少数字根据新闻稿数据估算。TEKEVER 的 £400M 反映 OVERMATCH 项目承诺,不只是股权融资。Anduril 2025 年收入($2.2B)引自 CEO 在 2026 年 5 月 H 轮融资博文中的说法。Harmattan AI 累计融资根据 $200M B 轮融资(2026 年 1 月)加上推断的前序轮次估算。“未披露”表示没有公开数据,不表示没有资金支持。
[CP001, CP003, CP006, CP008, CP009, CP013]Harmattan AI 位于高主权 / 宽能力象限。Anduril 能力宽度领先,但锚定美国;Helsing 欧洲主权属性领先,但能力范围更窄;TEKEVER 主权得分最高,但定位是 ISR 专家。
坐标位置是序数估计,依据截至 2026 年 6 月的公开证据,包括资金来源、国防部关系、总部司法辖区和有文档记录的产品组合宽度。没有数字指数或第三方评分支撑这些位置。X 轴:0 = 完全锚定美国,1.0 = 完整 EU/NATO 主权。Y 轴:0 = 单一能力专家,1.0 = 全谱系多域主承包商。
[CP036, CP037, CP005, CP023, CP031]3.2 Tier-1 融资同行:Helsing、Anduril 和 Shield AI
Helsing(约 2021 年创立于德国)最初是为传统军事平台提供 AI 软件叠加层,2024 年转向打击无人机硬件。其 HF-1 与乌克兰产业伙伴共同生产;更新的 HX-2 自杀式无人机在 Helsing 位于德国南部的第一座 Resilience Factory 制造,初始月产超过 1,000 架,并声称冲突情景下可扩展到每月数万架。Helsing 截至 2025 年 6 月 €600M 轮融资累计募集约 €761.5M;据报道,2026 年 5 月还在敲定由 Dragoneer 领投、估值 $18B、规模约 $1.2B 的新一轮融资。Helsing 的 Altra 连接平台支撑多无人机协同,直接类比 Harmattan 的 Kalahari C2 层。不同于 Harmattan,Helsing 是打击和软件专家:公开资料中没有其动能 C-UAS 能力或 SAR 载荷。Anduril(2017 年创立,Costa Mesa CA)是美国占主导地位的防务科技主承包商。CEO Brian Schimpf 称其 2025 年收入翻倍至 $2.2B;公司 2026 年 5 月完成 $5B Series H,估值 $61B,使累计融资超过 $11B。Anduril 的 Lattice 平台在异构系统之间提供自主 C2 和传感器融合——包括 Roadrunner 自主拦截器、Ghost 巡飞弹药和协同作战飞机。Anduril 正在扩张欧洲业务:2026 年 5 月赢得荷兰国防部合同。不过,ITAR 和军民两用出口法规限制了欧洲国防部要求的主权技术转移;相较 Harmattan 的法国公司身份以及法国、英国正式列装项目,这是结构性劣势。Shield AI(2015 年创立,San Diego CA)于 2026 年 3 月完成 $1.5B Series G,估值 $12.7B,由 Advent International 领投、JPMorganChase 联合领投。其 Hivemind 自主软件已驾驶包括 F-16 战斗机在内的 26 类载具,并被 US Air Force 选为 Collaborative Combat Aircraft 任务自主能力提供商。V-BAT UAS 已部署在几乎所有类别的 US Navy 舰艇上,并覆盖全部七支 Marine Expeditionary Units。Shield AI 正在收购 Aechelon Technology,将高保真仿真与 Hivemind 训练集成。Shield AI 的 EdgeOS 中间件为确定性自主控制实现亚毫秒级本地延迟。Reuters 2026 年 6 月发布的调查记录了一家估值约 $13B 的美国防务科技初创的安全事件,包括无人机坠毁和工人受伤;这凸显全行业运营安全审查,可能影响包括 Harmattan 在内的自主系统公司的监管姿态。[CP001, CP002, CP003, CP004, CP005, CP006]
| 能力维度 | Harmattan AI | Helsing | Anduril | Shield AI | TEKEVER | Quantum Systems | MyDefence |
|---|---|---|---|---|---|---|---|
| 远程 ISR(光学 / EO) | ✓(Sahara SAR + Sonora UAS) | 部分覆盖(HF-1 光学;软件叠加层) | ✓(Ghost + ISR 套件) | ✓(V-BAT + ViDAR 覆盖) | ✓(AR5、AR3 长航时) | ✓(Vector AI) | ✗ |
| 全天候 ISR(SAR) | ✓(Sahara L–S 波段) | 部分(仅软件;无 SAR 硬件) | Unknown | ✗ | ✓(AR5 变体集成 SAR) | ✗ | ✗ |
| 自主打击 / 巡飞弹 | ✓(Barkhan) | ✓(HF-1、HX-2) | ✓(Ghost、Roadrunner) | ✗ | ✗ | ✗ | ✗ |
| C-UAS 动能拦截 | ✓(覆盖 Gobi Grp 1–2;Gobi Tempest Grp 2–3) | ✗ | ✓(Roadrunner 可重复使用) | ✗ | ✗ | ✗ | 部分(仅 RF 软杀伤) |
| AI C2 / 任务编排 | ✓(Kalahari) | ✓(Altra 平台) | ✓(Lattice) | ✓(Hivemind + EdgeOS) | 部分(服务层 C2) | 部分(QBase Tactical) | ✗ |
| 蜂群 / 协同饱和 | ✓(核心作战理念) | ✓(Altra + HX-2) | ✓(Lattice 多智能体) | 部分(多载具 Hivemind) | ✗ | 部分(开发中) | ✗ |
| GNSS 拒止环境自主导航 | ✓(公司称) | ✓(公司称) | ✓(公司称) | ✓(Hivemind 已演示) | 部分 | ✓(Vector AI,乌克兰实战验证) | N/A |
| Tier-1 OEM / 平台集成 | ✓(Dassault Rafale F5 / UCAS) | 部分(Mistral AI;无 OEM 飞机) | ✓(Boeing YFQ-44A CCA) | ✓(USAF CCA / Anduril YFQ-44A) | 部分(CNES/DGA SAR;英国陆军 Project NYX) | 部分(Hensoldt + Airbus D&S 投资方) | ✗ |
| 欧洲主权资质 | ✓(法国公司;法国 + 英国 PoR) | ✓(德国公司;获 EU 资助) | 部分(拿下荷兰国防部;适用美国 ITAR) | 部分(EU 办公室;适用 ITAR) | ✓(葡萄牙 / 英国 / 法国运营) | ✓(德国公司;NATO IF 投资方) | ✓(丹麦公司;NATO 成员) |
| 电子战 / SIGINT | 部分(列有 EW 产品线) | ✗ | 部分(接近干扰能力) | ✗ | 部分(与 Avantix 签署 EW/SIGINT MoU) | ✗ | 部分(C-UAS 中的 RF 干扰) |
| 制造规模(声称 >1,000 台 / 月) | ✓(公司称数千台 / 月;未验证) | ✓(HX-2 已确认 ≥1,000 台 / 月) | ✓(大型美国制造基地) | Unknown | 部分(Cahors 工厂 2026 年投产) | 部分(Series C 后扩产) | 部分(累计 5,000+;月产率未知) |
| 实战 / 运营部署证明 | 部分(已获 PoR;Skyeton 合作) | ✓(HF-1 已交付乌克兰实战) | ✓(多个美国 DoD 合同) | ✓(V-BAT 已用于美国海军 7 个 MEU) | ✓(引用 350,000+ 实战飞行小时) | ✓(Vector AI 经乌克兰实战测试) | ✓(已部署 5,000+ 套系统) |
"✓" 表示能力已被抓取到的官方产品页面或可信媒体报道确认,或与这些资料一致。 "部分" 表示能力有声称或部分文件支持,但缺少足够独立验证。"✗" = 未找到证据; "未知" = 无公开数据。基于截至 2026 年 6 月可公开获得的信息; 不是独立技术评估。
[CP004, CP009, CP011, CP014, CP015, CP017]| 公司 | 核心产品 | 定价模式 | 已知定价信号 | 合同结构 | 不透明度 | 对买方的含义 |
|---|---|---|---|---|---|---|
| Harmattan AI | ISR / 打击 / C-UAS / C2 组合 | 正式列装项目(PoR)/ 政府合同 | 未披露 | 法国和英国国防部长期项目 | 很高 | PoR 锁定预算;近期市场限于立场一致的 NATO 国防部 |
| Helsing | HX-2 / HF-1 打击无人机 | 按台生产合同 | 声称单台成本低于 AeroVironment Switchblade;未披露数字 | 政府 + 乌克兰直接采购 | 很高 | 单台经济性具备竞争力的说法;未经独立验证 |
| Anduril | Roadrunner;Lattice OS;Ghost | 混合:硬件平台 + Lattice 软件订阅 | 部分 DoD 合同金额通过采购公告披露;标价未披露 | 美国 DoD 长期平台合同 | 高 | 政府企业级模式;IRAD 投入可能补贴单台定价 |
| Shield AI | Hivemind 软件;V-BAT;X-BAT(开发中) | 软件许可 + 硬件平台 | 未披露 | 政府项目;潜在 Hivemind SaaS 授权开始出现 | 高 | AI 飞行员软件授权模式可能独立于硬件收入放大 |
| TEKEVER | AR5;AR3 UAS + IaaS | 情报即服务订阅 | €30M EMSA 框架(多年);其他合同未披露 | 多年服务合同 | 中 | 经常性收入;买方前期成本更低;契合采购服务化趋势 |
| Quantum Systems | Vector AI;Trinity Pro;Drone Port 产品组合 | 系统销售 + 支持 | 未披露 | 按套 + 维护;商业 / 政府双轨 | 中 | 商业分销扩大买方基础;招标让部分价格可见 |
| Delair | DT26X;监视变体 | 按台销售 + 数据分析 SaaS | 未披露(有商业定价信号) | 商业 / 政府招标混合 | 低 | 价格最透明的竞品;商业惯例可能形成价格上限压力 |
| MyDefence | Wingman;Pitbull;固定站点阵列 | 系统销售 + 支持合同 | 未披露;政府招标部分可见 | 按套 + 服务 | 中 | 便携 / 可穿戴设备单元经济性低于平台系统;与 Harmattan 重叠有限 |
所有定价信息均来自截至 2026 年 6 月公开可得的合同公告、新闻稿和官方产品页面。 没有竞品公布标价;公开资料无法支持一对一价格比较。TEKEVER 的 EMSA 框架(€30M) 是本组对比中唯一量化的合同金额。"不透明度" 是对公开定价透明度的定性评估, 不是竞争力指标。
[CP020, CP023, CP039]Shield AI 的 Hivemind 在已展示平台类别中的多机 AI 飞行员宽度领先;Harmattan 的 Kalahari 和 Helsing 的 Altra 是欧洲 C2 / 自主能力的对应方案;在欧洲 ISR 专家中,Quantum Systems 的 Vector AI 在经战场验证的边缘计算 ISR 自主能力上领先。
评级(无 / 低 / 中等 / 高)是基于截至 2026 年 6 月的公开产品页、新闻稿和可信新闻来源形成的定性评估。未做独立技术基准测试。“未知”表示没有公开数据。
[CP004, CP011, CP018, CP027, CP028]3.3 欧洲相邻竞争者:TEKEVER、Quantum Systems 和 Delair
TEKEVER(Lisbon, Portugal;在英国和法国有重要业务)是结构上最不同的欧洲竞争者。TEKEVER 不销售平台,而是提供 情报即服务(Intelligence-as-a-Service):以订阅或按任务计费方式交付无人机监视。该模式带来罕见组合:快速增长和盈利能力;在公司获得 NATO Innovation Fund 支持的 €70M Series B 时,领投方 Baillie Gifford 确认了这一点。TEKEVER 的 AR5 持有一份 €30M EMSA 框架合同,用于欧洲海事 UAS 监视。公司在乌克兰部署系统用于远程 ISR,并入选 Project NYX,为 UK Army Apache 直升机开发忠诚僚机旋翼平台。法国 Cahors 工业基地目标在 2026 年夏季启动生产;North Carolina 办公室于 2026 年 5 月开设,以进入美国市场。TEKEVER 在英国 OVERMATCH 投资计划下的总承诺达到 £400M。TEKEVER 的服务层收入模式意味着它不太可能在平台级国防部项目中取代 Harmattan,但其在欧洲海事和边境监视中的 ISR 即服务 主导地位,与 Harmattan Sahara SAR 载荷用例重叠。Quantum Systems(Munich,2015 年创立)于 2025 年 5 月完成 €160M Series C,由 Balderton Capital 领投,Hensoldt 和 Airbus Defence and Space 战略参投。随后 2025 年 11 月的一次融资事件据称使公司估值增至原来的三倍以上,超过 €3B。公司在德国、乌克兰、澳大利亚和罗马尼亚拥有 550 名员工,并连续多年实现约 100% 收入增长,是可信的欧洲 ISR-sUAS 竞争者。其 Vector AI 平台配备双 Jetson Orin 边缘计算板,可在 GPS/GNSS 拒止环境中自主导航;系统已在乌克兰和澳大利亚经受实战测试。不过,Quantum Systems 只是 ISR 玩家:没有与 Kalahari 相当的打击、C-UAS 动能拦截或 C2 平台能力。Delair(Toulouse, France)生产企业测绘和监视无人机。其安全导向平台支持实时视频跟踪、军用频段通信,并可在 −15°C 到 +50°C 温度范围运行。Delair 主要是一家企业数据采集公司,防务只是相邻细分;它不提供自主打击、C-UAS 动能拦截或 AI 驱动 C2 层,因此对 Harmattan AI 更像边缘竞争者,而非直接竞争者。[CP019, CP020, CP021, CP022, CP023, CP024]
3.4 C-UAS 专家、传统厂商与现状替代方案
MyDefence(Copenhagen, Denmark)是最直接相关的 C-UAS 专家。公司有 134 名员工,全球部署超过 5,000 套系统,客户组织超过 100 家,积累的部署基数很可能超过 Harmattan 公开确认的 C-UAS 足迹。公司 2026 年 3 月以 Made-in-USA 模式在 Oklahoma City 开设美国生产设施,2026 年 4 月获得 Australian Department of Defence A$9.81 million 合同,并支持 US DHS 为 FIFA World Cup 2026 保护做准备。MyDefence 的产品——可穿戴 Wingman、便携式 Pitbull 和固定站点多传感器阵列——仅属于电子战软杀伤系统。它们执行 RF 探测、干扰和扰乱,但不提供动能直接撞击拦截,不提供 ISR、打击或 C2。相对 Harmattan 的 Gobi 拦截器(Group 1–2 动能)和 Gobi Tempest(Group 2–3 动能),MyDefence 只在软杀伤层竞争。多数 NATO 军队的现状替代方案仍是 Thales、Raytheon、L3Harris 等成熟主承包商提供的传统 C-UAS 系统。这些传统厂商提供采购安全感、既有集成路径和经验证互操作性,但缺少嵌入式 AI 驱动自主能力,也没有 Harmattan 为 Gobi 声称的分钟内交战周期。资金充足的 NATO 军队理论上仍可内部自研,但鉴于在相关时间窗口内部署可用 Group 1–3 拦截系统所需的自主能力开发速度,这种路径不太可能。[CP033, CP034, CP035]
3.5 护城河耐久性与商品化风险
Harmattan AI 的主要护城河来自结构和关系两端。Dassault Aviation 合作给了它一条进入北约标准作战飞机项目(Rafale F5 与 UCAS)的、公开可查的一线 OEM 集成路径;在欧洲防务自主系统初创公司里,这一点独一份。Helsing、TEKEVER 和 Quantum Systems 尚未公开复制同等项目深度。法国和英国国防部的正式采购项目,靠漫长政府合同周期、高切换成本,以及嵌入 Harmattan Sonora UAS 与 Kalahari C2 生态的训练依赖,形成采购锁定。第二个结构优势是纵向整合的全谱系产品组合——SAR 载荷(Sahara)、训练 UAS(Sonora)、打击系统(Barkhan)、C-UAS 动能拦截器(Gobi、Gobi Tempest)和 C2 (Kalahari)。欧洲国防部若要更换 Harmattan,竞争者必须复制整套体系级系统栈,而不是补一个单点能力。三类商品化风险已具实质性。第一,乌克兰战争推动欧洲 FPV 无人机产能扩张,乌克兰、芬兰和法国厂商同步加速,正压低全行业单机硬件成本,也削弱买方为先进平台设计支付溢价的意愿。第二,Anduril 估值 $61 billion、收入基数 $2.2 billion,有资本消化欧洲市场进入亏损;其 2026 年 5 月赢得荷兰国防部项目,指向更大范围的欧盟扩张。一旦 Anduril 拿到 OCCAR 或 EU Defence Fund 资格,Harmattan 的主权叙事会被挑战。第三,开源视觉模型和 LLM 衍生推理框架推动自主系统软件栈商品化,未来三到五年可能侵蚀软件自主层的差异化;不过 Harmattan 的 SAR 传感器集成和多系统蜂群协同仍是更可防守的细分壁垒。整个竞争集合的价格都不透明;Harmattan AI、Helsing、Anduril、TEKEVER 均未公开标价。没有直接采购尽调,就无法比较成本效益。[CP036, CP037, CP038, CP039, CP040, CP041]
| 护城河主张 | 主要威胁 | 严重性 | 缓释措施或尽调问题 |
|---|---|---|---|
| Dassault Aviation 合作 / Rafale F5 + UCAS 集成 | 竞品拿下可比的 Tier-1 OEM 合作(Airbus D&S 投资 Quantum Systems;Anduril 拥有 Boeing/USAF CCA) | 高 | 核实 Dassault 合作是否包含排他条款或 IP 授权,限制竞品获得同等交易;评估 Airbus D&S 对 Quantum Systems 的投资范围 |
| 法国和英国正式列装项目(PoR) | 续约时丢单,或在多供应商框架下重新竞标 | 高 | 确认合同期限、重新竞标条款和单一来源理由;注意公开资料未披露合同金额 |
| SAR 载荷(Sahara)的全天候 ISR 差异化 | 商业 SAR 即服务既有玩家(Capella Space、ICEYE)把机载 SAR 传感商品化 | 中 | Harmattan 的优势在系统集成;核实 Sahara L–S 波段规格是否自研,还是采用商用 SAR 模块 |
| 垂直整合的全谱系产品组合 | 竞品通过定向收购拼出同等技术栈,类似 Shield AI 收购 Aechelon | 高 | 跟踪 Anduril 和 Shield AI 在欧洲的收购动作;评估 Harmattan 向 EW 和地面机器人领域扩展能力的速度 |
| 欧洲主权与 NATO 阵营一致性 | Anduril 2026 年拿下荷兰国防部,以及潜在 EU Defence Fund 资格,削弱主权溢价 | 中 | 确认 Anduril 的 EU 主权技术转移承诺深度;ITAR 豁免不等于完整主权转移 |
| 嵌入式自主作战理念(Kalahari + 边缘计算) | 开源 LLM/CV 技术栈在 3–5 年内降低自主层进入门槛 | 中 | 核实 Harmattan 推理栈的自研属性;评估 SAR 集成和多系统蜂群协同是否真正差异化 |
| 规模制造(公司称数千台 / 月) | 乌克兰和欧洲的低成本 FPV 生产商在规模化后以硬件单元经济性压低 Harmattan | 高 | 没有单台成本数据;评估制造护城河前,需要独立工厂核验和成本基准对比 |
严重性评级是基于截至 2026 年 6 月收集到的竞争证据作出的定性判断。"高" 表示如果不缓释,威胁可能在 24 个月内实质侵蚀 Harmattan 的竞争位置。 "中" 表示威胁真实存在,但演进更慢或可被部分应对。这些评估没有独立生产审计或合同审计作为基础。
[CP036, CP038, CP040, CP041, CP042, CP005]Anduril 的 $61B 估值和 2025 年 $2.2B 收入构成上限锚点;Harmattan AI 的 $1.4B 估值约小 43 倍。TEKEVER 是唯一确认盈利的欧洲本土公司。MyDefence 在已验证部署系统数量上领先。
估值是已公布融资轮的投后估计。Helsing 约 $18B 的数字来自 2026 年 5 月新闻报道,截至运行日尚未获官方确认。Harmattan 的 $1.4B 基于多家可信新闻来源报道的 2026 年 1 月 Series B。TEKEVER 的 £400M 是项目承诺,不只是股权融资。
[CP001, CP002, CP008, CP009, CP013, CP019]3.6 图表
04财务情况
4.1 商业模式、收入流与变现方式
Harmattan AI 采用面向政府的 B2G 收入模式,核心锚点是防务采购合同。收入主要来自自主系统硬件交付——无人机、反无人机拦截器、电子战平台和 C2 软件——这些产品在正式采购项目下直接卖给国家国防部和盟军。公司披露的两个正式采购项目分别来自法国国防部(2025 年 7 月,1,000 架 SORONA 无人机)和英国国防部(2025 年 9 月,最多 3,000 套自主系统)。这确认核心收入机制是一次性总额或里程碑政府合同,而不是经常性订阅收入。 公司官方新闻稿称,Series B 资金将用于「在新作战区域部署 AI 赋能任务、把 Harmattan AI 的产品组合延伸到新领域,并扩大工业化制造」。这表明近期主导收入流仍是硬件交付合同。第二条新兴收入流,是为 Dassault Aviation 的 Rafale F5 和 UCAS 项目提供嵌入式 AI 集成服务;该业务仍处早期,未披露财务条款。第三条收入路径,是借助 Skyeton 合作做技术许可(把 Harmattan 的 Sahara SAR 载荷集成进 Raybird UAS);2025 年 10 月的合作公告已确认该路径,但收入贡献和结构并未公开。 Harmattan 官网明确表示,网站只是「静态目录」,没有电商功能;所有销售都要靠直接接触军方客户和国家采购机构推进(法国为 DGA,英国为 DE&S)。这确认渠道分销、经销安排和商业定价透明度在结构上缺位:公司没有标价,没有公开费率卡,也没有披露任何产品线的单价或单合同金额。 CEO 在 2026 年 2 月接受 Le Grand Continent 采访时提出一个很有分量的结构问题:M'Ghari 明确主张一种类似疫情期间疫苗制造合同的「take-or-pay」采购模式,即国家即使不承诺采购量,也为产能出资。若 Harmattan 拿到这类安排,将形成轻资本的经常性付款流;截至 2026 年 6 月,尚无证据显示已有生效的 take-or-pay 合同。[CI001, CI002, CI003, CI004, CI005, CI006]
| 收入来源 | 机制 | 单位 | 当前状态 / 价值 | 收入质量 | 尽调问题 |
|---|---|---|---|---|---|
| 直接硬件交付——ISR / 打击无人机 | 政府防务采购合同(正式列装项目,PoR);按里程碑或交付确认 | 按套或按批;基于合同 | 已确认 2 个正式列装项目(法国 1K 台、英国最高 3K 台);合同金额未披露 | 高——主权信用交易对手;回款风险低 | 要求提供完整合同 P&L、单价、逐合同毛利率 |
| 反无人机拦截器交付 | 面向 Gobi C-UAS 系统的政府采购;同一 PoR 合同模式 | 按套或按批;基于合同 | 公开未确认独立 PoR;管线由能力页面和 Dassault 合作范围推断 | 未知——潜在管线;无确认收入 | 确认现行 C-UAS 合同和合同金额 |
| 嵌入式 AI / 软件集成——Dassault 合作 | 面向 Rafale F5 和 UCAS 嵌入式 AI 的开发服务与软件授权 | 按项目 / 按里程碑 / 按许可;条款未披露 | 合作于 2026 年 1 月宣布;收入流未披露;收入来源仍早期 | 可能具备高质量经常性;由主权防务项目锚定 | 要求提供开发合同条款、里程碑付款和版税结构 |
| 传感器 / 载荷授权——第三方 UAS 集成 | 向 Skyeton Raybird 授权 Sahara SAR 载荷技术;可能有经常性费用 | 按集成或按台收取版税;结构未披露 | Skyeton 合作已于 2025 年 10 月确认;收入条款未披露 | 未知——早期;合作方收入未公开 | 要求提供 Skyeton 授权条款及任何收入分成协议 |
收入数据来自 Harmattan AI 官方新闻稿、Dassault Aviation 新闻资料包(2026 年 1 月)、 TechCrunch(2026 年 1 月)和 Defense News(2026 年 1 月)。未公开披露标价、合同金额或年收入数字。 所有收入状态评估仅基于公开合同公告。收入质量评估为分析判断,未基于已核验财务数据。
[CI003, CI004, CI005, CI006, CI007, CI008]| 产品 / 平台 | 定价模式 | 标价 vs. 实现价格 | 已披露金额 | 定价不透明度 | 来源 |
|---|---|---|---|---|---|
| SORONA ISR 无人机(四旋翼) | B2G 采购合同;按台或按批定价 | 无标价;通过 DGA 采购竞争性招标实现 | 法国国防部(1,000 台为数百万美元级;代理估算约 $2K–$20K / 台;实际未披露) | 完全不透明——无标价;单台成本未披露 | Army Recognition(2025 年 7 月);Harmattan AI 新闻稿(2026 年 1 月) |
| Gobi / Gobi Tempest C-UAS 拦截器 | B2G 采购合同;按台定价 | 未披露标价;拦截器定价通常高于 ISR 无人机 | 未披露 | 完全不透明——没有价格或合同数据 | Harmattan AI 网站(harmattan.ai/systems/gobi/) |
| Kalahari C2 平台 | 企业许可或平台集成;基于合同 | 未披露标价或企业许可条款 | 未披露 | 完全不透明——未披露定价 | Harmattan AI 网站(harmattan.ai/systems/kalahari/) |
| Sahara SAR 载荷 | 硬件 + 授权;条款取决于集成模式 | 未公布标价或授权费率;Skyeton 合作意味着存在集成定价 | 未披露 | 完全不透明——定价未披露;Skyeton 合作条款不透明 | Harmattan AI Skyeton 合作公告(2025 年 10 月) |
Harmattan AI 任何产品线都没有标价;网站明确说明所有销售都通过直接接触军事客户完成, 没有电商功能(使用条款,2025 年 11 月)。SORONA 单台估算值基于 1,000 台合同的 "数百万美元级" 描述以及防务无人机市场基准,属于分析估算。估算值未经核验, 不应作为谈判基准使用。
[CI009, CI010, CI005, CI006]Harmattan AI 的收入从政府国防部采购流向硬件交付合同,再进入收入确认;毛利未披露。Dassault 和 Skyeton 伙伴关系代表正在形成的次级收入节点。
收入和毛利节点是定性描述;公开来源没有可用数字。收入确认步骤反映对 B2G 合同模式的分析,不是披露的会计政策。Dassault 和 Skyeton 收入流仍在早期,对总收入的贡献未知。
[CI003, CI004, CI005, CI006, CI007, CI008]4.2 单位经济、成本结构与资本密集度
Harmattan AI 没有披露任何单位经济数据——毛利率、贡献毛利、CAC、回本周期或 LTV——这与其发展阶段和私营身份一致。公开证据可以搭出代理估算,但不确定性很高,不能替代尽调室里的财务资料。 纵向整合的防务硬件制造商,成本结构天然偏资本密集。2026 年 6 月,Harmattan 的运营成本基底包括 130 多名员工,中位从业经验 15 年,分布在法国、英国、瑞士、摩洛哥、UAE 和美国办公室。截至 2026 年 6 月,开放岗位包括制造岗位(Wissous 的生产操作员、产线主管、质量技术员)、系统架构师、总工程师、组长,以及多国 R&D 工程师,显示这是一个高成本、资深员工占比高的组织。按员工数粗估烧钱速度——130+ 名员工、欧洲防务行业混合人工成本约每人每年 $150K–$200K——意味着在资本开支、零部件采购和制造设施开销之前,年度现金运营费用约 $20–$35 million,约合每月 $1.5–$3 million。 收入侧,Harmattan 防务无人机硬件的毛利率未知。可比的硬件加软件防务主承包商显示,硬件交付合同毛利率约 15–40%,取决于 R&D 成本资本化还是费用化;软件或 AI 集成服务毛利率更高,约 50–70%。Harmattan 的纵向整合模式——自研机体、传感器和软件——理论上会减少第三方组件利润抽成,但在低产量时也会放大固定成本吸收风险。 CEO 在 Le Grand Continent 采访中点出两项结构性供应链成本风险:(1)无刷电机磁体含稀土元素,目前依赖中国稀土精炼链;(2)锂聚合物电池的电池化学也依赖中国稀土。两者都被认定为会持续三到五年的约束,意味着替代供应链成熟前,零部件成本和供应集中风险都会维持高位。这些供应链暴露会直接影响硬件合同的 COGS 和毛利率。 产量规模是单位经济的关键杠杆。Harmattan 声称具备「每月数千套系统」产能,DroneXL 的分析还引用公司最高每月 10,000 架无人机的计划。如果任一数字属实,固定制造开销将摊到很大的生产基数上,单位经济会明显改善。但两项数字都未获独立验证。[CI011, CI012, CI013, CI014, CI015, CI016]
| 指标 | 数值 / 代理估算 | 置信度 | 为什么重要 | 尽调问题 |
|---|---|---|---|---|
| 毛利率(硬件交付) | 未披露;防务硬件代理区间 15–40% | 低(仅代理估算) | 决定合同收入能否转化为经营杠杆 | 索取合同级 P&L;逐台 COGS 拆分 |
| 毛利率(AI / 软件集成) | 未披露;软件 / AI 集成代理区间 50–70% | 低(仅代理估算) | 若 Dassault 收入流放大,软件利润率会显著抬高混合毛利率 | 要求提供 Dassault 开发合同毛利结构 |
| 月度烧钱速度 | 未披露;按员工数代理估算约 $1.5M–$3M / 月 | 低(估算代理) | 决定现金跑道和下一轮融资时点 | 索取过去 12 个月经审计烧钱速度 |
| CAC(获客成本) | B2G 场景不适用;政府采购周期 6–18 个月 | N/A——指标与 B2G 模式无关 | B2G "CAC" 是关系和资格投入,不是 CRM 指标 | 梳理每份合同对应的政府关系和资质投入 |
| 平均合同金额 | 未披露;1K 台 PoR 的底线是 "数百万美元级" | 低(仅底线估算) | 决定单客户收入集中风险 | 索取法国国防部、英国国防部及其他客户的完整合同金额 |
| 人均收入 | 未披露;没有可用收入数字 | 低(无法计算) | 衡量 R&D 与商业生产率的代理指标 | 收入披露后可解决 |
| 单位制造成本——SORONA 无人机 | 未披露;配备 EO/IR 传感器的 sub-$1.8 kg 系统;市场代理估算 $1K–$8K / 台 | 低(仅代理估算) | 决定硬件毛利率底线 | 要求提供 COGS 拆分和逐台制造成本 |
| 经营利润率 | 未披露;按增长轨迹和制造投入看,尚未盈利 | 低(仅定性) | 确认通往盈利的时间窗口 | 索取覆盖下一轮触发点的 P&L 预测 |
所有代理值均为分析估算,来源包括员工数数据(Defense News,2026 年 1 月)、 欧洲防务硬件制造商行业基准,以及市场研究(BCG C-UAS Whitepaper,2026 年 5 月; StartUS Insights Military Drones 2026)。没有可用的已核验财务数据;所有代理值都只应视作方向性判断, 并在尽调中用公司实际财务数据验证。
[CI012, CI013, CI014, CI015, CI016, CI017]从合同授标到制造成本再到毛利率,单位经济流转只能做定性判断;所有利润率数字都是代理值或未知。供应链和人工成本是主要 COGS 驱动因素。
所有单位成本估计均为基于行业基准和披露员工数的代理值。没有经验证的 COGS 或毛利率数据。中国稀土依赖时间线来自 CEO 接受 Le Grand Continent 采访(2026 年 2 月)。人工成本代理值基于欧洲国防行业平均全包成本。
[CI013, CI014, CI015, CI016, CI019]4.3 资本充足性、融资历史与融资依赖
Harmattan AI 融资节奏很快,相比公司年龄,目前资金相当充足。「公司概况」章节已覆盖完整时间线融资记录;本节聚焦 2026 年 6 月时点的前瞻资本充足性。Series B 于 2026 年 1 月 12 日完成,由 Dassault Aviation 领投,带来 $200 million 新资金;Motier Ventures 续投。加上 Series B 前估计 $42 million 融资(Atlantic VC 领投种子轮;FirstMark Capital 领投约 $30 million 的 Series A),公司累计融资约 $242 million。Dassault 的具体出资额,以及 Motier 之外的共同投资方身份,均未披露。 Series B 资金用途——扩大制造、新产品领域和 AI 部署——符合防务硬件制造商规模化时的资本密集属性。无人机产线、零部件采购库存、测试基础设施和多国运营,都需要持续现金投入。公司未公开披露任何债务融资、项目融资安排或信贷额度,说明股权融资仍是唯一资本来源。 账上现金、烧钱速度和现金跑道均未披露。若按每月 $2–3 million 代理烧钱速度(基于员工数估算),并以 2026 年 1 月完成 $200 million 融资计算,在当前运营规模下可粗略推出 5–8 年跑道——但该估算忽略了制造资本开支、零部件采购营运资金,以及未来规模化招聘,这些都会显著压缩跑道。没有任何资本市场文件、注册公开债务或已发布季度报告,因此所有跑道估计都是分析代理,不是已验证数字。 Dassault 合作给资本充足性增加了超出资金投入的战略维度:Dassault 承诺把航空航天项目开发经验、供应链接入和国际业务拓展网络带给 Harmattan。虽然这类非货币资本不进资产负债表,却能降低公司最高价值潜在收入流(Rafale F5 集成、UCAS 项目)的商业化和市场进入风险。[CI020, CI021, CI022, CI023, CI024, CI025]
| 项目 | 数值 / 状态 | 置信度 | 来源 | 缺口 / 尽调问题 |
|---|---|---|---|---|
| 手头现金(Series B 后) | 未披露;2026 年 1 月融资 $200M 是总注资额 | 低——未披露资产负债表 | Harmattan AI Series B 新闻稿(2026 年 1 月) | 索取截至最近季度的经审计资产负债表 |
| 月度现金消耗 | 未披露;代理估算约 $1.5M–$3M / 月(基于员工数) | 低(代理估算) | Defense News(2026 年 1 月);分析估算 | 索取过去 6–12 个月管理账 |
| 估算现金跑道 | 未披露;仅按运营烧钱代理估算为 5–8 年;会被资本开支压缩 | 低(代理估算,不含制造资本开支) | 基于已披露 Series B 和员工数数据作出的分析估算 | 对照制造扩张计划核验资本开支之后的现金跑道 |
| Series B 募资计划用途 | 制造扩张、新产品领域、扩大 AI 部署 | 高 — 官方新闻稿已披露 | Harmattan AI Series B 新闻稿(Jan 2026);Dassault Aviation 新闻资料包(Jan 2026) | 要求提供资金投放时间表和里程碑触发条件 |
| 下一轮融资触发点 | 未披露;可能取决于制造规模化里程碑和 / 或第二波列装项目落地 | 低 — 未披露公开里程碑 | 基于 CEO Le Grand Continent 访谈(Feb 2026)的分析推断 | 询问下一轮融资条件、预期时间和隐含稀释 |
| 债务 / 项目融资义务 | 未公开披露;现有资本看起来全部来自股权 | 低 — 未披露不等于没有债务 | Harmattan AI 新闻稿;未提及债务融资安排 | 确认是否存在未披露授信、项目融资负债或设备租赁 |
资本充足性数据来自 Harmattan AI Series B 新闻稿(Jan 2026)和 Dassault Aviation 新闻资料包(Jan 2026)。公司未披露在手现金、烧钱速度、现金跑道和债务;代理估算使用 Defense News(Jan 2026)披露的员工数,以及公开可得的欧洲防务行业人工成本基准。所有代理值 只代表方向性估算。
[CI020, CI021, CI022, CI023, CI024, CI025]关键财务指标区间分为来源支撑和代理估计两类。估值和累计融资已确认;合同金额和烧钱速度是分析代理值,不确定性高。尽调中应对照公司财务数据验证所有估计。
估值和累计融资区间有来源支撑。合同金额区间和烧钱速度是分析代理估计,尚未获公司确认;它们反映行业基准,而非公司披露数据。合同金额可能按法国或英国国防采购规则保密。烧钱速度不包括制造资本开支。
[CI020, CI021, CI022, CI023, CI026, CI027]4.4 公开财务缺口与披露不透明
Harmattan AI 的财务披露姿态,符合一家私营阶段法国防务公司的常态:没有监管义务公布收入、利润率或详细资本结构数据。但从投资尽调视角看,缺少公开财务指标会造成基本面的承销不透明。截至 2026 年 6 月,任何公开可访问的新闻稿、投资者信、监管文件或财务报表中,都没有披露 ARR、收入运行率、毛利率、经营亏损、烧钱速度或现金跑道。 阻碍承销的具体缺口包括:两个正式采购项目的合同总额(可建立收入基线)、硬件交付毛利率(决定业务在规模化后是否根本可行)、烧钱速度与现金跑道(决定近期稀释风险),以及 Dassault 在 $200M 轮中的具体股权出资额(影响稀释和董事会动态)。研究期间访问 Crunchbase 档案返回 403 错误,未提供补充财务数据。CEO 在 Le Grand Continent 的访谈是最实质的公开披露,谈到供应链、地缘政治战略和产业政策,但未披露任何收入、利润率或烧钱数字。[CI028, CI029, CI030, CI031, CI032]
| 缺失指标 | 当前状态 | 对分析的影响 | 尽调路径 |
|---|---|---|---|
| ARR / 收入运行率 | 未披露;私人公司没有发布义务 | 无法计算收入倍数和评估增长率;$1.4B 估值缺少可比基准 | 索取管理账或董事会汇报包;询问过去 12 个月收入 |
| 毛利率(混合口径) | 未披露;没有公开的单元级或汇总毛利数据 | 无法判断硬件交付是否带来可投资回报;毛利率路径不明 | 索取法国国防部和英国国防部项目的合同级 P&L;询问各产品线毛利率 |
| 月度烧钱速度 / 现金消耗 | 未披露;只有员工数代理指标可用 | 无法判断现金跑道;若低估资本开支需求,可能触发稀释性的过桥融资 | 索取过去 6 个月 P&L 和现金流量表;对照薪资记录核验 |
| Series B 中 Dassault 出资额和共同投资人构成 | Dassault 领投;Motier Ventures 续投;其他投资人未具名 | 稀释和董事会动态不清;在军民两用市场,战略投资人可能带来利益冲突 | 索取股权结构表;确认 Dassault 是否持有董事席位或观察员权利 |
| 法国国防部和英国国防部合同金额 | 分别描述为“数百万美元”和“数百万英镑”;具体数字未公开 | 无法核验收入集中度、合同毛利,或按披露产量持续制造的能力 | 索取合同授予文件(如非保密);确认交付完成情况和付款节奏 |
| 债务、项目融资或授信安排 | 未公开披露;未披露并不确认不存在 | 未披露债务会影响真实现金跑道和资本结构;制造业常见设备融资 | 索取完整负债明细;确认没有未披露借款 |
| 收入确认政策 | 未披露公开会计政策;法国 SAS 没有强制公开披露义务 | 按里程碑还是按交付确认收入,会改变报告收入时点;这对管线分析很关键 | 索取会计政策备忘录;确认收入按交付还是按合同里程碑确认 |
所有缺口基于 2026 年 6 月的公开研究记录。Harmattan AI 是法国 SAS(简化股份公司),没有公开发布财务报表的法定义务。以其阶段和司法辖区看,财务不透明并不异常, 但仍是重大尽调限制。Crunchbase 资料页返回 403 访问错误;该来源没有可用的补充财务数据。
[CI028, CI029, CI030, CI031, CI032]$200M Series B 资本池投向制造扩产、新产品领域扩张和 AI 部署——这些都是资本密集型活动。供应链投资和多国运营让现金需求持续高于表面员工成本。
各用途类别的资本配置比例未披露;该流程展示 2026 年 1 月 Series B 新闻稿列明的募资用途类别,没有量化。公司未发布具体 capex 预算或 R&D 支出。
[CI024, CI025, CI031, CI034, CI035]4.5 财务结论——收入质量、利润率路径与尽调阻碍
2026 年 6 月,Harmattan AI 的财务图景由三件事并存构成:收入质量信号强、单位经济高度不透明、短期资产负债表资金充足。 凡是能评估的地方,收入质量都很高。与两个北约国家政府签订正式采购项目,意味着客户是主权信用交易方——这是防务公司能拿到的信用安全性最高的客户类别。政府合同通常按里程碑付款并带部分预付款,可降低回款风险。合同落地速度(成立 14 个月拿到法国国防部项目;15 个月拿到英国国防部项目;CEO 访谈称 7 个月拿到爱沙尼亚订单)显示出同阶段大多数防务初创公司不具备的运营交付能力。Dassault 合作增加第三层收入质量:为一家收入 $6.2B 的主承包商开发嵌入式 AI,可带来稳定、多年的开发收入,不受典型硬件采购周期完全约束。 利润率路径是关键未知数。防务硬件制造天然低于软件毛利率,Harmattan 的 SORONA 无人机还要在质量之外比拼成本效率——DGA 采购明确瞄准低单价。公司宣称「每月数千套系统」的产能会改善固定成本吸收,但中国稀土供应链依赖和欧洲高人工成本会压低近期毛利率。没有合同级定价或成本拆分,就无法以有用精度估算毛利率。 资本密集度是近期首要风险。扩大无人机制造需要重资本投入:产线、质量体系、供应链管理和多国运营都要持续投钱。$200M Series B 足以支撑当前增长阶段,但不足以支撑公司提到的每月 10,000 套目标。下一轮融资更可能由制造规模化里程碑驱动,而不是单纯由收入指标驱动。 主要尽调阻碍:(1)没有收入或 ARR 披露;(2)没有毛利率披露;(3)两个正式采购项目的合同金额被保密 / 暂不披露;(4)Dassault 出资额和共同投资方结构未披露;(5)没有债务或项目融资披露;(6)没有烧钱速度或现金跑道数据。[CI033, CI034, CI035, CI036, CI037, CI038]
4.6 图表
05产品与技术
5.1 双层自主架构
Harmattan AI 把自主能力拆成两层互补架构,意在把人类指挥权和机器执行速度解耦。上层是任务编排,指挥官和操作员可以在人工授权参数内定义目标、保持态势感知、排序威胁,并协调分布式系统交战。下层是平台自主,单个战术边缘系统可执行分配目标、适应战场变化,并在授权参数内维持作战连续性。这样的分工意味着,指挥官用 Kalahari 设定意图;每个平台(Gobi、Barkhan、搭载 Sahara 的 UAS)随后在批准边界内自主行动,不需要每个动作都由人确认。人类仍在决策环路中掌握授权,机器则以机器速度执行。 公司的总体作战理念是「协同饱和」:自主机器人力量在全作战谱系内像一个统一整体行动,把军事优势从单个平台性能,转向大规模部署和协调自主系统的能力。产品策略有三根设计支柱:可消耗(系统可批量部署、可承受损失)、可扩展(在机器人优先时代,制造和战时增产是核心差异点)和自主(不依赖集中式基础设施的边缘推理,同时保留人类决策权)。每个产品模块都设计成既能独立运行,也能在 Kalahari 编排下接入统一系统。这套架构也支撑 Dassault Aviation 合作:在同样的双层模型里,Dassault 系统定义意图,Harmattan AI 自主平台负责执行,目标是为 Rafale F5 和 UCAS 协同空战嵌入 AI 能力。[CE001, CE002, CE003, CE004, CE005, CE028]
| 层级 / 组件 | 系统角色 | 关键依赖 | 风险 |
|---|---|---|---|
| 硬件平台层(Sonora、Gobi、Gobi Tempest、Barkhan、Sahara) | 在战术边缘实际执行 ISR、拦截和打击任务 | 组件供应(电池、光学传感器、RF 链路);装配能力 | 供应链不透明;未指明制造设施;声称的产量未经验证 |
| 嵌入式边缘 AI(平台自主性) | 机载推理支撑目标检测、轨迹优化、末端制导 | 边缘计算硬件(芯片供应商未披露);ML 模型训练管线 | 芯片供应集中;新型威胁环境下模型可能漂移;未披露 OTA |
| Sahara SAR 处理层 | 机载 ACD/MTC/CCD;不靠云即可实时交付情报 | L–S 波段 RF 组件;电源总线(<65 W);UAS 宿主平台集成 | 宿主平台存在功耗预算约束;SAR 组件 ITAR 状态未披露 |
| Kalahari C2 编排层 | 意图驱动的任务分配;通用作战图;威胁优先级排序 | 网络基础设施(边缘到云或本地部署);与伙伴 C2 的 API 兼容性 | GPS / 通信拒止环境下的网络韧性;单一供应商 C2 锁定风险 |
| 互操作 / 标准层 | NATO 协议合规(Asterix、CoT、Lattice、Sapient);跨系统传感器融合 | 伙伴系统 API 可用性;NATO 认证流程 | Lattice 互操作性意味着依赖 Anduril 标准 API;认证未公开 |
| 安全 / 访问控制层 | E2E 加密;RBAC;审计轨迹;防务网络安全加固 | 加密库供应商;密钥管理基础设施;固件更新链 | 未发布独立渗透测试或网络安全审计;固件供应链未知 |
架构细节来自截至 2026 年 6 月的 Kalahari 官方页面和产品规格页。芯片供应商、ML 框架和云服务商身份均未公开披露。依赖和风险是分析师基于已披露架构特征作出的推断; Harmattan AI 未确认这些判断。
[CE001, CE003, CE022, CE024, CE025, CE026]从指挥意图到实体平台执行,六层视角拆解 Harmattan AI 自主系统架构,并展示任务编排与平台自主之间的分离。
[CE001, CE002, CE003, CE022, CE024]5.2 产品模块栈——Sonora、Sahara、Gobi、Gobi Tempest、Barkhan
Harmattan AI 提供五个硬件产品模块,覆盖从操作员准备、持续 ISR、动能 C-UAS 防御到自主打击的完整任务生命周期。 Sonora 是小型 UAS,用于单位级训练和操作员熟悉装备。规格包括:续航大于 40 分钟,1 分钟内起飞准备完成,航程 2 km,外形尺寸 35 × 42 × 11 cm,含电池重量低于 1.8 kg。EO 相机以 30 fps 输出 1,280 × 720 像素,支持 6× 变焦(3× 无损)。IR 相机以 30 fps 输出 320 × 240 像素,配 2 轴机械稳定。系统具备等同 IP53 的防护等级,支持夜间和低照度作业,并能在仿真模式下接入 Kalahari,保持训练连续性。 Sahara 是自主 UAS 搭载的 SAR 载荷,覆盖全天候 ISR、目标检测、地形分割和海上监视。关键规格:L–S 频段、双线性极化、2 km 时 0.25 m 空间分辨率、10 km 时 1.2 m 空间分辨率,物理尺寸 15 × 72.3 × 15 cm,重量低于 3.5 kg,功耗低于 65 W。工作模式包括 Stripmap 和 circular;机载变化检测算法包括幅度变化检测(ACD)、多时相相干(MTC)和相干变化检测(CCD)。CEO 公开称 Sahara 作为面向 150 kg 以下无人机的 SAR 雷达「世界独一无二」。该说法未获独立基准验证。 Gobi 是面向 Group 1–2 UAS 威胁的动能拦截器。巡航速度:250 km/h;最高速度:350 km/h;命中动能:大于 10,300 J;发射准备:在发射箱常开状态下为 0 秒;重量:2.2 kg;外形尺寸:32 × 34 × 34 cm。系统在无地面依赖下执行完全自主的直接撞击杀伤末端制导,从发射授权到威胁消除约需一分钟。 Gobi Tempest 针对 Group 2–3 威胁,外形更大。巡航速度:215 km/h;最高速度:310 km/h;战斗部兼容:800 g;航程:12 km;发射准备:5 秒;含电池和战斗部重量:4.4 kg;外形尺寸:46.5 × 46.5 × 69.5 cm。系统配备经认证的可逆 SAU,并支持协同空间作业。两款 Gobi 都包含完整系统,由发射箱、地面雷达、EO/IR 转塔和 C2 集成组成。 Barkhan 是自主打击平台,用于争夺环境中的快速目标交战。速度:120 km/h;航程:20 km;外形尺寸:43 × 43 × 40 cm;机体重量低于 2 kg;载荷能力:3 kg;续航:25 分钟;无线电链路:160–2,500 MHz。Barkhan 支持 GNSS 拒止环境导航、自主末端制导、光纤兼容,可单机运行,也可在协同蜂群配置中运行。五个硬件模块都被描述为平台无关、可集成任何 C2 平台,Kalahari 是推荐编排层。[CE006, CE007, CE008, CE009, CE010, CE011]
| 模块 / 资产 | 主要用户 | 状态 / 成熟度 | 关键差异化 | 尽调缺口 |
|---|---|---|---|---|
| Sonora(训练 UAS) | 操作员 / 教官 | 已部署 — 列装项目(法国国防部 1,000 台) | >40 min 续航,<1 min 部署,IP53;EO + IR,并接入 Kalahari 仿真模式 | 没有独立测试报告;相对真实 UAS 的训练效果未验证 |
| Sahara(SAR 载荷) | ISR 操作员 / 情报分析师 | 已部署 — 已接入 Skyeton Raybird;Q1 2026 首批交付 | 2 km 距离 0.25 m 分辨率;<3.5 kg;全天候;机载 ACD/MTC/CCD 处理 | 分辨率数字来自公司披露;没有独立基准;成像延迟未知 |
| Gobi(Group 1-2 拦截器) | 反无人机操作员 / 部队防护 | 已部署 — 英国国防部合同(据报道 3,000 台) | >10,300 J 撞击能量;0 s 待命;~1 min 压制;自主命中即摧毁 | 没有公开记录的实地交战;附带损害测试未获公开验证 |
| Gobi Tempest(Group 2-3 拦截器) | 反无人机操作员 / 部队防护 | 在售 — 可用;部署规模未披露 | 800 g 弹头;12 km 航程;5 s 待命;合格可逆 SAU;全天候 | 弹头认证机构未披露;交战记录保密 |
| Barkhan(打击 UAS) | 打击操作员 / 特战部队 | 在售 — 可用;部署规模未披露 | 20 km 航程;GNSS 拒止导航;3 kg 载荷;兼容光纤;可集群作战 | 主动 EW 环境下的 GNSS 拒止性能未经独立验证 |
| Kalahari(C2 / 编排) | 指挥官 / 操作员 / 教官 | 在用 — 已用于列装项目;API 驱动的 SaaS 式部署 | Asterix/CoT/Lattice/Sapient 标准;RBAC;E2E 加密;本地部署或边缘到云 | 没有独立网络安全审计;Lattice 互操作性未验证;未披露 SLA |
状态来自截至 2026 年 6 月的官方产品页、新闻稿和国防部合同公告。Gobi Tempest 和 Barkhan 的部署规模未公开量化。“在售”指商业可用; “已部署”指至少有一项确认的国防部交付或作战集成记录。
[CE006, CE010, CE015, CE018, CE020, CE022]| 用户任务 / 任务能力 | 没有 Harmattan 时的现有流程 | Harmattan 方案 | 可衡量收益(公司声称) | 限制 / 缺口 |
|---|---|---|---|---|
| 部队战备 / UAS 操作员训练 | 真实飞行器熟悉训练;纯仿真工具缺少战场逼真度 | Sonora 训练 UAS 搭配 Kalahari 仿真模式,复刻真实任务参数 | 可扩展到分队级训练;<1 min 部署;AI 生成复盘洞察 | 相对对抗性 EW 环境的训练逼真度未验证;没有通过率数据 |
| 持续 ISR / 全天候监视 | 光学 ISR 平台受天气和光照限制;SAR 系统体积大 | Sahara SAR 载荷装在轻型 UAS(<3.5 kg)上;机载处理;不依赖云 | 2 km 距离 0.25 m 分辨率;ACD/MTC/CCD;全天候持续覆盖 | 分辨率声明没有独立基准;集成伙伴仅限 Raybird |
| 精确打击 / 协同交战 | 传统弹药或大型 UCAV;交付周期长;大规模饱和能力有限 | Barkhan 自主打击 UAS;可单机或集群;GNSS 拒止导航;20 km 航程 | 机体 <2 kg,载荷能力 3 kg;可快速大规模交战 | 实地交战记录未公开验证;对抗环境中的 EW 韧性未证明 |
| VSHORAD / 反无人机防御 | 软杀伤 RF 干扰器(MyDefence 等);缺少针对 Group 2-3 的动能命中即摧毁能力 | Gobi(Group 1-2)和 Gobi Tempest(Group 2-3)动能拦截器,配 AI 制导 | 发射后 ~1 min 交战;最高速度 350 km/h;12 km 航程(Tempest) | 没有独立杀伤链测试数据;附带损害认证未公开 |
| 多域 C2 / 任务编排 | 孤岛式平台 GCS(地面控制站);没有统一自主任务分配层 | Kalahari 统一 C2;实时威胁优先级排序;RBAC;嵌入式仿真 | 指挥官 / 操作员 / 教官共用单一视图;兼容 NATO 标准 | 没有面向 NATO C2 系统的独立集成测试;未披露 SLA 和正常运行时间 |
除非标为推断,收益均来自公司官方产品页和能力页的声明。截至 2026 年 6 月,任何任务能力都没有公开可得的独立客户案例研究或作战复盘报告。 限制列只反映公开记录的缺口,不代表保密缺陷。
[CE006, CE010, CE017, CE020, CE022, CE025]基于截至 2026 年 6 月公开记录的产品组合,比较 Harmattan AI 与四家欧洲 / 美国国防自主能力同行在六个任务维度上的覆盖。
评级基于截至 2026 年 6 月公开记录的产品页面、新闻稿和可信媒体报道。「✓」 表示已有公开文件佐证;「已声明」表示公司自称、但未获独立验证;「部分」 表示能力覆盖但不完整。未开展独立基准测试。Harmattan AI 的 GNSS 拒止环境表现和 Kalahari Lattice 互操作性仅为公司自述。
[CE010, CE015, CE020, CE039, CE040, CE042]5.3 Kalahari C2——编排架构与能力
Kalahari 是 Harmattan AI 的指挥控制系统,内嵌自主能力。它支持意图驱动任务下达、实时决策支持,以及跨分布式无人系统的闭环执行。系统采用 API 驱动、软件定义架构,可本地部署,也可用混合边缘到云配置部署;一条平台无关的数字骨干负责扩展,可把任何资产统一到多任务作战中,并支持作战中软件动态适配。 Kalahari 服务三类用户角色。指挥官借助统一的共同作战图,在战区范围内查看各扇区、任务和自主资产,定义任务目标、设定交战规则,并保持集中权威。操作员在被分配扇区内执行任务,拥有任务范围内可见性和控制权,实时协调自主系统,并在指挥设定的授权框架内适应战场变化。训练员在同一 C2 环境的仿真模式中工作,与实装系统断开,演练任务并验证作业流程,不影响正在进行的行动。 Kalahari 的数据层能力包括实时摄取和处理多源传感器数据,并完成航迹分类。系统把分布式传感器和任务输入整合成统一作战图,持续评估任务优先级和可用资产以协调响应,并在整个任务周期中传输经验证的指令,同时动态调整任务分配、监控交战结果。 互操作性建立在开放且防务级标准之上:Asterix(空中交通监视)、Cursor on Target(CoT,态势感知数据交换)、Lattice(Anduril 的传感器融合标准,指向跨平台兼容性主张)和 Sapient(NATO AI / 自主系统交换)。安全功能包括端到端加密、基于角色的访问控制、完整审计轨迹,以及按防务网络安全标准加固。内嵌仿真与数据织物支持训练、任务回放和行动后复盘,把训战转换链条和实战行动接起来。[CE022, CE023, CE024, CE025, CE026, CE027]
从指挥官定义意图,经 Kalahari 到自主平台执行,再到行动后反馈的逐步作战流程。
[CE022, CE023, CE027, CE034]5.4 部署、集成与路线图
Harmattan AI 的系统为快速部署而设计:Sonora 不到 1 分钟完成起飞准备;Gobi 在常开式发射箱中保持 0 秒准备状态;Gobi Tempest 5 秒准备就绪。每个系统都被描述为平台无关、可集成任何 C2 平台,因此除 Kalahari 本身外,也能与伙伴和北约标准 C2 系统互操作。 迄今最重要的集成里程碑,是 2025 年 10 月宣布的 Skyeton-Harmattan 战略合作:把 Sahara SAR 载荷和 Harmattan 的 AI 技术集成进 Skyeton 的 Raybird UAS。Raybird 已在乌克兰防务部队累计超过 350,000 小时战斗飞行,这使该合作成为 Harmattan 传感器技术接入经实战验证第三方平台的首个有文档记录案例。搭载 Sahara 的 Raybird 系统计划于 2026 年 Q1 初始部署。 Dassault Aviation 合作(2026 年 1 月,Series B 公告)面向 Dassault 未来作战航空系统——Rafale F5 和 UCAS 项目——嵌入 AI 能力,尤其是自主系统控制。这是最受关注的路线图承诺,但尚未公布生产认证级集成的交付时间线。 Harmattan 参加了 World Defense Show 2026(利雅得,2026 年 2 月)、UMEX 2026(阿布扎比,2026 年 1 月)和 SOF Week 2026(坦帕,2026 年 5 月);Eurosatory 2026(巴黎,2026 年 6 月)也确认其持续出现在北约市场防务展会上。公司声称每月交付数千套系统,但未披露生产设施名称,也未公布可独立核验的产量数字。[CE028, CE029, CE030, CE031, CE032]
| 日期 / 阶段 | 功能 / 里程碑 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| October 2025 | Skyeton-Harmattan 战略合作 — Sahara SAR 接入 Raybird UAS | 已完成(已公告) | 首个第三方平台集成;验证 Sahara 可作为授权传感器载荷 | 官方博客(harmattan.ai);skyeton.com;defence-industry.eu |
| Q1 2026(计划) | 首批搭载 Sahara SAR 的 Raybird 系统投入运行 | 公司计划(Oct 2025 公告) | Sahara 载荷在第三方平台上的最早作战证明点 | militarnyi.com;drone-actu.fr 媒体报道 |
| January 2026 | Dassault Aviation 领投 Series B $200M;Rafale F5 / UCAS 自主性集成纳入范围 | 已公告;路线图阶段;无交付时间表 | Dassault 合作让 Kalahari 有机会接入作战航空 C2 | 官方博客(harmattan.ai);Dassault 新闻稿 |
| February 2026 | World Defense Show 2026(Riyadh)— 产品组合展出 | 已完成 | 完整产品栈面向 NATO / 海湾市场推进业务拓展 | Harmattan 博客链接:harmattan.ai/blog/harmattan-ai-is-exhibiting-at-world-defense-show-2026/ |
| May 2026 | SOF Week 2026(Tampa)— 接触美国及盟军部队 | 已完成 | 释放进入美国市场的意图;目标买方群体是特种作战部队 | Harmattan 博客链接:harmattan.ai/blog/join-harmattan-ai-at-sof-week-2026/ |
| June 2026 | Eurosatory 2026(Paris)— 全产品组合展出 | 已排期 | 欧洲最大防务展;主要受众是法国国防部和 NATO 盟友 | Harmattan 博客链接:harmattan.ai/blog/harmattan-ai-is-exhibiting-at-eurosatory-2026/ |
所有里程碑来自公司公告(博客、新闻稿)或第三方新闻报道。Rafale F5/UCAS 集成时间表或生产交付计划没有公开可得的独立确认。 Q1 2026 Raybird-Sahara 部署状态在公告后尚未获得独立确认。
[CE028, CE029, CE030, CE031, CE032]支撑 Harmattan AI 产品堆栈的关键外部依赖,包括硬件供应、平台、标准组织、客户和伙伴;风险色调表示依赖严重度。
[CE028, CE030, CE031, CE035]5.5 信任、安全、合规与技术风险
Kalahari 的安全架构采用端到端加密、基于角色的访问控制和完整审计轨迹,并按防务网络安全标准加固。系统的角色分区模型(指挥官、操作员、训练员各有不同授权范围)是主要人在环路控制机制,确保自主执行只在人工授权参数内发生。Gobi 和 Gobi Tempest 都被描述为人在环路设计,但真实交战序列中的预授权程度没有独立文档证明。 Harmattan 任何产品都未公开披露针对自主致命系统的 NATO 或 EU 监管认证。根据其发布的隐私政策和使用条款,公司在数据基础设施层面遵守法国法律、GDPR 和 CCPA/CPRA。公司未公开发布任何涉密系统资质证书、REACH/RoHS 合规文件或电磁兼容性(EMC)认证。 竞品基准揭示出值得尽调的能力缺口。Quantum Systems 的 Vector AI 使用双 Jetson Orin 边缘计算板,已记录 GPS/GNSS 拒止环境自主导航和 180+ 分钟固定翼续航,并经实战条件验证;相比 Harmattan 自述规格,这是置信度更高的边缘 AI 基线。Shield AI 的 V-BAT 在海上环境提供有文档记录的 ViDAR ISR 覆盖率,达到 3,140 NM²/hr,但需要舰载发射基础设施;Gobi 的地面发射箱避开了这一点。TEKEVER 的 AR5 在 €30 million EMSA 框架合同下运行,是已验证的 ISR 海上 UAS,也提供一个获监管认可的欧洲主权基准。截至 2026 年 6 月,Harmattan AI 任何系统都没有公开可得的独立第三方测试报告、现场部署审计或安全事故披露。[CE026, CE039, CE040, CE041, CE042, CE046]
| 控制 / 认证 / 质量指标 | 状态 | 范围 | 缺口 / 尽调要求 |
|---|---|---|---|
| 人在回路(HITL)交战授权 | 已实施(公司声称) | 所有动能系统(Gobi、Gobi Tempest、Barkhan) | 预授权范围和接管延迟未公开记录 |
| 端到端加密和基于角色的访问控制(Kalahari) | 已实施(公司声称) | Kalahari C2 平台;所有接入 Kalahari 的资产 | 未发布加密标准;没有第三方安全审计或渗透测试 |
| NATO 标准互操作性(Asterix、CoT、Lattice、Sapient) | 自行声明;未获独立认证 | Kalahari 互操作层 | 未公开披露 STANAG 或 NATO 互操作认证 |
| GDPR / CCPA 数据隐私合规 | 已实施(隐私政策、使用条款;EU 托管基础设施) | 公司数据运营;Kalahari 数据织物 | 未覆盖保密作战数据处理或国防部数据主权要求 |
| 防务网络安全加固 | 公司自述;未引用标准 | Kalahari 和所有连接系统 | 没有 ISO 27001、CMMC 或同等认证;没有漏洞赏金或负责任披露计划 |
除非另有说明,所有合规状态均来自官方产品页和法律文件中的自行声明。截至 2026 年 6 月,尚未公开发布独立第三方认证。缺口列反映公开已知的证据缺失, 不代表已确认不合规。
[CE026, CE025]5.6 图表
06客户情况
6.1 客户基础分层与主要买方
截至 2026 年 6 月,Harmattan AI 的商业基础完全锚定在主权防务部门。公开记录中的主要付费客户是法国国防部(DGA)和英国国防部,二者各授予一个正式采购项目——这是防务领域最强的采购信号,意味着多年预算承诺和正式采办治理。2025 年 7 月,公司又宣布第三个未披露的 NATO 政府合同,使已确认正式采购项目总数达到三个。Dassault Aviation 是客户与投资人的混合体:2026 年 1 月 Series B 合作要求 Harmattan AI 为 Dassault 的 Rafale F5 和未来 UCAS 项目开发嵌入式 AI 能力,因此 Dassault 既是技术开发的付款方,也是进入更广泛空战采购的战略分销渠道。 Skyeton 合作(2025 年 10 月)引入渠道合作伙伴模式:Harmattan 的 Sahara SAR 雷达载荷被集成进 Skyeton 经实战验证的 Raybird UAS。该合作瞄准 Raybird 已有作战足迹的 NATO 市场,让 Harmattan 不必在每个市场单独赢得主权采购竞赛,也能间接触达盟国终端客户。Raybird 平台已随乌克兰防务部队累计超过 350,000 小时战斗飞行,为 Skyeton 渠道提供了相当强的运营可信度。 地理分层上,已确认足迹包括法国(国防部合同,加上 Dassault 总部和 Rafale 项目)、英国(国防部合同,加上新闻报道中的英国 Gobi 销售信号),以及通过 Skyeton 触达的更广泛、邻近乌克兰的 NATO 生态。公司出现在利雅得 World Defense Show(2026 年 2 月)、阿布扎比 UMEX(2026 年 1 月)、坦帕 SOF Week(2026 年 5 月)和巴黎 Eurosatory(2026 年 6 月),显示其有意进入海湾国家和美国 SOF 市场,但这些地区尚无公开宣布合同。客户基础完全由政府和机构组成;公司未披露任何商业或军民两用客户。[CU001, CU002, CU003, CU004, CU005, CU006]
| 客群 | 买方 / 用户 / 付款方 | 用例 | 规模(已知) | 收入 / 战略价值 | 关键缺口 |
|---|---|---|---|---|---|
| 法国武装部队 / DGA | 法国国防部(买方 / 付款方);法国陆军(用户) | ISR 训练无人机;Orion 2026 演习准备 | 1,000 台 Sonora | 列装项目;数百万 EUR 合同 | 具体合同金额未披露;交付确认未经独立验证 |
| 英国国防部 | 英国国防部(买方 / 付款方);英国陆军 / 国防部用户 | 面向紧急作战需求的自主系统 | 最多 3,000 台 | 列装项目;数百万 GBP;第二个主权级列装项目 | “最多 3,000”套系统的准确金额和构成未披露 |
| NATO / 盟国政府(未披露) | 盟国国防部(买方 / 付款方);盟国武装部队(用户) | AI 赋能自主无人机;数百万 USD 合同 | 数量未披露 | 首个列装项目于 July 2025 公告;DefenceTalks 称 >$10M | 客户身份、合同范围和交付时间表未公开 |
| Dassault Aviation(战略伙伴 / 客户) | Dassault(付款方);Armée de l'Air / UCAS 未来用户 | 面向 Rafale F5 和未来 UCAS 的嵌入式 AI;无人僚机控制 | 整个 Rafale F5 项目;未来 UCAS 机队 | 领投 $200M Series B;长期 AI 开发合作 | AI 集成仍处于预生产阶段;范围、费用和排他性未披露 |
| Skyeton / 乌克兰相关 NATO | Skyeton(渠道合作伙伴);盟国武装部队(最终用户) | Sahara SAR 雷达接入 Raybird UAS;ISR 任务 | Raybird 机队(350,000+ 飞行小时);NATO 市场部署 | 技术授权 / 集成合作 | 收入分成模式、单价和部署规模未公开 |
| 海湾国家主权买方(潜在) | GCC 国防部(潜在买方) | 防空、ISR、反无人机 | 仅处于展会阶段 | 无合同;只有战略意图(WDS Riyadh、UMEX Abu Dhabi) | 没有 LOI 或合同;出口许可要求未验证 |
| 美国特种作战(潜在) | SOCOM、盟军特战部队(潜在买方 / 用户) | 战术边缘训练、ISR、反无人机 | 仅处于展会阶段 | 无合同;只有战略意图(SOF Week Tampa) | 未确认 ITAR 清关;没有 LOI 或合同 |
收入 / 战略价值根据公开采购信号估算;所有客群的具体合同金额均未公开披露。缺口列反映尽调中最关键的未知项。
[CU001, CU002, CU003, CU004, CU005, CU009]Harmattan AI 的客户旅程,从主权政府展会亮相和采购信号,推进到竞争性资格评估、在册项目授标, 再到多机生产交付和由伙伴驱动的扩张循环。所有已确认付费客户都落在「生产——在册项目」或「开发合作」 阶段;海湾国家和美国 SOF 客户仍停留在发现阶段。
[CU001, CU004, CU009, CU013, CU021, CU039]6.2 具名客户证明与采用轨迹
Harmattan AI 的客户证据质量,在正式采购项目确认上很强,但在部署深度、使用指标和结果细节上偏薄。法国国防部合同(2025 年 7 月授予,计划 2025 年 10–12 月交付)覆盖 1,000 架 Sonora ISR 训练无人机,被描述为「数百万欧元」级别,定位是在 Orion 2026 多国演习前用于作战训练。专业防务新闻媒体 Army Recognition 在公告数日内报道了该合同;DGA 竞争流程要求约 20 项技术规格,并吸引多家欧洲竞争者,Harmattan 凭价格竞争力和量产准备度胜出。这是公开记录中独立佐证最充分的客户证明。 英国国防部合同(2025 年 9 月 DSEI London 宣布)覆盖最多 3,000 套自主系统,用于「紧急作战需求」。OpexNews.fr 和 Defense News 均独立报道该授标。「最多」措辞带来数量不确定性,准确 GBP 合同金额也未公开。公告时点是在 DSEI 这个英国旗舰防务展上,符合加速采购路径,也支持公司将其描述为量产部署而非试点项目。 Dassault Aviation 在 2026 年 1 月 Series B 公告中的公开声明确认了合作范围:在「主权、可控且受监测的 AI」原则下,为 Rafale F5 和 UCAS 开发 AI。这算是开发阶段的具名客户,而非生产部署。Skyeton 合作提供渠道伙伴证明:Skyeton CEO 在公告时公开背书合作,Raybird 平台随乌克兰部队积累的 350,000+ 飞行小时,也可作为 Harmattan 传感器集成进入运营验证机体的代理证据。不过,截至 2026 年 6 月,Harmattan Sahara 雷达自身在 Raybird 部署中的运营记录仍未获独立验证。 公司自报的产能吞吐说法——「每月交付数千套系统」——出现在 Dassault 新闻稿中,但未被任何独立审计机构、政府采购机关或调查媒体验证。该数字应按公司声称估计处理,而不是观察到的指标。[CU011, CU012, CU013, CU014, CU015, CU016]
| 指标 | 数值 | 日期 | 来源 | 置信度 | 含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| 首个列装项目 — NATO 政府 | 1,000 架 AI 赋能无人机;数百万 USD | 2025-07 | Army Recognition;Harmattan AI 新闻稿 | 高 | 成立 15 个月拿下首个主权级列装项目;验证 DGA 采购流程 | 准确 EUR 金额;交付确认;交付后使用率 |
| 第二个列装项目 — 英国国防部 | 最多 3,000 套自主系统;数百万 GBP | 2025-09 | OpexNews.fr、Defense News 与 Harmattan AI 来源 | 高 | 两个月内拿下第二个国家级列装项目;DSEI 公告显示英国作战需求紧迫 | 准确 GBP 金额;“最多”上限与实际承诺数量;交付节奏 |
| 公司声称的生产吞吐 | 每月数千套系统 | 2026-01 | Harmattan AI 新闻稿(Dassault 公告) | 低 | 暗示具备工业级制造规模;外部难以验证 | 独立审计;计入的具体系统类型;“已交付”的定义 |
| Skyeton 渠道合作伙伴 — Raybird 作战历史 | 350,000+ 战斗飞行小时(Raybird 平台) | 2025-10 | Skyeton 新闻稿;defence-industry.eu | 高 | 伙伴平台已经战场验证;Harmattan 传感器集成补上 SAR 能力 | Harmattan 传感器在作战小时中的占比;Sahara 专属战斗验证 |
| Eurosatory 2026 — 本土市场旗舰展会 | Hall 5B E430 展位;June 15–19, 2026 | 2026-06 | Harmattan AI 博客 | 高 | 欧洲盟友需求信号;欧洲最大陆军展 | 政府会谈次数;形成的 LOI;后续合同 |
| SOF Week 2026 — 进入美国市场信号 | Level 3-2055 展位;May 18–21, 2026;佛罗里达州坦帕 | 2026-05 | Harmattan AI 博客 | 高 | 首次确认在美国参展;接触 SOF 社群 | 美国采购兴趣;ITAR 状态;SOCOM 关系深度 |
数据来自新闻稿和专业防务媒体。“每月数千套系统”是公司说法,未经独立审计。null 值表示公开数据缺失。
[CU011, CU012, CU013, CU014, CU015, CU016]| 客户 | 客群 | 部署 / 用例 | 量产 / 试点 | 结果 | 限制 |
|---|---|---|---|---|---|
| 法国武装部队(DGA) | 政府 / 军方;法国 | Sonora ISR 训练无人机,用于单位级训练和 Orion 2026 演习备训 | 量产 — 正式列装项目(Jul 2025) | 已订购 1,000 架;交付 Oct–Dec 2025;在 DGA 竞争性采购中击败欧洲对手 | 尚无独立来源确认交付完成;未披露交付后使用情况;没有操作员层面的证言 |
| 英国国防部 | 政府 / 军方;英国 | 最多 3,000 套自主系统,用于紧急作战需求 | 量产 — 正式列装项目(Sep 2025) | 18 个月内第二个国家级 POR;DSEI 公布;合同金额达数百万英镑 | 未披露确切数量、构成和交付时间表;“最多”上限带来收入不确定性 |
| NATO / 盟国政府(未披露) | 政府 / 军方;盟国 | AI 驱动的自主无人机系统;金额达数百万美元 | 量产 — 正式列装项目(Jul 2025) | 首次公开宣布 POR;第三方报道称合同金额 >$10M;Dassault 新闻稿确认 | 客户身份涉密;项目范围和结果未公开;无法独立验证 |
| Dassault Aviation | 航空主承包商伙伴 / 战略客户;法国 | 为 Rafale F5 和未来 UCAS 开发 AI;无人僚机 AI 和控制算法 | 开发合作(Jan 2026);量产前 | 领投 $200M Series B;Dassault CEO 公开背书合作;法国总统 Macron 称赞该公告 | AI 集成仍处量产前阶段;目标系统(Rafale F5)计划约 2030 年上线;尚无已部署单位 |
| Skyeton(乌克兰 / 欧洲渠道) | 渠道合作伙伴 / 间接客户;乌克兰 + NATO | 将 Sahara SAR 雷达集成到 Raybird UAS;用于侦察和边境监控的 ISR | 活跃开发合作(Oct 2025);报道预计 Q1 2026 首次部署 | Skyeton CEO 公开背书合作;militarnyi.com 报道乌克兰无人机采用 Harmattan SAR 雷达;drone-actu.fr 确认合作 | Q1 2026 之后,Harmattan 的 Sahara 雷达是否随 Raybird 投入作战部署尚未获得独立确认 |
按证据质量排序。“量产”指已在正式列装项目(Program of Record)下交付系统;“开发” 指量产前 AI 集成合作协议。每行来源数代表找到的独立确认来源数量。
[CU011, CU013, CU015, CU017, CU019, CU020]主权政府兴趣面广,但收窄到 3 个已确认在册项目和 2 项战略合作。尚未确认商业或非主权客户群。 漏斗显示资格评估到成交的转化率高(公司称两项初始 PoR 均击败多家欧洲竞争者),但交付后可见度薄弱。
[CU011, CU013, CU015, CU017, CU021, CU037]6.3 留存、耐久性与扩张路径
Harmattan AI 的客户耐久性几乎完全建立在正式采购项目结构上,而不是可公开验证的留存数据。法国和英国的正式采购项目意味着多年预算授权和部长级承诺;相比单年度试点或可自由裁量预算项,这类项目结构上更难取消。2026 年 1 月 Dassault Aviation 新闻稿明确称 Harmattan 已被法国和英国国防部「授予多个正式采购项目」,用复数确认两个合同都具备这一称谓。Defense News、TechCrunch 和 AirForce-Technology.com 同日独立复述了这一说法,佐证它更接近事实表述,而不是营销话术。 公司未公开披露 NRR、GRR、流失率、合同续约或复购数据。法国国防部最初交付窗口(2025 年 10–12 月)和英国合同授标(2025 年 9 月)意味着,截至 2026 年 6 月,部分项目已过初始交付期;但 Harmattan 未发布任何追加订单公告、使用报告或客户满意度披露。这造成重大证据缺口:投资人无法独立验证这些项目是否带来复购、扩大部署,或来自客户采购人员的任何可量化运营结果背书。 扩张路径多方向,但结构上仍早期。Dassault Aviation 背书打开了 Rafale 项目国家(Qatar、India、Egypt、Indonesia、UAE、Greece、Croatia、Serbia、Colombia、Indonesia)这一批潜在二级市场,路径是主承包商分销网络。Skyeton 合作同样打开乌克兰防务工业基础,以及采购 Raybird 的 NATO 盟友,使 Harmattan 不必赢得主权采购竞赛,也能延伸传感器收入。公司计划进入的新领域——电子战、打击平台(Barkhan)和 C2 软件(Kalahari)——如果达到正式采购项目状态,也给现有法国和英国国防部客户创造交叉销售机会。[CU029, CU030, CU031, CU032, CU033, CU034]
| 指标 | 数值 / 状态 | 客群 | 置信度 | 尽调询问事项 |
|---|---|---|---|---|
| 净留存率(NRR) | null — 未披露 | 所有客户 | 低 | 向公司索取;可能受国防部 NDA 约束而保密 |
| 总留存率(GRR) | null — 未披露 | 所有客户 | 低 | 核实是否发生项目取消、范围缩减或合同终止 |
| 合同续签 / 后续订单(法国国防部) | null — 初始交付窗口为 Oct–Dec 2025;截至 Jun 2026 尚未宣布后续订单 | 法国国防部 | 低 | 向 DGA 索取交付确认和后续采购信号;检查 DGA 采购数据库 |
| 合同续签 / 后续订单(英国国防部) | null — 交付时间线未公开;截至 Jun 2026 尚未宣布后续订单 | 英国国防部 | 低 | 向英国采购办公室索取交付节奏和未来批次启动情况 |
| 正式列装项目(POR)结构性耐久度(代理指标) | 正向代理指标 — 法国和英国两份合同均带 POR 认定,意味着多年预算授权 | 法国国防部和英国国防部 | 中 | 确认合同期限、选择权期限以及任何便利终止条款 |
| 操作员层面满意度 / NPS | null — 终端用户(士兵、飞行员、操作员)没有公开证言 | 法国陆军;英国陆军 | 低 | 向 DGA / 英国陆军采购对口方索取支持信、用户反馈报告或操作员证言 |
所有 NRR/GRR/流失字段均为 null;国防合同留存未公开披露。正式列装项目(Program of Record)认定可作为结构性耐久度的代理指标,但不是财务指标。
[CU029, CU030, CU032, CU043, CU044]| 扩张驱动因素 | 集中度风险 | 严重性 | 影响 | 尽调路径 |
|---|---|---|---|---|
| Dassault Aviation 作为国际分销渠道 | 依赖:如果合作条款限制 Harmattan 直接向 Rafale 国家客户销售,集中度会加剧 | 高 | Dassault 合作打开卡塔尔、印度、埃及、阿联酋及约 8 个其他 Rafale 国家,但可能形成独家限制 | 索取合作合同范围;核实分销独家权是否适用于特定地域或领域 |
| 两个主权客户收入集中(法国 + 英国国防部) | 严重:公开披露硬件收入的 100% 来自两个西欧政府 | 严重 | 单个项目取消或预算削减可能抹去 40-60% 收入;目前没有商业收入对冲 | 要求披露客户收入集中度;核实是否有任何单一客户超过总 ARR 的 25% 或 50% |
| Skyeton 渠道 — 间接进入 NATO 市场 | 中:渠道合作伙伴带来间接触达,但也让 Harmattan 依赖 Skyeton 在 NATO 市场的采购成功 | 中 | 部署取决于盟国政府采购 Raybird;Harmattan 收入是间接的,并与 Skyeton 销量成比例 | 确认 MoU 条款;核实 Harmattan 收取单台授权费还是固定技术费;跟踪 Raybird 采购管线 |
| 电子战与打击领域扩张 | 低-中:在没有客户 LOI 确认的情况下扩张到 EW 和打击,带来研发资本风险 | 中 | 若成功可分散收入;若政府客户不规模化采购新品,则有烧钱风险 | 核实开发阶段,以及 DGA 或英国国防部是否已为 EW 或 Barkhan 项目出具任何 LOI |
| 进入中东和海湾主权市场 | 中-高:参展尚未转化为合同;出口许可未核实 | 中 | 可分散收入,但地缘政治敏感;适用两用物项分类和法国出口法 | 确认 AECA/ITAR 状态、DGA 出口批准,以及是否有任何 GCC 政府发起正式采购询价 |
集中度风险评估仅基于公开合同披露。扩张路径阶段反映参展和合作结构,不代表已签合同。
[CU033, CU034, CU035, CU036, CU041, CU042]在 5 个具名客户中,两个国防部在册项目的生产证据最强;结果具体性和续约可见度整体偏低, 更多反映防务行业不透明,而非采用弱。Dassault Aviation 因 CEO 级背书和法国总统验证, 参考质量得分较高。
[CU014, CU017, CU020, CU022, CU026]6.4 集中度风险与反向证据
Harmattan AI 客户画像里最显著的结构弱点,是客户极度集中。所有公开具名的创收客户都是西欧主权政府:法国和英国。这两个客户按定义接近已披露商业收入的 100%。考虑到 Sonora 合同是 Harmattan 第一个合同,且交付时间线记录最充分,仅法国政府就可能贡献硬件收入的大部分。任何一个政府的单项政策变化——采购原则转向、防务预算削减、项目要求变化,或涉及自主系统的安全事件——都可能实质损害 Harmattan 的收入。 留存和重复使用数据不透明,进一步放大该风险。SaaS 公司会披露 NRR,商业防务主承包商会公布积压订单;Harmattan AI 没有披露合同续约选项、交付确认、交付后使用情况,或任何可让投资人区分「刚赢得首批合同的公司」和「合同正在形成真实重复商业关系的公司」的指标。这一部分是结构性的——防务采购数据常常涉密或具商业敏感性——但在当前阶段,它让耐久性无法被独立验证。 Reuters 2025 年 11 月关于法国无人机制造商面临融资不确定性的报道,凸显更广泛行业风险:即使有政府支持,法国防务初创公司也可能遇到供应链瓶颈、产能爬坡困难和采购预算延迟,进而影响交付时间线和客户关系。文章提到,在投入成本波动中扩大生产,是整个行业的普遍挑战——Harmattan 已在内部承认这一风险,但未公开量化。Dronexl 对 Dassault 交易的分析也指出,$1.4 billion 估值假设政府合同流会持续,但这一点尚未通过第二轮采购周期独立验证。[CU041, CU042, CU043, CU044, CU045, CU046]
Harmattan AI 的任何合同都没有公开续约百分比数据。矩阵展示各项目在 6 个月、12 个月和 18 个月节点的已知活跃状态。 「进行中」= 项目仍在推进 / 未见取消报道;「未知」= 超出当前观察窗口或未被公开确认。图表类型 从 cohort 改为 matrix,正是为了反映这个数据缺口。
[CU029, CU030, CU043, CU044]07风险
7.1 监管、法律与出口管制风险
Harmattan AI 处在监管最严的商业赛道之一:向政府军方出售致命或潜在致命的自主系统。几条监管线叠加,形成多层风险暴露。 法国两用出口管制:所有从法国出口的无人机和自主武器,都受 SGDSN 与 DGA 依据防务物资出口法(Loi du 22 juin 2011)执行的法国出口许可制度约束。非军用两用子系统还并行适用 EU Regulation 2021/821,即欧盟两用出口管制框架。Expert Market Research 2026 Europe UAV report 明确称,「欧洲 UAV 出口需要在国家出口许可与欧盟共同立场框架之间协调,这让面向全球市场的欧洲制造商面临监管复杂性。」Harmattan CEO 在 2026 年 2 月 Le Grand Continent 采访中承认,要大批量生产,供应链能否摆脱出口许可束缚是实质约束。 NATO 与 EU 互操作认证:NATO 的技术互操作演习(TIE 26)于 2026 年 5 月在荷兰举行,用来测试 C-UAS 与 UAS 系统是否满足联盟 C2 集成标准,只有通过后才有资格接入 NATO 防务架构。NCIA 表示,「现代多层 C-UAS 行动中的互操作性不再是可选项,而是必需项。」公开资料没有证实 Harmattan 系统参加了 TIE 26。NATO 的 Latvia Innovation Range(2026 年 3 月)和 LCI-X Beacon Project,都要求系统先经过结构化验证,才可能被全联盟采购。若拿不到 NATO 互操作认证,Harmattan 只能留在与法国、英国的双边国家合同里,可触达市场会被封顶。European Parliament Research Service 2026 年对军用无人机系统的分析也强调,在 EU 法律框架下部署自主系统监管复杂;两用无人机平台同时受军方和民航主管机构监督。 致命性自主武器与 IHL 合规:Harmattan 的 Gobi 反无人机拦截器可在「每次拦截无需人在环」的情况下运行。这直接落入国际社会关于致命性自主武器系统(LAWS)的持续争论。UN Convention on Certain Conventional Weapons 的政府专家组自 2014 年以来一直讨论 LAWS 监管,但尚未形成有约束力的条约。Gobi 没有公开披露 AI 伦理章程、安全认证或 IHL 合规文件。一旦发生误识别事件,国防部可能暂停项目,监管也可能介入,并伤害 Harmattan 在 NATO 客户中的声誉。 EU 碎片化加大互操作不确定性:法国与德国在 FCAS 联合战斗机项目上仍僵持,Defense News 称双方「难以前进」。按 CEO 自己的判断,European Commission 也没有制定泛欧防务战略的授权。Eurocontrol 2024 C-UAS workshop 证实,EU 无人机监管在民用 / 两用空域采用 EASA 框架,但军用系统由各国规则单独管理,两用系统的合规不确定性因此上升。 隐私与数据法:Harmattan 隐私政策(2025 年 11 月)称,个人数据可依据 Standard Contractual Clauses 转移至 EEA 以外。国家安全例外下,ISR 无人机外场采集的影像数据如何治理,公开资料没有说明。[CR001, CR002, CR003, CR004, CR005, CR006]
| 风险 / 规则 / 许可证 | 司法辖区 | 状态 | 可能性 | 严重性 | 缓释措施 | 剩余敞口 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| Gobi 无 HITL 自主拦截 — LAWS / IHL 合规缺口 | 国际 / 法国 / 英国 | 未处理 — 未发布 AI 伦理章程或 IHL 合规声明 | 中 | 严重 | 部署受国家 ROE 约束;法国尚未采纳 LAWS 暂停令 | 高 — 英国国防部可能在后续合同中施加 HITL 要求 | 索取 Harmattan AI 安全文档,并确认英国国防部 HITL 政策立场 |
| 法国两用物项出口管制许可(SGDSN / DGA) | 法国 / 欧盟 | 持续存在 — CEO 确认出口许可依赖是实质性生产约束 | 高 | 高 | 法国垂直一体化制造降低美国 ITAR 适用性 | 高 — 海湾 / APAC 出口需要逐案批准 | 索取英国国防部项目及计划中非欧盟出口市场的许可状态 |
| 欧盟两用物项法规 2021/821 | 欧盟 | 适用 — 自主系统和 SAR 雷达属于两用物项 | 中 | 高 | 法国国防部合同依政府间条款豁免;商业出口必须获许可 | 中 — 合规工作量随出口量扩大 | 确认 Harmattan 欧盟两用物项分类和许可流程 |
| NATO C-UAS 互操作认证(TIE 26 / LCI-X) | NATO / 多国 | 未定 — 截至 June 2026,尚未确认 Harmattan 参加 TIE 26 | 高 | 高 | 法英双边合同绕过 NATO 认证;TIE 26 是在联盟范围销售的前置条件 | 高 — 除非获得认证,否则 TAM 受限于双边市场 | 确认 TIE 26 / LCI-X 参与时间线;索取 NATO 互操作路线图 |
| GDPR / CCPA/CPRA(网站隐私政策) | 欧盟 / 加州 | 合规 — 隐私政策已就位,November 2025 更新 | 低 | 中 | 欧盟托管 CRM;跨境数据传输采用 SCCs | 低 — 仅限网站范围;战场 ISR 数据治理仍有缺口 | 索取 ISR 现场采集数据的数据治理框架 |
| 欧盟 / EASA 关于 ISR 无人机部署的两用民用空域监管 | 欧盟 | 不明确 — Eurocontrol/JRC 2024 研讨会确认,军事系统适用单独的国家规则 | 低 | 中 | 军事 ISR 无人机合同在国家国防部框架下运行;民用 EASA 规则不适用 | 中 — 两用边境 / CNI 应用可能适用 EASA 规则,存在风险 | 确认 Harmattan 系统任何计划中民用 / CNI 部署的监管框架 |
按严重性排序(严重 → 中)。状态反映截至 June 2026 的公开信息;未通过法院数据库确认任何诉讼或执法记录。
[CR001, CR002, CR003, CR004, CR006, CR007]Harmattan AI 主要风险类别的发生可能性(列:低 / 中 / 高)与剩余影响(行:低 → 严重)。 单元格标签展示风险和缓释成熟度。
可能性和影响是基于公开证据的定性评估;并非来自精算或概率模型。
[CR004, CR005, CR013, CR035, CR036, CR040]7.2 供应链、制造与运营风险
Harmattan AI 已确认两个交付窗口紧的正式列装项目:向法国国防部交付 1,000 架 SORONA 无人机(2025 年底前交付),以及根据 2025 年 9 月合同向英国国防部交付最多 3,000 套系统、用于近期部署。同时,Series B 资金要支撑制造扩产。BCG 2026 年 5 月 C-UAS 白皮书直接点名这一品类:「C-UAS 系统仍然昂贵、难以规模化,并依赖尚未为持续消耗而建好的供应链。」Expert Market Research 2026 Europe UAV report 认为,欧洲无人机行业在自主能力、AI 集成和产能规模上「落后美国和以色列同类 5–10 年」。 部件与稀土依赖:无人机硬件——包括 LiPo 电池、微电子、RF 模块和电机组件——很大程度依赖中国制造和稀土材料。Group 1 UAS 的锂电池供应链仍高度集中在中国厂商。无人机电机所用稀土元素(钕、镝)也会周期性受到中国出口管制。Harmattan 采用 Lynred 供应的 IR 传感器(法国公司),说明部分零部件有主权来源,但电池、电机、处理器或 RF 部件的采购来源没有公开披露。SORONA 无人机被描述为「在法国组装」,但最终组装不等于全链条主权可控。 制造爬坡与质量风险:月产量从数百台拉到数千台,需要大量资本和流程能力。公司没有披露制造设施、产能指标或质量管理认证(例如 AS9100、DEF STAN)。DroneXL 报道 Harmattan 正规划每月 10,000 架无人机产能;这个规模需要完整的供应链、测试和质量基础设施。Reuters 2026 年 6 月报道,Anduril 的无人机坠毁和安全事件说明,国防无人机快速扩产会带来真实运营安全风险,这是行业层面的模式。 电子战与 GPS 拒止韧性:Harmattan CEO 强调,战场上可能出现 GNSS 不可用或 RF 通信被压制的场景。Gobi 拦截器与 SORONA 依赖嵌入式 AI 实现不依赖 GPS 的飞行。MarketsandMarkets 2026 Europe UAV report 将 GPS 韧性列为设计要求,并称这会「增加开发复杂度和成本」。Harmattan 任何系统都没有公开红队、软件审计或渗透测试披露。[CR011, CR012, CR013, CR014, CR015, CR016]
| 失效模式 | 可能性 | 严重性 | 缓释成熟度 | 剩余敞口 | 未解决缺口 |
|---|---|---|---|---|---|
| 中国零部件供应中断(电池、稀土、微电子) | 中 | 严重 | 低 — 未披露替代采购或库存策略 | 高 — 影响所有无人机产线 | 未公开披露电池 / 电机 / RF 零部件来源;稀土敞口未量化 |
| 国防部交付进度失误(4,000+ 台,2025–2026 窗口) | 中 | 高 | 低 — 未发布制造产能证据 | 高 — 合同罚款、项目取消、声誉受损 | 未披露生产设施、产能指标或外部制造合作 |
| Gobi 自主拦截误识别(友军误伤或平民事件) | 低-中 | 严重 | 低 — 未公开红队、安全测试或 IHL 合规审计 | 高 — 声誉、法律和项目终止风险 | 无 AI 安全认证、偏差审计或交战规则技术披露 |
| 针对自主制导系统的网络攻击 / 对抗性 AI 欺骗 | 中 | 高 | 低 — 未披露渗透测试或网络安全认证 | 高 — GPS 拒止环境扩大攻击面 | 未公开网络安全态势或渗透测试披露;GNSS 欺骗可能改向无人机 |
| 争夺频谱中的 EW 系统性能下降 | 中 | 中 | 中 — CEO 称 EW 是重点研发方向;正在收集乌克兰数据 | 中 — EW 性能尚未在受控测试条件下验证 | 未发布独立 EW 性能数据;仅依赖乌克兰作战数据 |
| 扩产质量缺陷使现场故障率超过可接受阈值 | 中 | 中 | 低 — 未披露 MIL-SPEC 或 AS9100 资质 | 中 — 交付缺陷触发检验暂停和返工成本 | 未公开披露国防部合同中的质量管理标准或损耗率阈值 |
按严重性排序。供应链来源不透明,以及 Gobi 无 HITL 拦截架构,是主要未缓释运营风险;尚未公开披露制造审计或质量认证。
[CR011, CR013, CR014, CR015, CR016, CR017]7.3 合作伙伴、客户与依赖风险
Harmattan AI 的商业和战略位置在结构上依赖少数交易对手,形成很高的双边集中风险。 Dassault Aviation 作为领投方带来双重忠诚张力:Dassault 同时是 Series B 领投方、战略伙伴,也是 Rafale F5 与 UCAS 平台嵌入式 AI 的潜在未来客户。如果关系因商业分歧、IP 争议或 Dassault 收购竞争性 AI 能力而恶化,Harmattan 会同时失去资本来源、合作叙事和锚定合同。Defense News 提到,Dassault CEO Trappier 曾表示,如有必要,法国可以「独自」开发下一代战斗机。Harmattan CTO 在 Series B 时写道,「国际秩序正在脱轨」;这种表述可能给美国或海湾市场开发添阻力。 客户集中:所有已披露的收入客户都是政府实体,且只有法国国防部和英国国防部。截至 2026 年 6 月,没有公开资料证实公司已经拓展到其他 NATO 政府、非 NATO 盟友或商业细分市场。两个项目下没有公开报道交付失败或合同履约问题,但客户只集中在两家仍是结构性风险。 Skyeton(乌克兰)依赖:2025 年 10 月与 Skyeton Raybird 的合作,把 Harmattan 的 Sahara SAR 集成进去,带来地缘政治敏感性。俄罗斯对乌克兰工业设施的打击可能扰乱 Skyeton 运营。将搭载 Harmattan 技术的 Raybird 再出口到第三国,也会增加出口许可复杂度。 技术伙伴集中:Lynred(IR 传感器)是 SORONA 传感器栈的单一来源依赖。BCG 2026 C-UAS 白皮书强调,欧洲 C-UAS 供应链整体工业规模有限,替换供应商既困难又昂贵。[CR021, CR022, CR023, CR024, CR025, CR026]
| 依赖 | 交易对手 | 角色 | 集中度 | 失效情景 | 严重性 | 缓释措施 | 剩余敞口 |
|---|---|---|---|---|---|---|---|
| Series B 领投方和战略 AI 合作伙伴 | Dassault Aviation | 资本提供方、Rafale F5 AI 与 UCAS 合作伙伴、未来客户 | 极高 — 单一领投方兼主要战略伙伴 | Dassault 将 AI 能力内部化,或合作因 IP 争议解体 | 严重 | 合作协议和 Series B 投资让短期激励一致 | 高 — 未披露次级战略伙伴或备用领投方 |
| 法国国防部 SORONA 1,000 台合同 | 法国国防部 / DGA | 锚定客户;首个正式列装项目(POR) | 极高 — 唯一确认的法国收入来源 | 不续约、履约争议或失去后续招标 | 高 | 2025 年底前交付提供验证点;采购通道简化 | 高 — 截至 June 2026 未披露商业多元化 |
| 英国国防部 3,000 套自主系统合同 | 英国国防部 | 第二个锚定客户;第二个正式列装项目(POR) | 极高 — 唯一确认的英国收入来源 | 交付失败、HITL 政策变化或冲突后采购常态化 | 高 | 紧急作战需求提供短期保护;未确认长期框架合同 | 高 — 采购常态化可能要求与主承包商重新竞争 |
| 乌克兰 UAS 集成合作伙伴 | Skyeton | Sahara SAR 集成的 Raybird 平台合作伙伴 | 中 — 多条产品线之一;Harmattan 核心业务独立 | Skyeton 运营中断;再出口许可复杂化 | 中 | Harmattan 核心 ISR / C-UAS 产品不依赖 Skyeton | 中 — 地缘政治波动增加 SAR 部署时间线不确定性 |
| IR 传感器独家供应商 | Lynred(法国) | SORONA 四旋翼的单一来源 IR 传感器 | 高 — 关键传感器单一供应商 | 生产压力下 Lynred 产能受限、质量问题或涨价 | 中 | Lynred 是法国主权供应商;相较中国替代方案,降低地缘政治供应风险 | 中 — 替代传感器认证需要 DGA 重新批准,并推迟时间线 |
按严重性排序。客户集中在两个政府交易对手,是主要商业风险。除 Lynred 外,所有条目都没有披露的次级关系缓释。
[CR021, CR022, CR023, CR024, CR025, CR043]Harmattan AI 的关键交易对手和供应链依赖;节点色调表示中断潜在严重程度。
[CR021, CR022, CR024, CR025]7.4 人员、治理与执行风险
Harmattan AI 不到两年扩张到约 130 名员工;据 Defense News,公司称员工经验中位数为 15 年。这种罕见的人才密度,反映公司有意从 Safran、Isar Aerospace 和相邻防务公司招聘资深人员。但关键人物集中度很高:Mouad M'Ghari(CEO)是所有主要新闻稿、投资人公告和 Le Grand Continent CEO 访谈中唯一具名高管。公司没有公开任命 CFO、COO、CRO 或运营负责人。Martin de Gourcuff(CTO)只在技术联合创始人语境中被提及。 治理不透明:Harmattan AI 没有披露外部董事会构成。Dassault Aviation 作为 Series B 领投方大概率拥有董事会席位,但未确认。FirstMark Capital、Atlantic VC 和 Tholus Capital 的投资人董事席位权利也未披露。一家国防独角兽在没有披露董事会监督的情况下运营,治理风险偏高,产品敏感性又放大了这一点。截至 2026 年 6 月,公开招聘岗位包括 Chief Engineer、Electronic Warfare Team Lead 和 System Architect,这些高级岗位仍未填补。 人才竞争:Harmattan 争夺 AI、EW 和自主系统工程师时,面对 Anduril($61 B 估值、最新一轮 $5 B)、Shield AI($12.7 B、最新一轮 $1.5 B)、Helsing(据报道正以约 ~$18 B 估值融资 $1.2 B)以及 Quantum Systems(已融资 €310 M)。这些同行能提供明显更高薪酬、更成熟职业路径和更大运营规模。Helsing 的 $600 M 融资专门瞄准乌克兰无人机制造扩产,直接争抢 EW 和 AI 无人机人才。 执行负荷:一家 130 人公司要同时交付两个国防部项目、扩制造、延伸到 EW 与 Rafale F5 AI、管理 Skyeton 集成,并参加 Eurosatory 2026 和 SOF Week 2026,执行负荷很重。公司没有公开项目管理框架或独立质量保证披露。[CR027, CR028, CR029, CR030, CR031, CR032]
| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| CEO(Mouad M'Ghari) | 所有新闻和投资者沟通中唯一具名高管;政府关系的主要负责人 | 低-中 | 高 | 联合创始团队提供一定冗余;Dassault 合作分散关系风险 | 索取 CEO 接班计划、雇佣协议和股权锁定条款 |
| CTO 角色(Martin de Gourcuff) | 技术语境中唯一提到的另一位联合创始人;EW 和 AI 架构负责人 | 低-中 | 高 | 来自 Safran 和 Isar Aerospace 的资深工程师在联合创始人之下提供技术纵深 | 确认技术领导层梯队图和二级工程管理层 |
| 外部董事会监督 | 未披露董事会构成;投资方董事席位权利未确认 | 高 | 高 | Dassault Aviation 作为 Series B 领投方推测持有董事席位;未确认 | 索取董事会构成、董事会纪要访问权和独立董事资历 |
| CFO / 财务负责人 | 未公开具名 CFO;未披露经审计财务报表 | 高 | 高 | Series B 尽调推测包含财务审阅;未公开确认 | 索取 CFO 姓名和任期,以及最近一期经审计财务报表 |
| 国防工程人才管线 | 与 Anduril、Helsing、Shield AI 争夺 AI、电子战(EW)和自主系统工程师 | 高 | 高 | 股权、使命感和法国国家冠军地位有差异化;薪酬基准未知 | 核查开放岗位填补率、招聘周期和高级工程师流失率 |
| 制造运营负责人 | 产量爬坡到每月数千台,需要有经验的生产管理 | 中 | 中 | 已披露从 Safran 航空供应链招聘;运营团队完整度未知 | 确认运营副总裁或制造负责人已到岗;索取组织架构图 |
按严重性排序。董事会构成、CFO 缺位和人才管线竞争,是主要未解决的执行-治理缺口。
[CR027, CR028, CR029, CR030, CR031]7.5 竞争、财务与采购周期风险
Harmattan AI 的 $1.4 billion 估值和约 ~$242 million 累计融资,明显低于最直接的同行。Anduril 2026 年 5 月完成 $5 billion Series H,估值 $61 billion,2025 年收入翻倍至 $2.2 billion。Shield AI 2026 年 3 月累计融资 $2 billion,估值 $12.7 billion。Helsing 据报道正以 $18 billion 估值融资 $1.2 billion。Quantum Systems 累计融资 €310 million,并报告年收入增长 100%+。Tekever 从 Baillie Gifford 与 NATO Innovation Fund 融资 €70 million,且已经盈利。正如 Anduril CEO 在 Series H 时所说:「国防过去不是吸引大量风险投资的品类。现在已经发生了实质变化。」资本涌入让这个赛道在合同、人才和战略伙伴上都极度拥挤。 战时紧迫性导致采购周期集中:Harmattan 的两个正式列装项目都通过简化或紧急采购机制授予——法国 DGA 发布的征集只有 20 项技术要求,英国国防部则引用「紧急作战需求」。在当前环境下,压缩采购通道是竞争优势;但如果地缘政治缓和、紧迫性下降,它会变成负担。冲突后采购可能回到完整生命周期评估,届时拥有几十年合规基础设施的传统主承包商会有结构优势。 财务不透明与资本强度:Harmattan 没有披露收入、毛利率、烧钱速度或盈利路径。国防硬件业务资本强度高:营运资金需求包括制造、库存、政府应收款(通常 90–180 天)和保修义务。$200 million Series B 正用于制造扩产和领域扩张(EW、Rafale F5 AI)。如果采购收入确认延迟——政府合同中很常见——公司可能在达到现金流盈亏平衡前遭遇营运资金挤压。 EU 碎片化风险:BCG 2026 C-UAS 报告称,欧洲格局「碎片化,国家项目并行、R&D 重叠、工业规模有限」。CEO 也承认 European Commission 没有防务战略授权,欧洲战略自主行动「仍处于起步阶段」。这种碎片化限制了大型泛欧采购工具的出现,而这类工具本可让 Harmattan 超越双边合同实现规模化。[CR033, CR034, CR035, CR036, CR037, CR038]
| 风险 | 可监控触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| LAWS / 自主武器监管 | 英国 / 法国出台国家级人在回路(HITL)强制要求,或 UN CCW 就 LAWS 达成有约束力条约 | 任何覆盖 Gobi 拦截场景的有约束力 HITL 要求,或 Harmattan 承认存在监管约束 | 除非 Gobi 修改,否则投资逻辑破裂;继续投资前要求给出 HITL 路线图 |
| 中国供应链中断 | 中国限制稀土元素或锂聚合物电池出口 | 限制覆盖钕、LiPo 电芯或 MCU,影响无人机 BOM 的 >20% | 必须立即分散供应;对法国 / 英国国防部的交付排期承压 |
| 国防部合同交付失败 | DGA 或英国国防部正式通知不合规或暂停排期 | 任何公开通知显示合同延期、质量暂停或交付数量低于承诺排期 | 执行可信度投资逻辑破裂;增加敞口前先做根因调查 |
| Dassault Aviation 合作破裂 | Dassault 公开表示 AI 内部化,或 Harmattan 发布合作重组新闻稿 | Dassault 收购竞争性 AI 能力,或 Harmattan 停用 Rafale F5 合作表述 | 重新评估投资逻辑;Dassault 关系是进入作战航空市场的核心 |
| 竞争对手赢下法国或英国核心细分市场 | Anduril、Helsing 或 Shield AI 赢得法国或英国国防部 ISR、C-UAS 或战术无人机合同 | 任何披露的竞争对手在 Harmattan 主要市场细分中的中标 | 密切监控;市场份额被侵蚀,会给未来收入假设发出下行信号 |
| 治理红旗 | 联合创始人离职,或未披露关联方交易 | 任何联合创始人退出、AMF 调查或重大未披露投资者关系 | 立即升级尽调;联合创始人集中度高,领导层变化高度重要 |
| NATO 认证失败 | Harmattan 未出现在 TIE 26 / LCI-X 参与者名单或目录中 | 到 2027 年底,NATO 发布的任何互操作目录中都没有 Harmattan 系统 | 可触达市场收缩到双边合同;重新评估 TAM 假设 |
终止标准代表投资逻辑失效,需要立即采取组合层面的行动。监控触发项按季度检查。
[CR033, CR035, CR036, CR039, CR040, CR041]主要风险节点及其向收入集中度、市场准入、人才和投资论点传导的下游路径。
[CR023, CR024, CR031, CR033, CR034, CR037]7.6 图表与附录
08估值
8.1 投资逻辑与反向逻辑
Harmattan AI 的乐观逻辑建立在四个相互叠加的支柱上。第一,结构性需求:欧洲 C-UAS 与自主系统市场年增速超过 50%,BCG 预计到 2029 年达到 $5–7 billion;已承诺的国防部支出将从 2025 年的 $600 million 增至 2029 年的 $800 million,基线之上还有大量未分配防务支出。Harmattan 的垂直整合模式覆盖 ISR 无人机、反无人机拦截器、电子战平台、打击 UAS 和 Kalahari C2 平台,使其能够跨四个细分承接需求,窄型竞争对手做不到。第二,主权侧验证:公司成立 15 个月内拿下两个国家国防部正式列装项目(法国 2025 年 7 月、英国 2025 年 9 月),在欧洲防务采购里执行速度领先同龄同行,也建立了纯技术同行在相近年龄缺少的经常性收入基础。第三,Dassault Aviation 护城河:2026 年 1 月战略合作形成双向锁定——Harmattan AI 为 Rafale F5 标准和未来 UCAS 提供嵌入式 AI,Dassault 则提供系统架构深度、任务系统集成经验和国际市场通道,大幅缩小可竞争者集合。第四,防务 AI 的长期顺风:European Parliament Research Service 在 2026 年研究中确认,军用无人机系统已从实验走向作战基线;NATO 2026 年 5 月 TIE 26 演习测试了 60+ 项 C-UAS 技术的互操作性,形成一条更有利于 Harmattan 这类已有合同背书供应商的采购管线。 反向逻辑同样关键。收入不透明是影响确信度的第一约束:Harmattan 没有公开披露 ARR、合同总价值、毛利率、烧钱速度或单位经济模型。因此,$1.4 billion 估值完全定价在战略溢价、合同证明和同行倍数上,而不是财务模型输入。没有收入披露,市场报道的估值隐含了无法压力测试的收入倍数;一旦合同延迟、重谈,或 Rafale F5 时间线滑后,倍数可能快速收缩。供应链脆弱性是 CEO 承认过的风险:无刷电机磁体的中国稀土依赖和电池化学体系依赖仍在,电池主权替代方案还要 3–5 年。在中国出口管制收紧的情景下,产能目标会更像愿景,而不是可执行计划。治理不透明是第三个反向因素:董事会构成未公开披露,关键人物集中在 CEO Mouad M'Ghari 和 CTO Martin de Gourcuff 身上,且没有公开章程说明投资人权利或股东保护。最后,防务科技整体倍数压缩风险是真实的:Reuters 2026 年 6 月报道,一家估值 $1.3 billion 的硅谷防务创业公司因无人机坠毁和安全事件陷入可靠性危机;这提醒市场,一旦运营故障浮出水面,战略溢价会迅速蒸发。Harmattan 必须维持无瑕疵的可靠性记录,才能守住溢价估值。[CV001, CV002, CV003, CV004, CV005, CV006]
| 论点 | 证据锚点 | 改变判断的条件 |
|---|---|---|
| 15 个月内拿下两项国家级国防部 PoR,验证产品市场匹配和交付能力 | 法国国防部(1,000 架 SORONA 无人机,2025 年 7 月);英国国防部(最多 3,000 套系统,2025 年 9 月);CEO M'Ghari 在 2026 年 2 月 Le Grand Continent 访谈中确认 | 未能按时履行现有 PoR 承诺,或已交付系统出现客户投诉 / 可靠性事件 |
| Dassault 合作形成复制壁垒,并带来 Rafale F5 AI 授权 | 2026 年 1 月联合新闻稿;Dassault CEO Eric Trappier 公开声明;Defense News 和 TechCrunch 对 Rafale F5 与 UCAS AI 集成授权的报道 | Thales cortAIx 或竞争 AI 供应商被正式选为 Rafale F5 主要 AI 供应商,把 Harmattan 挤出核心授权 |
| 欧洲防务市场年增速 50%+,降低执行风险 | BCG 2026 年 5 月 C-UAS 白皮书:C-UAS 需求预计到 2029 年达到 $5–7B,CAGR 58%+;2026 年 5 月 NATO TIE 26 互操作演习验证管线 | NATO 防务开支收缩、乌克兰通过外交解决,或预算大规模从自主系统转向其他方向 |
| 收入不透明,无法形成确信定价(反向逻辑) | 无公开财务指标;私营 SAS 公司无申报义务;TechCrunch 确认 Series B 时收入未披露;Reuters 指出,法国无人机制造商尽管收入目标激进,融资仍不确定 | Harmattan 披露可验证 ARR 或合同总额;或 Series C 轮估值独立验证其财务指标 |
| 电池供应链主权化仍需 3–5 年(反向逻辑) | CEO M'Ghari 2026 年 2 月 Le Grand Continent 访谈:无刷电机磁体和电池化学体系依赖中国稀土;主权电池方案还需 3–5 年;主权电机方案预计 2026 年底落地 | 独立验证显示主权电池供应链提前锁定;或供应链中断实际发生,坐实该风险 |
| 治理不透明限制投资者保护(反向逻辑) | 未公开披露董事会组成;联合创始人控制高管决策;未发布治理章程;法国 SAS 架构公开监督有限 | Harmattan 披露董事会组成、投资者权利或治理章程;或治理失效在媒体报道的纠纷中显现 |
投资逻辑行引用证实性证据;反向逻辑行引用反向或中性证据。改变判断的条件是投资监控的首要触发清单。
[CV004, CV005, CV006, CV007, CV008, CV009]建议链条从市场验证、产品证据、估值分析,最后走到「有条件关注」建议;两个论点锚点和三项阻断性尽调缺口, 决定了附带条件的限定。
[CV001, CV002, CV003, CV013, CV014, CV042]8.2 估值背景、融资与入场纪律
Harmattan AI 的融资历史压缩得很快。Atlantic VC 领投的种子轮金额未披露,之后是 FirstMark Capital 领投、Atlantic VC、Tholus Capital、Motier Ventures 和 Sisyphus Ventures 参与的约 ~$30 million Series A;TechCrunch 报道 Series B 前累计融资为 $42 million。2026 年 1 月 12 日,Dassault Aviation 领投 $200 million Series B,确立 $1.4 billion 投后估值——Defense News、TechCrunch、Tech.eu、AirForce-Technology.com 以及 Harmattan 自己的新闻稿均确认了这一点。截至 2026 年 1 月,累计融资约 $242 million(Series B 前 $42 million 加 $200 million Series B)。Crunchbase 也确认融资轨迹,尽管平台有速率限制、无法直接查询;DefenceJobs.org 列出 Dassault Aviation、FirstMark Capital、Atlantic VC、Motier Ventures 和 Tholus Capital 为投资方,进一步印证了财团结构。 入场纪律需要调和四组张力。第一,估值 vs. 收入证明:$1.4 billion 估值约等于累计融资($242 million)的 5.8×。对一家拥有两个国家国防部正式列装项目、且有 Dassault 锚定战略护城河的公司来说,这个水平合理——但前提是合同收入正在规模化落地。没有披露,投资者面对的是二元执行风险。第二,Series A 到 Series B 的跃升很大:Harmattan 称 Series B 相比 Series A「显著提高」,但既没有披露 Series A 估值,也没有披露跃升倍数,稀释模型因此不精确。第三,优先股堆叠动态未知:Dassault 投资份额相对于财务 VC 投资者如何处理、清算优先权、反稀释条款以及任何信息权,都是未披露的私人条款。第四,主权防务溢价:法国总统在 X 上背书,并将其定位为法国首家防务独角兽,带来一个政治估值底;在法国政府相邻的场景里(被防务主承包商收购、法国政府战略持股),这可能提供下行保护,但对财务投资者而言不能直接变成可流通价值。 欧洲同行对比很有参考意义。德国 Helsing 据报道 2026 年 5 月接近以 $18 billion 估值融资 $1.2 billion,运营 5 年,对比 Harmattan 的 2 年;如果 Harmattan 持续执行,按年龄调整后的估值轨迹对它有利。Quantum Systems 在 2025 年 5 月完成 €160 million Series C 后,又在 2025 年 11 月一轮融资中将估值提高到超过 €3 billion;Reuters 指出,这次估值三倍增长由 100% 年收入增长驱动,而 Harmattan 尚未做出这种财务披露。总部在里斯本的 ISR 无人机龙头 TEKEVER 以超过 £1 billion 估值融资 £400 million;关键差异在于 TEKEVER 已经盈利,估值下方有财务底,Harmattan 目前还不能这么说。Anduril 2025 年收入翻倍至 $2.2 billion 后,在 Series H 达到 $61 billion 估值,代表自主防务平台上限;Shield AI 的 $12.7 billion 估值反映了 AI pilot 已部署到 26 类载具上的软件厚度。Harmattan 的 $1.4 billion 位于同行区间低端,说明如果未来披露收入且数据强劲,仍有重估空间;但也说明当前入场倍数已经默认这些指标会不错。[CV013, CV014, CV015, CV016, CV017, CV018]
| 维度 | 评估 | 证据依据 |
|---|---|---|
| 建议 | 有条件观察 / 继续研究 | 两项国防部 PoR、Dassault 合作、50%+ 市场 CAGR 支撑战略溢价;但收入不透明和治理缺口不足以支持确信买入。 |
| 置信度 | 中 | 合同验证可信度高;收入、毛利率、生产成本等估值输入仍属私有信息,未获任何第三方来源验证。 |
| 风险评级 | 高 | 关键人物集中、供应链脆弱、收入不透明且缺乏公开治理披露,留下重大未解决敞口。 |
| 估值立场 | 对战略投资者合理到偏满;对纯财务投资者偏高 | $1.4B 在 Dassault 合作和国防部验证下站得住,但财务回报建模仍必须验证收入。 |
| 决策含义 | 获取收入数据;影子跟踪 Series C 节奏;启动供应链尽调 | 首次收入披露后重新评估;若毛利率确认高于 50% 且第二项 Dassault AI 授权得到确认,可上调为观察。 |
评估基于截至 2026-06-07 的公开证据;Harmattan AI 未公开披露收入、毛利率和生产成本数据。
[CV001, CV013, CV014, CV042]以 2026 年 1 月 $1.4B 的 Series B 入场估值为参照,给出 Harmattan AI 2028 年牛市、基准、熊市估值区间, 并列出情景假设和关键触发事件。
估值区间为作者基于同业对比和情景逻辑作出的估算;没有收入披露,无法做 DCF。概率信号是定性评估, 不是精算概率。回报数字假设 Series C 不稀释;实际回报取决于稀释、优先权包袱和退出时点。
[CV036, CV037, CV038, CV039, CV040, CV041]面向 IC 的七项投资维度评分,反映第 1–8 章基于证据的评估。分数代表作者对当前证据质量和风险调整后强度的判断, 采用五分制(1=很弱,5=很强)。
IC 评分是研究分析师基于截至 2026 年 6 月 7 日收集证据作出的定性判断。一旦收入披露(经济性 / 财务) 或 Rafale F5 AI 任务确认(护城河),分数会出现实质变化。
[CV001, CV002, CV003, CV007, CV008, CV013]8.3 可比公司组与战略溢价分析
收入倍数不适用于 Harmattan AI,因为收入没有披露。估值分析只能用情景逻辑、先例融资倍数(累计融资倍数)、战略溢价归因和可比公司画像,三角定位一个可辩护区间。最相关的五个同行是 Helsing(德国 / 欧洲,防务 AI 软件 + UAS)、Anduril Industries(美国,自主防务平台)、Shield AI(美国,AI pilot 自主软件 + UAS 硬件)、TEKEVER(葡萄牙 / 英国,ISR 无人机服务)和 Quantum Systems(德国,ISR UAS 硬件)。每家公司提供不同视角:Helsing 对应欧洲主权 AI 定位;Anduril 是规模化全栈模式的最清晰类比;Shield AI 隔离出纯软件溢价;TEKEVER 展示财务证明溢价;Quantum Systems 展示有披露收入增长的硬件 ISR 倍数。 Harmattan 相比纯 ISR 硬件同行(如 TEKEVER,约为累计融资 1× 到 1.5×)拿到的战略溢价,来自三点:Dassault 合作带来防务主承包商集成授权,形成复制壁垒;公司成立 15 个月后即拿到两个国防部正式列装项目;以及跨 ISR、C-UAS、EW、打击四个防务细分的垂直整合平台架构。这些因素支撑其相对 TEKEVER 和 Quantum Systems 的溢价,但 Harmattan 只有 2 年历史且收入不透明,限制了它相对 Helsing(5 年)和 Anduril(8+ 年)的定价。因此,中性估值立场是:对能看见合同爬坡的战略投资者,$1.4 billion 当前入场价算合理;对没有收入披露、无法锚定回报模型的财务投资者,则已经偏满 / 偏高。 同行证据里的反向因素包括:Reuters 2026 年 6 月报道一家硅谷防务创业公司的无人机坠毁和安全事件,说明运营可靠性失守时倍数会迅速压缩;Quantum Systems 估值三倍增长明确建立在已披露 100% 年收入增长之上,强调重估所需的同行门槛是财务披露;Reuters 2025 年 11 月提到法国无人机融资不确定性,特别指出防务创业公司野心与支撑估值的未披露收入之间存在张力。这些反向信号强化了投资建议中的条件限定。[CV027, CV028, CV029, CV030, CV031, CV032]
| 公司 | 阶段 / 最新轮次 | 估值 / 融资额 | 已披露收入或指标 | 与 Harmattan 的相关性 | 局限 |
|---|---|---|---|---|---|
| Helsing(德国) | Series D(据报道 2026 年 5 月接近完成一轮 $1.2B 融资,估值 $18B);2025 年 6 月以约 €12B 估值融资 €600M | $18B(估计,待交割) | 未公开披露;防务 AI 软件 + 面向乌克兰的 HX-2 无人机 | 最接近的欧洲可比:主权 AI + 自主无人机;成立 5 年,而 Harmattan 2 年;更高估值反映更长验证轨迹 | 年龄溢价 5×,累计融资 $600M+,而 Harmattan 为 $242M;两家公司都未披露收入;HX-2 制造规模未确认达到 Harmattan 水平 |
| Anduril Industries(美国) | Series H(融资 $5B,投后估值 $61B,2026 年 5 月) | $61B | 2025 年收入 $2.2B(同比翻倍);成立 9 年 | 全栈自主防务主承包商可比;Lattice 平台对应 Kalahari C2;Roadrunner C-UAS 对应 Gobi / Gobi Tempest;展示垂直整合防务平台在披露收入后的估值天花板 | 总部在美国;客户以 DOD 为中心;成立 9 年,而 Harmattan 2 年;披露收入 $2.2B 支撑其 $61B 估值;Harmattan 缺乏可比披露 |
| Shield AI(美国) | Series G(累计融资 $1.5B,投后估值 $12.7B,2026 年 3 月) | $12.7B | 未公开披露;AI 飞行员软件 + V-BAT / X-BAT UAS;成立 11 年 | 软件占比高的 AI 自主系统可比;Hivemind AI 飞行员对应 Harmattan 嵌入式 AI;显示硬件之上的软件溢价;Series G 由 Advent International 领投 | 美国防务重点,且有明确空军 CCA 授权;成立 11 年,而 Harmattan 2 年;未披露收入;$12.7B 估值反映纯软件溢价,Harmattan 的硬件集成模式需要匹配这一溢价 |
| TEKEVER(葡萄牙 / 英国) | 2025 年以 >£1B 估值融资 £400M | >£1B (~$1.3B) | 已盈利(TEKEVER 确认 Series B 时盈利);ISR 无人机服务模式;已在乌克兰部署;英国 Home Office 和 EMSA 合同 | 财务验证最接近的欧洲 ISR UAS 同行;盈利证明为估值提供财务底部;获 Baillie Gifford + NATO Innovation Fund 支持 | 仅 ISR(无 C-UAS、EW 或打击);服务收入模式(Intelligence-as-a-Service)对比 Harmattan 的系统销售;增长轨迹低于 Harmattan |
| Quantum Systems(德国) | €160M Series C(2025 年 5 月),估值约 €1B+;Reuters 2025 年 11 月称,在 100% 年收入增长后,估值翻至 >€3B | >€3B (~$3.3B) | 100% 年收入增长(已披露);550 名员工;收购 AirRobot + Nordic Unmanned UK | 硬件 ISR UAS 可比,且披露收入增长;Series C 获 Balderton Capital、Hensoldt、Airbus Defence & Space 支持;成立 10 年 | 收入增长披露支撑其 €3B 估值;仅硬件 ISR(无完整 C2 平台、EW 或 C-UAS);成立 10 年,而 Harmattan 2 年;欧洲防务可比但没有 Dassault 级战略投资者 |
| Harmattan AI(法国)— 参照 | Series B($200M,投后估值 $1.4B,2026 年 1 月) | $1.4B | 未公开披露;两项国防部 PoR;Dassault Aviation 合作 | 参照公司 | 收入未披露;成立 2 年;$1.4B 处在同行区间低端,若执行延续则有重估上行;若出现运营失败或财务表现不及预期,则有压缩风险 |
所有估值均为独立新闻来源报道的私募轮投后估值,未经独立核验。Harmattan、Helsing 或 Shield AI 未披露收入,因此无法计算倍数。Quantum Systems 披露收入增长;Anduril 的 $2.2B 收入是同行组里唯一直接财务数据点。
[CV017, CV018, CV019, CV020, CV021, CV022]用最新披露或估算轮次,对防务自主化同业的隐含收入倍数做示意性敏感性分析,显示收入已披露同业 (Anduril、Quantum Systems)与收入不透明同业(Harmattan、Helsing、Shield AI)之间的宽幅差异。 柱形表示累计融资额相对于最新投后估值的倍数;披露收入时展示收入倍数。
所有数字来自公开报道的轮次披露和估算累计融资;私营公司没有经过核验的资本结构表。Harmattan $242M 累计融资为估算值,来自两项独立公开披露(B 轮前 $42M + Series B $200M)。Quantum Systems 累计融资估算来自连续披露轮次。收入倍数仅供方向性参考。
[CV013, CV017, CV018, CV019, CV020, CV021]8.4 乐观、基准与悲观情景
乐观情景假设 Harmattan AI 到 2027 年底前再拿下两个或更多 NATO 成员国正式列装项目,在 2030 年目标前启动 Rafale F5 标准的 AI 集成工作,实现每月 5,000+ 套系统产能,并披露年经常性收入(ARR)超过 $300 million。在这些假设下,Dassault 合作会形成事实上的防务主承包商合作估值底,2028–2029 年向 Helsing 类倍数(成立约 5 年时估值 $18 billion)重估就变得可达。乐观情景估值区间为 2028 年 $4–8 billion,意味着相对 2026 年 1 月 Series B 入场有 3×–6× 回报。打开乐观情景的关键触发点是 Rafale F5 AI 授权:如果 Harmattan 的嵌入式 AI 被确认为法国下一代战斗机项目的主要 AI 供应商,可触达合同总价值将是变革性的(仅法国的 Rafale F5 目标就有 40+ 架机体,并伴随数十年支持周期)。 基准情景假设到 2027 年底前再拿下一个欧洲 NATO 正式列装项目,现有法国和英国合同稳定交付并验证利润率,没有重大运营故障或供应链中断,最终披露部分收入,确认单位经济模型可观但并不异常优秀。Dassault 合作稳步推进但不加速。基准情景估值区间为 2028–2029 年 $2.5–4 billion,意味着相对当前入场有 1.8×–3× 回报。这个情景与同行区间一致:Quantum Systems 在披露 100% 收入增长后估值提高三倍,说明 Harmattan 只要一次可信的财务披露,就可能催化类似跃升。 悲观情景假设出现运营可靠性事件(无人机坠毁、任务失败、供应链中断导致交付延迟),Rafale F5 AI 集成被授予其他供应商(例如 Thales cortAIx,Dassault 与其也有 2025 年 11 月协议),且收入持续不透明,外部验证无法完成。在这些条件下,Harmattan 在假设的 Series C 中会面对平轮或下轮,$1.4 billion 估值会向累计融资锚点($242 million 基数)压缩。悲观情景估值区间为 $800 million 到 $1.2 billion,意味着只有 Series B 价格的 0.6×–0.9×,财务投资者会遭受实质减值。关键悲观触发点不是单独的中国供应链中断——CEO 承认过这个风险,也给出了解决时间线——而是供应链中断与 Thales 赢得竞争性 AI 授权同时发生。[CV036, CV037, CV038, CV039, CV040, CV041]
| 情景 | 关键假设 | 估值区间(2028) | 概率信号 | 关键风险或触发项 |
|---|---|---|---|---|
| 乐观情景(30% 权重) | 到 2027 年底新增 2 项以上 NATO PoR;确认 Rafale F5 AI 以 Harmattan 为主;产量达 5,000+ / 月;披露 ARR >$300M;Dassault 合作加深 | $4B–$8B | 需要确认 Rafale F5 AI 授权并披露收入;先例:Helsing 在可比轨迹下第 5 年估值 $18B;Anduril 收入翻倍后估值 $61B | Thales 在 Rafale F5 AI 上取代 Harmattan;或在 F5 标准确认(约 2030 年)前出现运营失败 |
| 基准情景(50% 权重) | 到 2027 年底新增 1 项欧洲 NATO PoR;现有法国和英国合同按时交付;无重大供应链中断;部分披露收入;Dassault 合作推进但不加速 | $2.5B–$4B | Quantum Systems 在披露收入增长 100% 后估值翻三倍;Harmattan 可能在首次可信财务披露后重估;TEKEVER 在盈利基础上估值约 £1B+ | 收入披露低于预期;供应链中断拖延交付;或防务开支收缩压低估值倍数 |
| 悲观情景(20% 权重) | 发生运营可靠性事件或交付失败;Rafale F5 AI 授权给 Thales;收入继续不透明;Series C 平轮或下轮;中国供应链中断落地 | $0.8B–$1.2B | Reuters 2026 年 6 月无人机坠毁先例显示,战略溢价会快速消退;法国无人机融资不确定性(Reuters 2025 年 11 月)显示市场对未披露收入存疑 | Reuters 对美国同行无人机坠毁的反向报道(2026 年 6 月关于一家 $1.3B 防务初创公司的文章)给出时间线:不利运营消息后一个季度内,溢价即崩塌 |
概率权重是指示性信号,不是精算概率。估值区间为作者基于可比私募轮倍数的估计;没有收入披露,无法做 DCF。回报测算以 2026 年 1 月 Series B 入场估值 $1.4B 为基础。
[CV036, CV037, CV038, CV039, CV040, CV041]8.5 投资建议、退出准备度与最终尽调问题
投资建议是有条件观察 / 继续研究。证据支持 $1.4 billion 估值里确有战略溢价:15 个月内拿下两个国家国防部正式列装项目、Dassault Aviation 里程碑式 AI 集成授权、欧洲防务市场 50%+ 年增长顺风,合在一起足以支撑明显高于单纯累计融资倍数的估值。但五个阻断性尽调缺口让我们无法给出明确买入:(1)没有任何公开财务披露(收入、ARR、利润率、烧钱);(2)董事会构成和治理结构未披露;(3)除了 CEO 声称的「每月数千台」外,没有验证产能规模;(4)Dassault 合作条款以及 Harmattan 在 Rafale F5 AI 中的具体角色没有独立确认;(5)供应链主权化时间线依赖 CEO 表述,未被独立验证。 退出准备度仍然过早。Harmattan AI 成立于 2024 年 4 月,注册为法国 SAS,没有公开申报义务,也没有释放 IPO 或战略出售的信号。Dassault Aviation 的战略合作和持股创造了一条合乎逻辑的收购路径(Dassault 整合自身 AI 能力),但这仍属推测;如果乐观情景假设兑现,更可能的近期结果是在更高估值上完成 Series C。Harmattan AI 2026 年 6 月在巴黎 Eurosatory 2026 参展,以及美国团队扩张,都更像近期国际市场开发,而不是退出准备。 战略收购方应采用不同于财务投资者的估值方式。对希望掌握空战主权 AI 的欧洲防务主承包商或国家政府而言,Harmattan 是稀缺资产;如果执行持续,Series C 后大概率不会再以当前价格出现。对财务投资者而言,当前入场价已经包含战略溢价,回报取决于财务披露能否验证底层模型——这件事尚未发生,也无法从公开证据预测。 估值立场是在战略溢价和执行不透明之间求平衡:对使命匹配的战略投资者是合理至偏满;对没有收入可见性的纯财务回报则偏满 / 偏高。建议的尽调优先级,是在提高超出战略跟踪仓位的敞口前,拿到保密收入数据、利润率结构和单台生产成本数据。[CV042, CV043, CV044, CV045]
| 触发项 | 阈值或事件 | 对投资逻辑的传导 | 行动含义 |
|---|---|---|---|
| 运营可靠性失败 | Harmattan 系统发生无人机坠毁、任务失败或安全事件,并收到负面媒体报道或触发国防部审查 | 战略溢价立即消退;Reuters 2026 年 6 月报道一家 $1.3B 美国防务初创公司无人机坠毁,显示投资者信心在一个季度内被侵蚀;PoR 合同承压 | 全面退出,或在调查期间暂停持有;监控国防部关于合同状态的公开声明 |
| Rafale F5 AI 授权输给竞争对手 | Dassault Aviation 宣布 Thales cortAIx 或另一家供应商成为 Rafale F5 标准的主要 AI 合作方,把 Harmattan 挤出核心授权 | 投资逻辑锚点(Dassault 合作护城河)被打断;没有战略溢价,将重估到纯硬件 ISR 倍数;乐观情景概率接近归零 | 立即减仓;下调至悲观情景估值 $800M–$1.2B |
| 中国出口管制造成供应链中断 | 中国政府限制电机磁体稀土或电池材料出口,并实质性拖延或停止 Harmattan 产线 | 产量扩张目标变成愿望;PoR 交付承诺承压;主权可信度叙事受损;基准和乐观情景都受伤 | 监控中国稀土出口政策;要求管理层提供供应链备份方案和库存水位 |
| 收入披露低于预期 | 首次自愿或强制收入披露(例如 Series C 时)显示 ARR 低于 $50M 或毛利率低于 40% | 估值倍数压缩;悲观情景变成基准;按战略溢价定价的财务投资者出现减值;Anduril 和 Helsing 可比不再支持 $1.4B 入场 | 下调至减持;基于披露收入重新建模;关注毛利率轨迹和烧钱跑道 |
| 治理或法律事件 | 公开报道创始人纠纷、投资者诉讼、监管行动(ITAR/EAR 出口管制违规)或非自愿董事会重组 | 关键人物集中风险落地;与 Dassault 的战略投资者关系受扰;法国国防部合同可靠性遭质疑 | 鉴于美国团队扩张,全面复核并立即让法律顾问介入 ITAR/EAR 合规状态 |
终止触发项按对投资逻辑的传导严重性排序。运营失败和 Rafale F5 授权丢失会打破投资逻辑;供应链和治理触发项会削弱投资逻辑,但若在 6–12 个月内处理,仍可恢复。
[CV008, CV011, CV040, CV041, CV044, CV045]| 主题 | 缺失证据 | 重要性 | 负责人或尽调路径 |
|---|---|---|---|
| 收入与财务模型 | ARR、合同总额、毛利率、烧钱速度、单位经济性(单架无人机成本、每份 PoR 合同贡献毛利)、现金跑道 | 没有财务数据,估值完全依赖战略溢价;无法建模回报情景或压力测试悲观情景;没有可比项可验证倍数 | 直接接触管理层(Series C 流程数据室);需签 NDA;向法国国防部采购官做参考访谈,核实合同金额区间 |
| Dassault 合作范围与 AI 授权 | Harmattan 是 Rafale F5 项目的主要 AI 供应商,还是多个供应商之一;UCAS AI 开发授权的合同价值和范围;排他条款;来自 Dassault 与国防部客户的收入归因 | Dassault 合作是最重要的投资逻辑锚点;若 Harmattan 只是多个 AI 合作方之一而非主要供应商,战略溢价会明显下降 | 直接访谈 Dassault Aviation CEO 或项目副总裁;复核联合新闻稿中的排他性表述;联系法国 DGA 项目办公室 |
| 产能规模与单位经济性 | 经验证的月产量(CEO 所称“每月数千台”);制造设施产能;SORONA 单位成本和英国合同单位定价;供应链备份方案和原材料库存水位 | 产能规模是核心运营逻辑;若产量实质性低于说法,悲观情景已经出现;单位经济性决定规模是否转化为利润 | 工厂走访;供应链审计;向法国国防部索取生产日志作为客户参考;通过 DroneXL 或防务媒体独立核验 |
| 董事会组成与治理 | 董事姓名和背景;投资者代表与投票权;任何股东协议条款;信息权;反稀释保护;Dassault 股权与财务 VC 的待遇 | 治理不透明带来关键人物和保护性条款风险;没有董事会可见度,就无法评估投资者保护;对任何超过战略性小票的投资都至关重要 | 向管理层或法律顾问直接索取;通过法国公司注册处(Registre du commerce et des sociétés)查 Infogreffe 董事会备案 |
| 出口管制合规 | 无人机系统的 ITAR/EAR 分类;美国团队扩张,以及美国原产技术是否转移给非美国国防部客户;法国 CIEEMG 对英国和爱沙尼亚合同的出口许可状态 | 防务技术出口管制违规会带来严厉刑事和商业处罚;Arlington VA 的美国团队暗示可能涉及美国原产技术;对 NATO 盟友客户而言,法国出口法合规具有重大影响 | 法律顾问复核;检查 CIEEMG 数据库;直接询问管理层所有系统的 ITAR/EAR 分类 |
| 供应链主权化进展 | 当前对中国稀土的依赖水平;具体电池和电机磁体供应链备份供应商;主权电机磁体(预计 2026 年底)和电池(3–5 年)的时间表;暴露于中国供应商的 COGS 占比 | CEO 已承认该问题重大;不利地缘政治事件(台湾海峡升级、中国出口限制)可能在没有主权替代方案的情况下打断规模化生产 | 供应链审计;CEO Le Grand Continent 访谈提供基线;与公开报道提及的电机和电池供应商做参考访谈 |
议题按对建议的重要性排序;财务模型和 Dassault 范围是阻塞性尽调项。鉴于美国团队扩张和多 NATO 客户基础,出口管制合规至关重要。
[CV006, CV008, CV009, CV010, CV011, CV012]8.6 图表与附录
免责声明
本报告仅基于截至 2026-06-07 的公开来源研究,不构成投资、法律、出口管制或采购建议。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Harmattan AI is incorporated as a simplified joint-stock company (SAS) under French law, registered at 1 rue du Mail, 75002 Paris, France, with VAT number FR39978035392. | 中 | SO013 |
| CO002 | Harmattan AI was founded in April 2024 in Paris, France. | 高 | SO009, SO022 |
| CO003 | Harmattan AI describes itself as a defense technology company developing vertically integrated autonomous systems including layered air-defense solutions, coordinated ISR and strike UAVs, electronic-warfare products, and C2 platforms. | 高 | SO009, SO001 |
| CO004 | Harmattan AI's stated strategic doctrine is "coordinated saturation" — deploying autonomous robot forces at mass scale to operate as a unified whole across the full operational spectrum. | 中 | SO001 |
| CO005 | The company builds hardware, embedded AI software, and C2 systems in-house under a vertically integrated model; its three strategic pillars are autonomous AI accumulation, mass production at scale, and rapid operational deployment. | 中 | SO001, SO024 |
| CO006 | As of June 2026, Harmattan AI operates in six countries: France, United Kingdom, Switzerland, Morocco, the United Arab Emirates, and the United States. | 中 | SO010 |
| CO007 | The company's open-roles page in June 2026 lists active positions in Paris, Lausanne (Switzerland), Wissous (France), London, Arlington VA (USA), UAE, and Poland across R&D, manufacturing, software, and business development functions. | 中 | SO011 |
| CO008 | Mouad M'Ghari is CEO and Co-Founder of Harmattan AI; he is a graduate of École Polytechnique and MIT and serves as the company's primary public spokesperson in all major press releases and media interviews. | 高 | SO009, SO016, SO024 |
| CO009 | Martin de Gourcuff is CTO and Co-Founder of Harmattan AI. | 中 | SO009, SO016 |
| CO010 | Edouard Rosset is a Co-Founder of Harmattan AI, listed in Tech.eu, TechCrunch, and DefenceJobs.org profiles; his specific functional role has not been publicly disclosed. | 中 | SO020, SO025 |
| CO011 | Marc Grelet is identified as a Co-Founder of Harmattan AI in the DefenceJobs.org company profile; no background or functional role has been publicly disclosed. | 低 | SO025 |
| CO012 | Harmattan AI's open-roles page lists a Chief Engineer and System Architect among senior technical openings, indicating that the CTO is supported by senior systems-level engineering staff. | 低 | SO011 |
| CO013 | The full board of directors composition of Harmattan AI, including investor board-seat rights, has not been publicly disclosed as of June 2026. | 低 | |
| CO014 | Harmattan AI closed a $200 million Series B funding round on January 12, 2026, led by Dassault Aviation. | 高 | SO009, SO015, SO016, SO017 |
| CO015 | The Series B round values Harmattan AI at $1.4 billion post-money, making it France's first defense-sector unicorn. | 高 | SO009, SO017, SO020 |
| CO016 | Prior to the Series B, Harmattan AI had raised approximately $42 million in total, comprising a seed round led by Atlantic VC and a Series A led by FirstMark Capital, with Motier Ventures and Sisyphus Ventures also participating, according to TechCrunch. | 中 | SO016, SO025 |
| CO017 | The Series A round is separately reported at approximately $30 million, led by FirstMark Capital with Atlantic VC and Tholus Capital participating, consistent with a prior seed of approximately $12 million bringing the pre-Series B total to ~$42 million. | 中 | SO025 |
| CO018 | Total lifetime capital raised by Harmattan AI through January 2026 is estimated at approximately $242 million, combining the $42 million pre-Series B total and the $200 million Series B. | 中 | SO016, SO025 |
| CO019 | Motier Ventures renewed its investment in Harmattan AI as part of the Series B round, according to a Motier Ventures LinkedIn post cited by Defense News. | 中 | SO017 |
| CO020 | French President Emmanuel Macron praised the Dassault–Harmattan partnership on X, calling it "excellent news for our strategic autonomy" and welcoming Harmattan's unicorn status. | 中 | SO017, SO020 |
| CO021 | The Dassault Aviation Series B partnership includes a mandate to develop embedded AI capabilities for the Rafale F5 fighter standard and for Dassault's future unmanned combat aerial systems (UCAS). | 高 | SO009, SO015, SO017 |
| CO022 | As part of the Dassault partnership, Dassault Aviation will provide expertise in system architecture, mission-system integration in high-intensity environments, and international business development support. | 中 | SO009, SO015 |
| CO023 | Harmattan AI stated that Series B proceeds will fund expansion of AI-enabled mission deployments across new operational theaters, new product domains, and scaling of ISR, drone-interception, and EW industrial manufacturing. | 中 | SO009, SO015 |
| CO024 | In July 2025 — approximately 15 months after founding — Harmattan AI was awarded its first Program of Record, a multi-million-dollar NATO government contract for AI-enabled small drones. | 中 | SO009, SO022 |
| CO025 | France's Armed Forces Ministry (DGA) ordered 1,000 AI-enabled quadcopter drones (SORONA model) from Harmattan AI for delivery by end of 2025, constituting the first Program of Record. | 中 | SO022, SO009 |
| CO026 | The French-ordered Sonora drone weighs under 1.8 kg, has a flight endurance of over 40 minutes, a 2 km operational range, and carries electro-optical and infrared cameras (infrared sensor supplied by French firm Lynred). | 中 | SO005, SO017 |
| CO027 | The UK Ministry of Defence awarded Harmattan AI a contract in September 2025 for up to 3,000 autonomous drone systems for "urgent operational needs," constituting Harmattan's second Program of Record. | 高 | SO009, SO017, SO019 |
| CO028 | Harmattan AI announced a strategic partnership with Ukrainian UAS maker Skyeton on October 2, 2025, to integrate Harmattan's sensor technology with Skyeton's Raybird UAS for French and NATO markets. | 中 | SO014, SO009 |
| CO029 | The Skyeton Raybird UAS has accumulated over 350,000 combat flight hours with Ukrainian defense forces; the Harmattan partnership will integrate the Sahara SAR payload and embedded AI into the Raybird platform. | 中 | SO014 |
| CO030 | Harmattan AI's Sahara SAR payload delivers 0.25 m spatial resolution at 2 km and 1.2 m at 10 km, weighs under 3.5 kg, and operates in all-weather, all-light conditions. | 中 | SO006, SO024 |
| CO031 | Harmattan AI's Kalahari C2 platform uses open and defense-grade standards including Asterix, CoT, Lattice, Sapient, and NATO standards; it supports on-premise or hybrid edge-to-cloud deployment with end-to-end encryption and role-based access control. | 中 | SO007 |
| CO032 | Harmattan AI's product portfolio comprises six core systems: Sonora (ISR training UAS), Sahara (SAR payload), Gobi (Group 1-2 VSHORAD interceptor), Gobi Tempest (Group 2-3 VSHORAD interceptor), Barkhan (precision strike UAS), and Kalahari (C2 platform). | 中 | SO003, SO004, SO005, SO006, SO007 |
| CO033 | Harmattan AI had more than 130 employees as of January 2026, with a median experience of 15 years, according to Defense News citing the company. | 中 | SO017 |
| CO034 | Harmattan AI claimed in January 2026 press materials to be "delivering thousands of systems each month"; this production volume figure has not been independently verified. | 低 | SO009, SO014 |
| CO035 | Defense News reported in January 2026 that Harmattan AI had hired senior executives from Safran and Isar Aerospace within the prior year; individual names were not disclosed. | 中 | SO017 |
| CO036 | CEO M'Ghari confirmed in a February 2026 Le Grand Continent interview that brushless electric motor magnets and battery chemistry remain dependent on Chinese-controlled rare-earth supply chains; sovereign motor-magnet solutions are expected by end 2026, but battery-chemistry sovereignty remains 3–5 years away. | 中 | SO024 |
| CO037 | CEO M'Ghari stated in a February 2026 interview that the French MoD drone contract was fulfilled 14 months after founding, the UK MoD contract 15 months after founding, and an Estonian partner order was completed within 7 months. | 中 | SO024 |
| CO038 | DroneXL noted in January 2026 that the $1.4 billion valuation might seem aggressive for a company barely 18 months old relative to conventional benchmarks, while also arguing the valuation is justified by actual military contracts and battlefield-proven technology. | 中 | SO021 |
| CO039 | A DroneXL analysis cited Harmattan AI plans for manufacturing facilities capable of producing up to 10,000 drones monthly; this forward-looking figure has not been corroborated by independent sources. | 低 | SO021 |
| CO040 | OpexNews.fr explicitly flagged that Harmattan AI's execution challenge post-Series B includes sustaining production cadence, securing the sensor, compute, and radio-link supply chain, and proving reliability in live operational conditions. | 中 | SO023 |
| CO041 | No lawsuits, regulatory enforcement actions, sanctions, or adverse legal proceedings involving Harmattan AI were identified in the sources reviewed for this chapter as of June 2026. | 中 | SO013, SO024 |
| CO042 | As of June 2026, no public disclosure of Harmattan AI's revenue, ARR, gross margin, burn rate, per-unit economics, or cash runway exists in any source reviewed; all financial metrics remain private. | 低 | |
| CO043 | Harmattan AI's blog page records five announced events and news items: Series B (Jan 12, 2026), Eurosatory 2026 (June 15-19), SOF Week 2026 (May 18-21), World Defense Show 2026 (Feb 8-12), UMEX 2026 (Jan 20-22), and Skyeton partnership (Oct 2, 2025). | 中 | SO008 |
| CO044 | Harmattan AI's terms-of-use confirms EU-hosted CRM infrastructure, GDPR data processing, and French law jurisdiction; the document was last updated November 10, 2025. | 中 | SO013 |
| CO045 | CTO and Co-Founder Martin de Gourcuff posted on LinkedIn at the time of the Series B: "As the international order goes off the rails, we are entering an era where, increasingly, power precedes law" — a statement that reflects the founders' geopolitical mission framing and risk appetite. | 中 | SO016 |
| CO046 | TechCrunch reported that Harmattan AI previously described itself as "a European Anduril" before the Dassault partnership, after which the company shifted to "defense technology company" rather than "next-generation defense prime." | 中 | SO016 |
| CO047 | The precise size of Dassault Aviation's individual investment tranche in the Series B and the identities of other co-investors were not disclosed at the time of the announcement. | 中 | SO016, SO017 |
| CO048 | The pre-Series B (Series A) post-money valuation of Harmattan AI was not publicly disclosed; the company described the Series B valuation only as "significantly increases" from the prior round. | 中 | SO017 |
| CM001 | BCG (May 2026) estimates European MoD-committed C-UAS spending at approximately USD 600 million in 2025, growing to approximately USD 800 million in committed programmes by 2029. | 中 | SM016 |
| CM002 | BCG projects total European C-UAS demand (including unallocated defence spend and CNI protection) to exceed USD 5–7 billion by 2029, at an annual growth rate of over 50% (cited as 58%+ p.a.). | 中 | SM016 |
| CM003 | BCG estimates that an aggregate of 4,000–5,000 C-UAS systems will be required in Europe alone, spanning fixed-site, land/naval vehicle, and man-portable deployment configurations. | 中 | SM016 |
| CM004 | BCG estimates that military C-UAS requirements across NATO Europe will account for 1,500–2,000 systems. | 中 | SM016 |
| CM005 | BCG projects that civil critical national infrastructure (CNI) protection—airports, nuclear plants, energy nodes, and large event venues—will require 2,000–3,000 C-UAS systems in Europe, making it one of the largest and fastest-rising C-UAS segments. | 中 | SM016 |
| CM006 | BCG documents that Europe has more than 35 announced C-UAS platform variants versus approximately 13 in North America, reflecting fragmented national programmes that prevent the manufacturing scale needed for efficient mass production. | 中 | SM016 |
| CM007 | NATO's TIE 26 exercise (May 11–22, 2026, Netherlands) involved approximately 300 participants, 40 companies from 11 Allied nations plus Ukraine and Australia, testing more than 60 commercial systems and 40 C2 software applications for C-UAS. | 中 | SM001 |
| CM008 | NATO established five pilot Innovation Ranges under the Rapid Adoption Action Plan (RAAP), including the Latvia range for UAS/C-UAS, the Estonia cyber testbed, the Finland-Sweden connectivity range, the Italy underwater range, and the Netherlands shallow-water range. | 中 | SM002 |
| CM009 | NATO ACT's LCI-X Crucible 1-26 (Romania, 2026) assembled approximately 500 personnel and 215 technical systems representing 21 Allied nations, testing layered C-UAS integration with Ukrainian-derived threat-informed scenarios. | 中 | SM004 |
| CM010 | MarketsandMarkets projects the Europe UAV (all applications) market to grow from USD 5.00 billion in 2025 to USD 7.98 billion by 2030 at a CAGR of 9.8%. | 中 | SM008 |
| CM011 | Expert Market Research estimates the Europe UAV market at USD 9.91 billion in 2025, growing to USD 18.29 billion by 2035 at a 6.32% CAGR; Q1 2026 conflict in the Middle East is cited as a catalyst accelerating European procurement decisions. | 中 | SM009 |
| CM012 | IMARC Group places the Europe drones market at USD 275.6 million in 2025 growing to USD 600.2 million by 2034 at 9.03% CAGR, a figure approximately 36× smaller than MarketsandMarkets's 2025 estimate for the same geography. | 中 | SM010 |
| CM013 | Grand View Research places the Europe drone market at USD 17.05 billion in 2024, growing to USD 36.34 billion by 2030 at a 13.4% CAGR, with hardware as the largest revenue-generating component. | 中 | SM011 |
| CM014 | Market Research Future isolates Europe military drone revenue at USD 2.15 billion in 2024, growing to USD 4.65 billion by 2035 at a 7.26% CAGR, with Germany as the largest market and the UK as the fastest-growing national segment. | 中 | SM014 |
| CM015 | The order-of-magnitude spread between IMARC ($276M) and Grand View ($17B) for "Europe drone market" in 2024–2025 reflects incompatible perimeter definitions rather than methodological error: IMARC captures primarily commercial/consumer segment while Grand View and MarketsandMarkets include defence hardware as the dominant category. | 中 | SM010, SM011, SM008 |
| CM016 | StartUs Insights projects the global military drone market will reach USD 22.81 billion by 2030 at a 7.6% CAGR; the European segment represents approximately 20–25% of global spend by standard regional share estimates. | 低 | SM007 |
| CM017 | StartUs Insights projects the global military drone market to reach USD 98.24 billion by 2033 at an 8.9% CAGR from 2026, supported by approximately 510 startups and 2,780+ companies in the broader ecosystem. | 低 | SM007 |
| CM018 | Germany approved EUR 536 million for contracts with Helsing and Stark to supply strike drones for the Bundeswehr, representing a rapid-procurement template for attritable autonomous systems outside traditional multi-year capital cycles. | 中 | SM007 |
| CM019 | The US Department of Defense FY2026 budget request included a USD 13.4 billion autonomy line: USD 9.4 billion for unmanned and remotely operated aerial vehicles and USD 3.1 billion for counter-UAS, setting a global reference for paired drone and counter-drone investment scale. | 中 | SM007 |
| CM020 | The European aerospace and defence sector recorded EUR 325.7 billion in turnover in 2024, employed approximately 1.1 million people, and spent EUR 25.2 billion on R&D, establishing a large industrial base but one where drone-specific capability lags the US and Israel by an estimated 5–10 years. | 中 | SM007, SM009 |
| CM021 | The European Parliament issued an own-initiative report (January 2026) calling for rapid integration of drone and counter-drone capabilities at every EU defence level, and for a robust, self-sufficient European drone industry free of non-EU supplier dependence. | 中 | SM006 |
| CM022 | The EU European Defence Fund and PESCO jointly fund and coordinate drone capability development and counter-drone research across member states, and the European Defence Agency is advancing unmanned aerial system technology through joint projects and its innovation hub. | 高 | SM013, SM012 |
| CM023 | Ukrainian military leadership stated in May 2024 that drones were the leading cause of casualties on both sides of the Ukraine conflict, surpassing all other weapons—a finding that drove European Parliament and NATO to treat C-UAS as a baseline requirement rather than a niche capability. | 中 | SM013 |
| CM024 | The European Parliament's EPRS report notes that the EU's drone defence initiative faces uncertainty because some member states argue that major defence initiatives should be led by national governments or NATO rather than the European Commission, creating governance risk for centralised EU-level drone procurement. | 中 | SM006 |
| CM025 | SIPRI data (updated March 2026) shows global military expenditure reached USD 2.7 trillion in 2024, a 9.4% year-on-year increase extending a decade-long rise in global defence outlays, with Europe a major contributor to spending growth. | 高 | SM007, SM017 |
| CM026 | Expert Market Research notes that the European UAV industry is estimated to be 5–10 years behind US and Israeli equivalents in autonomy, AI integration, and production scale, requiring sustained investment to close the capability gap. | 中 | SM009 |
| CM027 | Q1 2026 operational data from the Iran-US-Israel conflict validated loitering munitions, counter-drone systems, and AI-enabled drone swarms as decisive capability areas, accelerating European UAV procurement and industrial investment decisions. | 中 | SM009 |
| CM028 | NATO's Rapid Adoption Action Plan (RAAP) established five pilot innovation ranges specifically to accelerate testing, experimentation, and adoption of innovative defence technologies across operational domains. | 中 | SM002 |
| CM029 | LCI-X is explicitly designed as a recurring-cycle experimentation programme to help NATO move from demonstration toward usable capability at operational speed, with Crucible events building on each other across European test sites. | 中 | SM004 |
| CM030 | The JRC C-UAS model describes a multi-stakeholder architecture requiring coordination across defence, federal police, regional police, local government, airspace managers, and frequency authorities—illustrating that CNI C-UAS interoperability is as much an organisational as a technical challenge. | 中 | SM003 |
| CM031 | BCG characterises Europe's C-UAS landscape as fragmented, with parallel national programmes, overlapping R&D, and limited industrial scale that threaten both affordability and the ability to deploy interoperable, networked defences at speed. | 中 | SM016 |
| CM032 | BCG estimates that each time cumulative C-UAS production doubles, unit costs typically fall by 10–20%, following learning-curve dynamics observed across defence and aerospace programmes; European fragmentation prevents most suppliers from advancing down this curve. | 中 | SM016 |
| CM033 | Harmattan AI's VSHORAD capability comprises the Gobi (kinetic interception in controlled environments against smaller threats—Type 1–2) and Gobi Tempest (charge-carrying interception in contested environments against high-intensity threats—Type 2–3), both orchestrated via the Kalahari C2 system. | 中 | SM019, SM025 |
| CM034 | Harmattan AI's persistent ISR capability is delivered through the Sahara SAR payload, which provides high-resolution synthetic aperture radar imaging independent of visibility conditions and operates in all-weather environments. | 中 | SM020 |
| CM035 | Harmattan AI raised USD 200 million in a Series B round led by Dassault Aviation in January 2026 at an implied valuation of approximately USD 1.4 billion, making it a defence unicorn at the time of the raise. | 高 | SM021, SM022, SM024 |
| CM036 | The EU defence readiness roadmap 2030, presented by the Commission and High Representative in October 2025, includes a European drone defence initiative targeting development of an interoperable system for countering and deploying drones, scheduled for launch in early 2026. | 中 | SM006 |
| CM037 | NCIA evaluates TIE 26 systems against track stability, sensor-to-C2 integration, identification performance, and engagement-chain effectiveness, and only systems demonstrating technical and operational interoperability are eligible to advance to Baltic Trust 26. | 中 | SM001 |
| CM038 | TIE 26 (May 2026) included over 60 commercial C-UAS systems and 40 command-and-control software applications tested by 300+ participants in a controlled environment across multiple scenarios, including interceptor capabilities against simulated enemy drones. | 中 | SM001 |
| CM039 | Systems that demonstrate interoperability during TIE 26 are eligible to participate in Baltic Trust 26, an operational exercise scheduled for Latvia in August 2026, which represents the next procurement qualification stage before framework agreement eligibility. | 中 | SM001 |
| CM040 | The NATO Latvia Innovation Range's first TEVV campaign (March 9–13, 2026) brought together defence industry, operational users, and government representatives for high-speed interceptor flights and electronic warfare testing in an open environment. | 中 | SM002 |
| CM041 | The European Defence Agency advances unmanned aerial system technology through joint EU capability projects and its innovation hub, coordinating member-state investment and avoiding duplication in drone development programmes. | 高 | SM012, SM013 |
| CM042 | The Eurodrone MALE RPAS consortium (France, Germany, Spain, Italy) gained Q1 2026 urgency with participating governments accelerating development funding and production commitments following conflict validation of MALE-class capabilities. | 中 | SM009 |
| CM043 | BCG concludes that counter-UAS is shifting from niche capability to baseline requirement across European defence and security, making it one of the fastest-growing segments in the sector. | 中 | SM016 |
| CM044 | BCG estimates C-UAS system assembly and integration costs run approximately €120M per design, with subsystem costs ranging from approximately €40M (design/architecture) to approximately €260M (intercept subsystems), and MRO representing a significant lifecycle cost beyond initial procurement. | 中 | SM016 |
| CM045 | BCG's analysis of the Shahed-136 drone evolution notes that adversary drones can undergo multiple major upgrades (warheads, decoys, propulsion, AI targeting) within two years, implying that C-UAS systems must iterate at the same pace—favouring software-first, open-architecture suppliers. | 中 | SM016 |
| CM046 | Anduril Industries won a 10-year, USD 642 million Programme of Record contract with the US Marine Corps for C-UAS, which BCG cites as a reference demonstrating how rapid iteration and continuous software upgrades can be built into long-term defence contracts. | 中 | SM016 |
| CM047 | SIPRI's Arms Transfers Database (updated March 2026) provides continuous arms transfer data from 1950–2025, documenting European states as historically significant net importers of major conventional weapons systems from the US and Israel. | 中 | SM017 |
| CM048 | Harmattan AI's Kalahari command-and-control system coordinates sensing, threat evaluation, and engagement execution across distributed autonomous systems in real-time under human authorization, serving as the orchestration layer across all Harmattan mission capabilities (ISR, VSHORAD, strike, training). | 中 | SM025, SM018 |
| CM049 | France announced a deal for approximately 1,000 AI-driven training drones (attributed to Harmattan AI) in 2025 to transform military training and deterrence, representing one of the first production-scale attritable drone procurement events by a European sovereign nation. | 中 | SM023 |
| CM050 | NATO's SAPIENCE programme (NATO Science for Peace and Security) supported a university competition in 2025 (Alabama) demonstrating multi-platform autonomous drone cooperation for crisis management, with a third competition planned for the Netherlands in 2026—evidence of NATO's long-run investment in autonomous multi-drone coordination across civilian and defence domains. | 中 | SM005 |
| CP001 | Anduril raised a $5 billion Series H round at a $61 billion post-money valuation in May 2026, led by Thrive Capital and Andreessen Horowitz. | 高 | SP016, SP001 |
| CP002 | Anduril's revenue doubled to $2.2 billion in 2025, according to CEO Brian Schimpf's statement in the May 2026 Series H announcement. | 中 | SP016 |
| CP003 | Anduril has raised more than $11 billion from investors in total across all funding rounds through May 2026. | 中 | SP016 |
| CP004 | Anduril's Lattice platform provides autonomous C2 and sensor fusion across heterogeneous unmanned systems, creating a data-layer integration dependency that makes buyer switching costly. | 中 | SP002, SP001 |
| CP005 | ITAR and dual-use export regulations constrain the degree of sovereign technology transfer that US defense-technology firms including Anduril can offer to European Ministries of Defence seeking autonomous systems with full intellectual property control. | 中 | SP016, SP026 |
| CP006 | Anduril won a contract with the Dutch Ministry of Defence in May 2026, signalling active expansion into the European defense market and direct competitive pressure on European-sovereign defense primes. | 中 | SP016 |
| CP007 | Anduril's Roadrunner is an autonomous air vehicle designed to intercept aerial threats and return to base for reuse, distinguishing it from single-use loitering munitions and offering a structural cost advantage in repeated engagements. | 中 | SP003, SP002 |
| CP008 | Helsing was reportedly in advanced discussions to raise approximately $1.2 billion at an approximately $18 billion valuation in May 2026, with Dragoneer expected to lead and existing investor Lightspeed co-leading. | 高 | SP014, SP025 |
| CP009 | Helsing's first Resilience Factory in southern Germany began manufacturing HX-2 kamikaze strike drones at an initial monthly rate exceeding 1,000 units, with claimed scalability to tens of thousands per month in a conflict scenario. | 高 | SP015, SP024 |
| CP010 | Helsing has raised approximately €761.5 million in total funding through its June 2025 Series C round, making it the most highly valued European defense-autonomy startup by publicly reported figures. | 中 | SP015, SP014 |
| CP011 | Helsing's Altra connectivity platform is designed to link multiple drone systems and enable coordinated multi-drone operations, serving as Helsing's functional equivalent to Harmattan AI's Kalahari C2 layer. | 中 | SP015, SP024 |
| CP012 | Helsing's HF-1 strike drones are produced in partnership with Ukrainian industry; the HX-2 is manufactured at Helsing's proprietary Resilience Factory, reflecting a dual production-sovereignty model. | 中 | SP015 |
| CP013 | Shield AI raised a $1.5 billion Series G at a $12.7 billion post-money valuation in March 2026, led by Advent International and co-led by JPMorganChase's Security and Resiliency Initiative. | 高 | SP017, SP016 |
| CP014 | Shield AI's Hivemind autonomy software has piloted 26 classes of vehicles including F-16 fighter jets, jet-powered UAVs, helicopters, drone boats, and ground vehicles as of March 2026. | 中 | SP017, SP005 |
| CP015 | The US Air Force selected Shield AI's Hivemind as a mission autonomy provider for the Collaborative Combat Aircraft programme and is conducting flight tests aboard the Anduril YFQ-44A CCA. | 中 | SP017, SP016 |
| CP016 | Shield AI is acquiring Aechelon Technology, a defense simulation and synthetic-reality company, to integrate high-fidelity simulator-based training with its Hivemind Foundation Model for autonomous AI pilots. | 中 | SP017 |
| CP017 | Shield AI's V-BAT UAS has been deployed on nearly every class of US Navy ship and with all seven Marine Expeditionary Units, establishing it as the primary maritime-ISR UAS for US expeditionary forces. | 中 | SP004, SP017 |
| CP018 | Shield AI's EdgeOS middleware achieves sub-millisecond local communication latency and sustains millions of messages per second in production deployments, enabling deterministic real-time autonomous control for mission-critical robotics. | 中 | SP005 |
| CP019 | TEKEVER raised €70 million in a Series B round led by Baillie Gifford and backed by the NATO Innovation Fund, and was described by lead investors as already profitable at the time of the raise. | 中 | SP008 |
| CP020 | TEKEVER secured a new €30 million multi-year framework contract with the European Maritime Safety Agency (EMSA) for AR5 UAS maritime surveillance operations. | 中 | SP008, SP007 |
| CP021 | TEKEVER is opening an industrial production site in Cahors, France, targeting operational start-up ahead of summer 2026, strengthening its French manufacturing sovereign capability. | 中 | SP006 |
| CP022 | TEKEVER was selected for Project NYX by the British Army to rapidly develop a loyal-wingman rotary platform designed to operate alongside Apache attack helicopters, with a Bristol engineering hub announced. | 中 | SP006 |
| CP023 | TEKEVER delivers drone-based surveillance primarily as Intelligence-as-a-Service—a subscription or per-mission model—rather than direct platform sales, producing a recurring-revenue structure structurally distinct from Harmattan AI's hardware-sale Program-of-Record model. | 中 | SP008, SP007 |
| CP024 | TEKEVER opened a North Carolina office in May 2026 for dedicated in-country support to US defense partners and partners, representing the company's formal US market entry. | 中 | SP006 |
| CP025 | Quantum Systems raised €160 million in Series C funding in May 2025, led by Balderton Capital with strategic participation from Hensoldt and Airbus Defence and Space, bringing total funding to €310 million. | 中 | SP010, SP014 |
| CP026 | Quantum Systems' valuation reportedly tripled to more than €3 billion following a subsequent November 2025 funding event, based on press coverage citing the round. | 中 | SP014 |
| CP027 | Quantum Systems employs approximately 550 people across Germany, Ukraine, Australia, and Romania as of the May 2025 Series C press release. | 中 | SP010 |
| CP028 | Quantum Systems' Vector AI platform incorporates dual Jetson Orin edge-compute boards for GPS/GNSS-denied autonomous navigation and has been combat-tested in Ukraine, Germany, and Australia. | 中 | SP009, SP010 |
| CP029 | Quantum Systems reported consecutive years of approximately 100% year-over-year revenue growth preceding its May 2025 Series C close. | 中 | SP010 |
| CP030 | Delair's defense-focused drone platforms support real-time data transmission, silent operations with minimal radar signature, and simultaneous video tracking of four targets for security and surveillance missions. | 中 | SP011 |
| CP031 | Delair is primarily an enterprise digital-applications company—focused on data collection and business intelligence for commercial industries—with military defense as an adjacent rather than core market segment. | 中 | SP011, SP012 |
| CP032 | Delair's platforms are rated for operation across temperatures from −15°C to +50°C and support interoperability and communications between multiple drones, meeting basic environmental resilience requirements for tactical deployments. | 中 | SP011 |
| CP033 | MyDefence has deployed more than 5,000 Counter-UAS systems globally and serves over 100 customer organisations worldwide as of June 2026 per company homepage data. | 中 | SP013 |
| CP034 | MyDefence opened a US counter-drone production facility in Oklahoma City in March 2026, operating under a Made-in-USA model to serve US federal and state defense customers. | 中 | SP013 |
| CP035 | MyDefence and partner HighCom secured an A$9.81 million contract with the Australian Department of Defence in April 2026 for counter-drone systems. | 中 | SP013 |
| CP036 | Harmattan AI is the only European defense-autonomy startup with a documented Tier-1 OEM integration path into a NATO-standard combat aircraft programme, via its Dassault Aviation partnership for Rafale F5 and UCAS embedded AI. | 中 | SP018, SP019, SP020 |
| CP037 | Harmattan AI's portfolio spans ISR (Sahara SAR payload, Sonora training UAS), precision strike (Barkhan), kinetic C-UAS interception (Gobi and Gobi Tempest), and C2 mission orchestration (Kalahari), constituting a vertically integrated full-spectrum stack across the kill chain. | 中 | SP018, SP019 |
| CP038 | Harmattan AI holds multiple Programs of Record awarded by both the French Ministry of Defence and the UK Ministry of Defence for its autonomous systems portfolio. | 中 | SP018, SP019, SP020 |
| CP039 | No list pricing is publicly available from Harmattan AI, Helsing, Anduril, or TEKEVER; all pricing is disclosed only via government procurement processes, contract notices, or bilateral MoD negotiations. | 中 | SP001, SP006, SP024, SP018 |
| CP040 | A Reuters investigation published June 2026 documented safety incidents including drone crashes and worker injuries at a US defense-technology startup valued at approximately $13 billion, indicating sector-wide operational-safety scrutiny that may intensify regulatory oversight of autonomous-systems companies. | 中 | SP022 |
| CP041 | Rapid scaling of European first-person-view drone production driven by the Ukraine conflict is compressing per-unit drone hardware costs sector-wide, increasing commoditization pressure on all hardware-focused autonomous defense startups including Harmattan AI. | 中 | SP015, SP023, SP026 |
| CP042 | Harmattan AI's Kalahari C2 platform and Dassault Rafale F5 integration represent a potential platform-level lock-in moat for European MoD buyers, but no independent operational combat-theatre deployment of Kalahari has been publicly verified as of June 2026. | 低 | SP018, SP019 |
| CI001 | Harmattan AI raised $200 million in a Series B funding round led by Dassault Aviation, closing on January 12, 2026. | 高 | SI001, SI002, SI003, SI004 |
| CI002 | The Series B values Harmattan AI at $1.4 billion post-money, making it France's first defense unicorn. | 高 | SI001, SI002, SI003, SI004 |
| CI003 | Harmattan AI operates a business-to-government (B2G) revenue model, selling directly to national ministries of defence under formal Programs of Record. | 高 | SI001, SI022 |
| CI004 | Harmattan AI has been awarded two Programs of Record: the French MoD (1,000 SORONA drones, July 2025) and the UK MoD (up to 3,000 autonomous systems, September 2025). | 高 | SI001, SI004, SI005 |
| CI005 | The French MoD contract for 1,000 SORONA drones is described as a "multi-million-euro commitment" with deliveries scheduled October–December 2025; the exact contract value is not disclosed. | 中 | SI018, SI009 |
| CI006 | The UK MoD contract for up to 3,000 autonomous systems is described as a "multi-million-pound programme"; the exact contract value is not disclosed. | 中 | SI004, SI009 |
| CI007 | Revenue recognition for Harmattan AI's hardware delivery contracts is milestone-based or delivery-on-completion; no public accounting policy or revenue recognition note has been disclosed. | 低 | SI001, SI018 |
| CI008 | Harmattan AI CEO M'Ghari discussed a take-or-pay procurement model in his February 2026 Le Grand Continent interview, advocating for states to finance production capacity analogous to pandemic-era vaccine manufacturing contracts; no such contract has been publicly confirmed. | 中 | SI008 |
| CI009 | The Harmattan AI website operates as a "static catalogue" with no e-commerce functionality; no orders, payments, or sales contracts can be placed via the site, confirming direct B2G sales-only model. | 高 | SI012, SI022 |
| CI010 | No list pricing, per-unit pricing, or rate card has been publicly disclosed for any Harmattan AI product line including SORONA, Gobi, Kalahari, or Sahara. | 高 | SI012, SI022, SI027 |
| CI011 | Harmattan AI has more than 130 employees as of January 2026, with a median experience of 15 years, according to the company; the exact headcount breakdown by function or country is not disclosed. | 中 | SI004, SI013 |
| CI012 | A headcount-based proxy estimate of Harmattan AI's monthly operating cash burn — 130+ employees at blended European defense-sector labor cost of ~$150K–$200K/year — suggests $1.5–$3.0 million per month; this is an analytical estimate, not a disclosed figure. | 低 | SI004, SI016 |
| CI013 | CEO M'Ghari confirmed in the February 2026 Le Grand Continent interview that brushless motor magnets in Harmattan's drone systems contain rare-earth elements requiring Chinese rare-earth refining chains. | 中 | SI008 |
| CI014 | M'Ghari confirmed that lithium-polymer battery chemistry in Harmattan systems also carries Chinese rare-earth dependency, and that resolution of both supply chain exposures is expected to take three to five years. | 中 | SI008 |
| CI015 | No public reports of delivery delays, product failures, or contract disputes on Harmattan AI's Programs of Record were identified in research conducted June 2026; the French MoD delivery was reported as completed by end 2025 per company statements. | 中 | SI001, SI008, SI018 |
| CI016 | The CEO mentioned in the February 2026 Le Grand Continent interview that an Estonian partner order was completed within seven months of founding, confirming at least a third Program of Record customer; contract value is undisclosed. | 中 | SI008 |
| CI017 | Harmattan AI's production cost structure includes high R&D overhead: more than 130 senior engineers operating on a stated six-week technology iteration cycle drive a continuous R&D spend that is amortized across hardware delivery contracts. | 中 | SI008, SI004 |
| CI018 | Defense hardware industry proxies for gross margin on hardware delivery contracts range from 15–40%; AI/software integration services typically carry 50–70% gross margins; Harmattan's blended gross margin structure is unknown. | 低 | SI016, SI017 |
| CI019 | No verified unit manufacturing cost for the SORONA drone has been disclosed; the system weighs under 1.8 kg and incorporates EO optics and optional Lynred IR sensors; market proxy for comparable ISR drones suggests $1K–$8K per unit COGS. | 低 | SI018, SI016 |
| CI020 | Total lifetime capital raised by Harmattan AI through the Series B is approximately $242 million: ~$12 million seed (estimated) + ~$30 million Series A + $200 million Series B; this is an analytical estimate based on TechCrunch reporting that pre-B total was $42 million. | 中 | SI003, SI010, SI011 |
| CI021 | The Series B was led by Dassault Aviation with Motier Ventures renewing its investment; Dassault's specific equity tranche and the identities of other co-investors were not disclosed at the time of announcement. | 高 | SI001, SI003, SI004 |
| CI022 | The Series A raised approximately $30 million, led by FirstMark Capital with Atlantic VC, Tholus Capital, Motier Ventures, and Sisyphus Ventures as participants, per TechCrunch reporting; the post-money valuation at Series A was not disclosed. | 中 | SI003, SI011 |
| CI023 | No debt, project-finance obligations, or credit facility has been disclosed for Harmattan AI in any press release, news coverage, or company document reviewed during research conducted June 2026. | 低 | SI001, SI002 |
| CI024 | Harmattan AI's stated use of Series B proceeds is to expand deployment of AI-enabled missions across new operational theaters, extend the product offering into new domains, and scale industrial manufacturing of ISR, drone interception, and EW platforms. | 高 | SI001, SI002 |
| CI025 | Beyond the $200 million equity injection, Dassault Aviation contributes program management expertise, military systems integration knowledge, and international business development network access as non-monetary strategic capital to the partnership. | 高 | SI001, SI002, SI004 |
| CI026 | At a proxy monthly operating burn of $1.5–$3.0 million (headcount-based estimate) and $200 million raised in January 2026, an analytical runway estimate of 5–8 years at operating expenses alone is indicated; this does not account for manufacturing capex or working capital needs. | 低 | SI004, SI016 |
| CI027 | Harmattan AI's cash on hand post-Series B deployment is not publicly disclosed; the $200 million gross injection does not reflect net cash after ongoing capex and operational expenditures since January 2026. | 中 | SI001, SI002 |
| CI028 | Harmattan AI has disclosed no ARR, revenue run rate, annual revenue, gross margin, operating loss, or net income in any public statement or document as of June 2026. | 高 | SI001, SI003, SI011 |
| CI029 | The aggregate contract value of Harmattan AI's two confirmed Programs of Record (French MoD and UK MoD) is not publicly disclosed; both are described as "multi-million" without further quantification. | 高 | SI001, SI018, SI005, SI009 |
| CI030 | The Le Grand Continent CEO interview, the most substantive public disclosure available, addresses supply chain, geopolitical strategy, and industrial policy but contains no revenue, margin, burn, or specific financial metrics. | 高 | SI008, SI003 |
| CI031 | Harmattan AI's French SAS legal structure entails no statutory obligation to publish annual financial statements publicly, consistent with the complete absence of any publicly available P&L or balance sheet for the company. | 中 | SI012, SI003 |
| CI032 | The Crunchbase organizational profile for Harmattan AI returned a 403 access error during research conducted June 2026, yielding no supplementary financial or investor data. | 中 | SI003 |
| CI033 | Harmattan AI's Programs of Record with sovereign NATO governments represent high-quality revenue anchors: France and the United Kingdom are investment-grade sovereign counterparties with low default risk and established defense procurement frameworks. | 中 | SI001, SI004, SI017 |
| CI034 | Harmattan AI's capital intensity is driven by vertically integrated manufacturing (in-house airframes, sensors, and AI), multi-country operations, and a stated target of producing "thousands of systems per month" requiring production line capex. | 中 | SI001, SI022, SI007 |
| CI035 | DroneXL reported company plans for up to 10,000 drone units per month production capacity; this figure is unverified and has not been confirmed in official Harmattan AI communications. | 低 | SI007 |
| CI036 | CEO M'Ghari confirmed in Le Grand Continent that the French MoD order was fulfilled in 14 months from founding and the UK MoD in 15 months, demonstrating delivery execution capability within the timeframes committed at contract award. | 中 | SI008 |
| CI037 | Dassault Aviation reported €6.2 billion in 2024 revenue and employs 14,600 people; as lead Series B investor and strategic partner, it provides Harmattan access to established international sales networks and defense program integration expertise. | 高 | SI002, SI004 |
| CI038 | Harmattan AI at ~$242 million total raised and $1.4 billion post-money valuation is significantly more capital-efficient than Anduril Industries ($5 billion raised by 2026) or Helsing (€600 million raised through 2026), though revenue comparisons are not available for Harmattan. | 中 | SI003, SI016 |
| CI039 | The Opex News analysis from September 2025 explicitly named supply chain resilience (sensors, electronics, communications links) and field reliability as the key tests facing Harmattan in the period following the UK MoD contract award. | 中 | SI009 |
| CI040 | No next-round trigger, financing milestone, or target revenue metric for a future Series C or debt raise has been publicly disclosed by Harmattan AI as of June 2026. | 中 | SI001, SI003 |
| CI041 | The Skyeton partnership (October 2025) integrates Harmattan's Sahara SAR payload into the Raybird UAS platform (350,000+ combat flight hours), targeting French and NATO markets; licensing revenue terms and structure are not disclosed. | 中 | SI015, SI020, SI024, SI028 |
| CI042 | Open roles at Harmattan AI as of June 2026 include production operators, line supervisors, quality technicians (Wissous), a US Manufacturing Manager (Arlington), and multiple engineering and business development roles across six countries, consistent with an active manufacturing scale-up investment. | 高 | SI013, SI014 |
| CI043 | The BCG May 2026 C-UAS white paper identifies the European C-UAS market as growing from $600M in 2025 to $5–7 billion by 2029 (58% CAGR), providing a secular demand tailwind for Harmattan's counter-UAS product lines. | 中 | SI017 |
| CI044 | The StartUS Insights Military Drones 2026 report notes the sector records an average investment value of $16.3 million per funding round, against which Harmattan's $200 million Series B represents approximately 12× the sector average round size. | 中 | SI016 |
| CI045 | Payment terms for Harmattan AI's French MoD and UK MoD Programs of Record are not publicly disclosed; based on standard DGA and UK DE&S procurement practice and the CEO's stated "deliver fast in volume" model, milestone or delivery-based payment structures are most consistent with the available evidence, but Harmattan-specific terms have not been confirmed. | 低 | SI008, SI018 |
| CE001 | Harmattan AI structures autonomy across two complementary layers: mission orchestration, which enables commanders and operators to define objectives and coordinate distributed systems under human authorization, and platform autonomy, which enables tactical-edge systems to execute assigned objectives and adapt to battlefield conditions within authorized mission parameters. | 高 | SE001, SE010 |
| CE002 | Mission orchestration enables commanders and operators to define mission objectives, maintain operational awareness, prioritize threats, and coordinate engagements across distributed systems under human authorization, forming the upper layer of the Harmattan autonomy model. | 中 | SE001 |
| CE003 | Platform autonomy enables systems at the tactical edge to execute assigned objectives, adapt to changing battlefield conditions, and sustain operational continuity within authorized mission parameters, forming the lower execution layer of the Harmattan autonomy model. | 中 | SE001 |
| CE004 | Harmattan AI's core doctrine is "coordinated saturation" — autonomous robot forces acting as one unified whole across the full operational spectrum — shifting military advantage from individual platform performance to the ability to produce and coordinate autonomous systems at mass. | 高 | SE011, SE001 |
| CE005 | Harmattan AI's product strategy rests on three design pillars: attritable (deployable in volume, resilient to loss), scalable (manufacturing capacity as the primary differentiator in the robotic-first era), and autonomous (edge inference without centralized infrastructure dependency while preserving human decision authority). | 中 | SE011 |
| CE006 | Sonora is a small unmanned aerial system designed for unit-level training and operator familiarization; it is designed for repeatable use to make training scalable and to prepare future operators to manage UAS in operational environments. | 高 | SE007, SE005 |
| CE007 | Sonora's published specifications are: flight time greater than 40 minutes; ready-for-takeoff in under 1 minute; range 2 km (1.2 mi); size 35 × 42 × 11 cm (13.7 × 11 × 4.7 in); weight including batteries under 1.8 kg (3.5 lbs); IP-rating equivalent to IP53. | 中 | SE007 |
| CE008 | Sonora's EO camera is a low-light sensor with 6× zoom (3× lossless), end-user resolution of 1,280 × 720 pixels at 30 fps. The IR camera delivers 320 × 240 pixels at 30 fps with 2-axis mechanical stabilization. The system supports night and low-light operations. | 中 | SE007 |
| CE009 | Sonora integrates with Kalahari in simulation mode, enabling operators to rehearse missions and after-action reviews within the same C2 environment used for live operations, maintaining train-to-fight continuity. | 中 | SE007, SE005 |
| CE010 | Sahara is an autonomous UAS-borne SAR (Synthetic Aperture Radar) payload designed for all-weather ISR, detecting man-made assets, performing terrain segmentation, and supporting maritime surveillance independent of weather, light, or environmental conditions. | 高 | SE008, SE013 |
| CE011 | Sahara's published technical specifications: frequency range L–S band; dual linear polarization; spatial resolution 0.25 m at 2 km and 1.2 m at 10 km; physical size 15 × 72.3 × 15 cm (5.9 × 28.5 × 5.9 in); weight under 3.5 kg (8 lbs); power consumption under 65 W; operating modes Stripmap and circular. | 中 | SE008, SE020 |
| CE012 | Sahara implements onboard change-detection algorithms including amplitude change detection (ACD), multi-temporal coherence (MTC), and coherent change detection (CCD), delivering real-time actionable intelligence without cloud dependency. | 中 | SE008 |
| CE013 | Sahara's onboard processing accelerates terrain analysis, object detection, and mission-relevant insights to support faster operational decision-making; the system detects manufactured objects, terrain modifications, vehicle movement, and maritime assets under camouflage. | 中 | SE008 |
| CE014 | Harmattan AI CEO Mouad M'Ghari claimed in a February 2026 Le Grand Continent interview that Sahara is "unique in the world" as a SAR radar for drones weighing less than 150 kg, enabling reconnaissance through clouds and fog without external processing. | 中 | SE024, SE020 |
| CE015 | Gobi is a high-speed UAS designed for kinetic interception of Group 1–2 hostile UAS. Published specifications: cruise speed 250 km/h (155 mph); maximum speed 350 km/h (217 mph); kinetic energy at impact greater than 10,300 J; ready-for-launch 0 seconds (always-on in launch box); size 32 × 34 × 34 cm; weight including batteries 2.2 kg. | 中 | SE002, SE012 |
| CE016 | Gobi is stored in an always-on state within its launch box, remaining continuously ready for immediate deployment. Following launch authorization, Gobi can neutralize a target in approximately one minute from launch, enabling critical threat engagement with minimal collateral damage. | 中 | SE002 |
| CE017 | Gobi performs fully autonomous hit-to-kill interception against aerial threats during the terminal engagement phase, with onboard sensing and computer vision refining the impact solution in real time without ground dependency, while maintaining human-in-the-loop authorization at the launch decision stage. | 中 | SE002, SE012 |
| CE018 | Gobi Tempest is an autonomous interceptor for Group 2–3 aerial threats. Published specifications: cruise speed 215 km/h (134 mph); maximum speed 310 km/h (193 mph); warhead compatibility 800 g; flight range 12 km (7.5 mi); ready-for-takeoff 5 seconds; size 46.5 × 46.5 × 69.5 cm; weight including batteries and warhead 4.4 kg. | 中 | SE003 |
| CE019 | Gobi Tempest features a qualified and reversible safety and arming unit (SAU), supports coordinated space operations, operates day and night in all weather conditions, and is described as human-in-the-loop. The complete system includes launchbox, ground radar, EO/IR turret, and C2 integration. | 中 | SE003 |
| CE020 | Barkhan is an autonomous strike platform for rapid target engagement. Published specifications: speed 120 km/h (75 mph); flight range 20 km (12.4 mi); size 43 × 43 × 40 cm; airframe weight under 2 kg (4.4 lbs); payload capacity 3 kg (6.6 lbs); flight time 25 minutes; radio link 160–2,500 MHz. | 中 | SE004, SE006 |
| CE021 | Barkhan supports GNSS-denied navigation, autonomous flight plan execution, autonomous terminal guidance, and optic fiber compatibility. It can be deployed standalone or in coordinated swarm operations, and operates day and night. | 中 | SE004 |
| CE022 | Kalahari is a command and control system with embedded autonomy that enables intent-driven tasking, real-time decision support, and closed-loop execution across distributed unmanned systems, serving as the orchestration layer for all Harmattan AI platform modules. | 高 | SE009, SE010 |
| CE023 | Kalahari serves three user roles with distinct authority scopes: commanders (theater-wide visibility, mission objectives, rules of engagement, centralized authority); operators (sector-scoped visibility, real-time coordination within commander-defined framework); trainers (simulation mode only, disconnected from live systems for rehearsal). | 中 | SE009 |
| CE024 | Kalahari's deployment architecture is API-driven and software-defined, supporting on-premise or hybrid edge-to-cloud deployment, with an agnostic digital backbone that unifies any asset for multi-mission operations and enables on-the-fly software adaptation to operational contingencies. | 中 | SE009 |
| CE025 | Kalahari's interoperability layer is built on open and defense-grade standards: Asterix (air traffic surveillance data exchange), Cursor on Target (CoT, NATO situational awareness), Lattice (Anduril sensor-fusion and C2 API standard), and Sapient (NATO AI/autonomous systems interchange standard). | 中 | SE009 |
| CE026 | Kalahari implements end-to-end encryption, role-based access controls (RBAC), a full audit trail, and is described as hardened to defense cybersecurity standards, though no specific standard (ISO 27001, CMMC, ANSSI) or independent audit is cited. | 中 | SE009 |
| CE027 | Kalahari includes embedded simulation and a data fabric for training, mission replay, and after-action review, enabling train-to-fight continuity across all Harmattan platform modules within the same C2 environment. | 中 | SE009, SE005 |
| CE028 | The January 2026 Dassault Aviation Series B partnership includes a commitment to embed Harmattan AI's autonomous capabilities into Dassault's future combat air systems—Rafale F5 and the UCAS programme—particularly for control of unmanned aerial systems. No production-qualified integration timeline has been publicly announced. | 中 | SE017 |
| CE029 | Harmattan AI stated in its January 2026 Series B announcement that it delivers thousands of systems per month; the company has not named a production facility or published an independently verifiable production count or delivery schedule. | 低 | SE017 |
| CE030 | The Skyeton-Harmattan partnership (announced October 2025) integrates Harmattan AI's Sahara SAR sensor and AI technologies into Skyeton's Raybird UAS platform, with initial operational deployments planned for Q1 2026 in global markets including France and NATO allies. | 中 | SE014, SE022, SE023 |
| CE031 | Skyeton's Raybird UAS has accumulated over 350,000 combat flight hours with Ukrainian defense forces, providing the most significant operational credentialing proxy for Sahara's integration into a battle-proven platform. | 中 | SE022, SE021 |
| CE032 | Harmattan AI exhibited at World Defense Show 2026 (Riyadh, February 2026), UMEX 2026 (Abu Dhabi, January 2026), and SOF Week 2026 (Tampa, May 2026), and has confirmed presence at Eurosatory 2026 (Paris, June 2026), demonstrating active multi-region defense market development. | 高 | SE015, SE016 |
| CE033 | Kalahari's scalability is described as an agnostic digital backbone capable of unifying any asset type for multi-mission operations, with on-the-fly software adaptation to operational contingencies and support for any unmanned system regardless of manufacturer. | 中 | SE009 |
| CE034 | Kalahari performs real-time ingestion and processing of multi-source sensor data with track classification, continuously evaluating mission priorities and available assets to coordinate responses across distributed systems while reducing operator workload under high-tempo conditions. | 中 | SE009 |
| CE035 | All Harmattan AI hardware systems (Gobi, Gobi Tempest, Barkhan, Sahara, Sonora) are described as agnostic and integrable with any C2 platform, with Kalahari as the recommended orchestration layer but not a mandatory dependency. | 中 | SE002, SE003, SE004, SE008 |
| CE036 | Barkhan performs autonomous reconnaissance of the operational area, identifying and validating hostile targets through onboard sensing and mission intelligence inputs before autonomous strike execution, with terminal guidance continuously refining the strike solution without ground dependency. | 中 | SE004 |
| CE037 | Gobi and Gobi Tempest are sold as complete systems including the launchbox, ground radar, EO/IR turret, and C2 integration module, making them integrated kinetic C-UAS solutions rather than standalone interceptors. | 中 | SE002, SE003 |
| CE038 | Sahara processes SAR data onboard and delivers actionable intelligence to operators and command systems in real time, operating without cloud dependency even in environments where network connectivity is degraded or denied. | 中 | SE008 |
| CE039 | Shield AI's V-BAT maritime ISR UAS provides a documented ViDAR sensor coverage rate of 3,140 NM²/hr—claimed by Shield AI as over 2.5 times that of its nearest competitor— and requires ship-based or maritime infrastructure for launch and recovery. | 中 | SE026 |
| CE040 | Quantum Systems' Vector AI sUAS uses dual Jetson Orin edge-compute boards enabling GPS/GNSS-denied autonomous navigation, has been combat-tested in Ukraine, and offers 180+ minutes fixed-wing endurance; it is an ISR-only platform with no strike, kinetic C-UAS, or C2 orchestration capability comparable to Harmattan's stack. | 中 | SE027 |
| CE041 | In a February 2026 Le Grand Continent interview, Harmattan AI CEO M'Ghari described the Sahara SAR as a system "unique in the world" for drones under 150 kg. The militarnyi.com coverage independently summarized this claim with corroborating specification data from the company website (0.25 m at 2 km, <3.5 kg), but provided no independent benchmark. | 中 | SE020, SE024 |
| CE042 | TEKEVER's AR5 is a long-endurance maritime ISR UAS operating under a €30 million EMSA framework contract for European maritime surveillance. TEKEVER has no kinetic C-UAS interceptor, no AI-driven C2 orchestration platform analogous to Kalahari, and no autonomous strike capability. | 中 | SE028 |
| CE043 | Harmattan AI's open-roles page as of June 2026 lists active positions including GNC (guidance, navigation, and control) Engineer in Lausanne and RF Hardware Engineer in Paris, confirming a real R&D engineering organization with active hardware development across navigation and sensing disciplines. | 中 | SE019 |
| CE044 | Barkhan is described as capable of operating in coordinated swarm formations as well as standalone, giving it the ability to execute mass-precision engagement across distributed targets under Kalahari orchestration. | 中 | SE004, SE006 |
| CE045 | Harmattan AI operates no publicly accessible GitHub repository, open-source software package, API documentation portal, or developer SDK as of June 2026. The absence of any public developer surface for a company whose core value proposition is AI-embedded autonomy software makes independent software-maturity assessment impossible without non-public access. | 中 | SE019, SE018 |
| CE046 | The NATO Support and Procurement Agency C-UAS Capabilities 2025 reference document establishes NATO requirements for counter-UAS kinetic intercept across Group 1–3 threat classes, providing a standards framework against which Harmattan AI's Gobi (Group 1–2) and Gobi Tempest (Group 2–3) systems must be evaluated for NATO programme qualification. | 中 | SE033 |
| CU001 | The French Ministry of Defence (DGA) awarded Harmattan AI a Program of Record contract in July 2025 for 1,000 AI-enabled micro-drones. | 高 | SU005, SU018, SU016 |
| CU002 | The UK Ministry of Defence awarded Harmattan AI a Program of Record contract in September 2025 for up to 3,000 autonomous systems. | 高 | SU016, SU018, SU023 |
| CU003 | Harmattan AI has been awarded multiple Programs of Record by the French and UK Ministries of Defence, per its own press release corroborated by Dassault Aviation. | 高 | SU013, SU015 |
| CU004 | Dassault Aviation is both a Series B lead investor and a strategic customer of Harmattan AI for embedded AI in the Rafale F5 and future UCAS programs. | 高 | SU015, SU016, SU017 |
| CU005 | Skyeton is a channel partner of Harmattan AI, integrating the Sahara SAR radar into the Raybird UAS for deployment in French and global NATO markets. | 高 | SU006, SU007, SU014 |
| CU006 | Harmattan AI has deployed its technology with multiple NATO and allied partners, per the January 2026 Dassault press release. | 中 | SU013, SU015 |
| CU007 | Harmattan AI exhibited at World Defense Show 2026 (Riyadh, February 2026), targeting Middle East defense authorities. | 中 | SU003 |
| CU008 | Harmattan AI exhibited at UMEX 2026 (Abu Dhabi, January 2026), targeting UAE and regional defense leaders. | 中 | SU004 |
| CU009 | The French MoD contract required Harmattan AI to outbid multiple European competitors in a DGA competitive evaluation with approximately 20 technical specifications. | 中 | SU005, SU018 |
| CU010 | Harmattan AI's customer base is exclusively government and institutional; no commercial or dual-use customers have been publicly disclosed as of June 2026. | 中 | SU021, SU013 |
| CU011 | The French MoD contract (SORONA) covers deliveries scheduled for October through December 2025 per OpexNews reporting. | 中 | SU018, SU005 |
| CU012 | The Sonora (SORONA) drone weighs under 1.8 kg, offers >40 minutes endurance, 2 km range, and integrates a Lynred infrared camera. | 高 | SU018, SU016 |
| CU013 | The French MoD contract was described as "multi-million euros"; the exact contract value was not publicly disclosed. | 中 | SU005, SU018 |
| CU014 | The UK MoD contract was announced at DSEI London in September 2025 and described as a multi-million pound program covering up to 3,000 systems for near-term deployment. | 高 | SU016, SU018, SU023 |
| CU015 | The UK MoD contract for 3,000 systems was Harmattan AI's second Program of Record in less than 18 months of founding. | 高 | SU016, SU025 |
| CU016 | Harmattan AI claims to be delivering thousands of systems each month as of January 2026. | 低 | SU013, SU028 |
| CU017 | Dassault Aviation CEO Eric Trappier publicly endorsed the Harmattan partnership in a press statement, and French President Emmanuel Macron praised the announcement on X. | 高 | SU015, SU017, SU016 |
| CU018 | The Dassault partnership focuses on developing embedded AI for Rafale F5 (targeted ~2030) and the future UCAS unmanned combat air system, not current production deployment. | 高 | SU015, SU016, SU024 |
| CU019 | The Dassault partnership incorporates a strategy of "sovereign, controlled and monitored AI" integration into combat systems. | 高 | SU015, SU013 |
| CU020 | Avions Légendaires reported that the Gobi counter-UAS interceptor was "acquired by the British Army," suggesting UK MoD procurement beyond the Sonora-type system. | 低 | SU010 |
| CU021 | Harmattan AI stated in the Dassault press release that "multiple Programs of Record" had been awarded by both the French and UK Ministries of Defence. | 高 | SU013, SU015 |
| CU022 | Harmattan AI will exhibit at Eurosatory 2026 (Paris, June 15–19, Hall 5B E430), its home-market flagship European defense show. | 中 | SU001 |
| CU023 | Harmattan AI exhibited at SOF Week 2026 (Tampa, FL, May 18–21, Level 3-2055), targeting US SOF forces and allied partners. | 中 | SU002 |
| CU024 | Skyeton's Raybird UAS has accumulated more than 350,000 combat flight hours and is widely deployed with Ukrainian defense forces. | 高 | SU006, SU007, SU014 |
| CU025 | Militarnyi reported that Ukrainian drones will be equipped with Harmattan AI's Sahara SAR radars for all-weather, cloud-penetrating reconnaissance. | 中 | SU008, SU022 |
| CU026 | The Skyeton CEO Pavlo Shevchuk confirmed the partnership publicly, citing "reinforcing Europe's defense capabilities and accelerating Raybird's adoption in key NATO markets such as France." | 高 | SU006, SU007, SU014 |
| CU027 | First operational deployments of the Harmattan AI-equipped Raybird were expected in Q1 2026 per third-party analysis. | 低 | SU011, SU022 |
| CU028 | DefenceJobs.org profile lists French Ministry of Defence, UK Ministry of Defence, Dassault Aviation, and Skyeton as Harmattan AI's named customers and partners. | 中 | SU021 |
| CU029 | Programs of Record in France and the UK imply multi-year budget authorization; Harmattan's contracts carry this designation, which is structural evidence of durability. | 中 | SU013, SU015, SU016 |
| CU030 | No public NRR, GRR, or churn data has been disclosed by Harmattan AI for any of its government contracts as of June 2026. | 高 | SU013, SU021 |
| CU031 | The Dassault Aviation partnership opens approximately eight to ten Rafale-nation markets (Qatar, India, Egypt, UAE, Greece, Croatia, Serbia, Colombia, Indonesia) as secondary sales channels through Dassault's international network. | 中 | SU015, SU024 |
| CU032 | The Dassault partnership press release states that Dassault will "support international business development through its established network," creating an indirect distribution channel for Harmattan. | 高 | SU015, SU013 |
| CU033 | A new electronic warfare (EW) product line and the Barkhan precision-strike platform represent potential expansion into new product domains for existing MoD customers. | 中 | SU013, SU028 |
| CU034 | Harmattan AI's presence at Gulf-state exhibitions (WDS Riyadh, UMEX Abu Dhabi) signals intent to penetrate Middle East sovereign defense markets, but no contract has been publicly announced. | 高 | SU003, SU004 |
| CU035 | SOF Week 2026 participation signals Harmattan AI's entry into the US special-operations market, though no US contract has been publicly announced. | 中 | SU002 |
| CU036 | The Kalahari C2 platform and Barkhan strike system provide cross-selling opportunities to existing MoD customers who already operate Sonora and Gobi. | 中 | SU028, SU013 |
| CU037 | Harmattan AI will present at Eurosatory 2026 (June 15–19, Paris Nord Villepinte) "airborne intelligence, tactical sensing, counter-UAS defense, and autonomous mission management" products to armed forces and government delegations. | 中 | SU001 |
| CU038 | Harmattan AI presented at SOF Week 2026 (May 18–21, Tampa) its portfolio of "training, airborne intelligence, and counter drone defense" capabilities for the full mission lifecycle. | 中 | SU002 |
| CU039 | Harmattan AI participated in World Defense Show 2026 (Riyadh, February 8–12) to engage Middle East defense authorities with its counter-drone and ISR systems. | 中 | SU003 |
| CU040 | Harmattan AI participated in UMEX 2026 (Abu Dhabi, January 20–22) describing itself as targeting "UAE defense innovation ecosystem" and "regional sovereignty through autonomous technologies." | 中 | SU004 |
| CU041 | All publicly disclosed hardware revenue for Harmattan AI derives from two sovereign governments (France and UK), constituting a critical customer concentration risk. | 高 | SU021, SU013, SU026 |
| CU042 | The Dassault Aviation partnership as both investor and strategic customer creates a potential conflict of interest or exclusivity constraint that could limit Harmattan AI's ability to sell to non-Dassault-aligned markets. | 低 | SU015, SU024 |
| CU043 | No follow-on order, delivery confirmation, or post-delivery utilization data has been publicly disclosed for the French MoD Sonora contract as of June 2026. | 高 | SU021, SU013 |
| CU044 | No operator-level testimonials from French Army or British Army end-users of Harmattan AI systems are publicly available as of June 2026. | 高 | SU021, SU013 |
| CU045 | Reuters November 2025 reporting on French drone makers highlighted production scaling, supply chain, and procurement timing risks applicable to Harmattan AI's dependency on government customers. | 中 | SU026 |
| CU046 | The DroneXL analysis noted that Harmattan AI's $1.4 billion valuation "might seem aggressive" given no second procurement cycle has been independently validated. | 中 | SU024 |
| CU047 | Harmattan AI CEO M'Ghari stated in a February 2026 Le Grand Continent interview that "technological superiority does not necessarily produce operational superiority," signaling awareness of deployment and customer-adoption complexity. | 中 | SU019 |
| CU048 | The Reuters June 2026 reporting on drone startup risks (Anduril case) highlights that defense-tech program-of-record contracts can face operational deployment failures, reputational risks, and customer-relationship strains when autonomous systems scale rapidly. | 中 | SU027 |
| CU049 | Compared with defense-AI peers like Helsing and defense hardware companies generally, Harmattan AI's public retention transparency is consistent with early-stage defense hardware companies operating under government NDA constraints; no competitor discloses NRR for MoD contracts. | 中 | SU012, SU021 |
| CU050 | Harmattan AI raised approximately $242M in total lifetime capital through January 2026 (seed + ~$30M Series A + $200M Series B) and secured at least 4,000 systems committed across three Programs of Record. | 高 | SU013, SU015, SU017 |
| CU051 | NATO maintains an active institutional procurement and experimentation pathway for autonomous drone and counter-UAS systems, evidenced by the July 2025 SAPIENCE competition, the March 2026 Latvia counter-drone innovation range, and NCIA industry testing exercises. | 高 | SU030, SU031, SU032 |
| CU052 | The NATO Support and Procurement Agency publishes a C-UAS Capabilities catalogue that identifies member-nation systems as part of allied procurement coordination, creating a formal channel through which Harmattan AI's Gobi and Kalahari products could be evaluated. | 中 | SU033 |
| CU053 | The ICRC has called for new international legal rules on autonomous weapon systems, citing concerns about meaningful human control; this creates regulatory and reputational risk for sovereign government customers procuring AI-enabled autonomous systems. | 中 | SU035 |
| CU054 | NATO maintains a policy hub on autonomous weapons and AI, providing the institutional framework through which allied governments evaluate and procure autonomous defense systems like those built by Harmattan AI. | 中 | SU034 |
| CU055 | SIPRI research on autonomous weapon systems documents growing sovereign government procurement interest across NATO and allied nations, validating the TAM for Harmattan AI's products. | 中 | SU037 |
| CU056 | EU Regulation 2021/821 on dual-use export controls governs cross-border sales of autonomous drone systems; Harmattan AI's expansion into non-EU NATO and Middle East markets requires compliance with this framework. | 中 | SU038 |
| CR001 | Harmattan AI's privacy policy (November 2025) states personal data may be transferred outside the EEA under EU Standard Contractual Clauses. | 中 | SR002, SR001 |
| CR002 | Harmattan AI's terms of use explicitly prohibit use of site content for competitive analysis or to build a competing service. | 中 | SR001, SR014 |
| CR003 | As of June 2026, Harmattan AI has not published any AI ethics charter, LAWS compliance statement, or IHL compliance documentation for its autonomous weapons systems. | 中 | SR018, SR007 |
| CR004 | Harmattan's Gobi counter-UAS system is designed to operate without requiring a human in the loop for each intercept, using computer vision and autonomous guidance. | 中 | SR018, SR012 |
| CR005 | France has not adopted a national moratorium on Lethal Autonomous Weapons Systems (LAWS), leaving Gobi's no-HITL intercept architecture legally unresolved under emerging international standards. | 中 | SR011, SR034, SR037 |
| CR006 | Harmattan AI CEO stated in February 2026 that supply-chain independence from export licenses and components is critical when seeking to produce at large volumes. | 中 | SR005, SR032 |
| CR007 | EU Regulation 2021/821 on dual-use export controls applies to autonomous drone subsystems and SAR radar technology exported commercially from France. | 高 | SR032, SR033, SR038 |
| CR008 | NATO TIE 26 (May 2026) tests C-UAS and UAS interoperability for inclusion in NATO's defence architecture, requiring track stability, sensor-to-C2 integration, and engagement-chain effectiveness. | 高 | SR008, SR009 |
| CR009 | Harmattan AI systems are not publicly confirmed as TIE 26 participants as of June 2026, leaving their NATO interoperability status unverified. | 中 | SR008, SR011 |
| CR010 | The French DGA's simplified 2025 call for tenders used only 20 technical requirements, enabling Harmattan to win the 1,000-drone contract through an urgency-procurement mechanism. | 中 | SR006, SR004 |
| CR011 | BCG's May 2026 C-UAS whitepaper states that European C-UAS systems are dependent on supply chains not built for sustained attrition. | 高 | SR013, SR032 |
| CR012 | Harmattan uses a Lynred-supplied infrared sensor in the SORONA drone, demonstrating partial sovereign French sourcing for this key sensing component. | 高 | SR007, SR006 |
| CR013 | Harmattan AI's two Programs of Record require delivery of approximately 4,000 systems total: 1,000 to French MoD (by end 2025) and up to 3,000 to UK MoD on a near-term schedule. | 中 | SR006, SR004, SR023 |
| CR014 | No public evidence exists of Harmattan's manufacturing facility location, production capacity figures, quality management certification, or batch-production test results. | 中 | SR018, SR022 |
| CR015 | DroneXL reported that Harmattan AI is planning facilities capable of producing 10,000 drones monthly, requiring supply chains, skilled workers, and testing infrastructure. | 中 | SR003, SR007 |
| CR016 | Reuters reported in June 2026 that drone crashes and safety incidents at a US defense drone startup illustrate real operational safety risk from rapid scaling in the sector. | 中 | SR030, SR015 |
| CR017 | The MarketsandMarkets 2026 Europe UAV report identifies GPS resilience as a mandatory design requirement derived from Q1 2026 jamming data, adding development complexity and cost. | 中 | SR031, SR032 |
| CR018 | Harmattan's CEO acknowledged in the Le Grand Continent interview that GPS-denial and radio-wave jamming are operational realities requiring embedded AI without connectivity dependence. | 中 | SR005, SR018 |
| CR019 | Expert Market Research's 2026 Europe UAV report notes European UAV exports require coordination between national export licensing and EU common position frameworks. | 中 | SR032, SR033 |
| CR020 | The October 2025 Skyeton-Harmattan partnership integrating Sahara SAR into the Raybird creates geopolitical sensitivity for export licensing in third markets and re-export complications. | 中 | SR019, SR020, SR029 |
| CR021 | Dassault Aviation is simultaneously Harmattan AI's Series B lead investor, primary strategic partner for Rafale F5 / UCAS AI, and future customer — a triple-role concentration with no disclosed fallback. | 高 | SR012, SR007, SR021 |
| CR022 | All publicly disclosed revenue-generating customers of Harmattan AI are government entities: the French Ministry of Defence and the UK Ministry of Defence only. | 中 | SR018, SR023, SR012 |
| CR023 | Harmattan's CTO wrote at the Series B that 'the international order is going off the rails,' framing that could complicate US or Gulf market development if amplified in sales contexts. | 中 | SR021, SR022 |
| CR024 | Skyeton's ability to support the Raybird partnership could be disrupted by Russian strikes on Ukrainian industrial facilities, adding timeline uncertainty to Harmattan's SAR integration program. | 中 | SR020, SR019 |
| CR025 | France, Germany, and Spain remain at impasse on FCAS, reflecting EU defense-industrial fragmentation that limits pan-European procurement vehicles for Harmattan. | 中 | SR007, SR005 |
| CR026 | Harmattan's CEO stated in Le Grand Continent that the European Commission lacks a defence strategy mandate and concrete European strategic autonomy actions 'are still in their infancy.' | 中 | SR005, SR034 |
| CR027 | Defense News described Harmattan's workforce as unusually senior, with a claimed median experience of 15 years, implying higher personnel-cost rigidity and retention risk than a junior-skewed startup workforce. | 中 | SR007, SR018 |
| CR028 | Mouad M'Ghari (CEO) is the sole named executive in all major press releases and investor announcements; no CFO, COO, or head of operations has been publicly named. | 高 | SR021, SR012, SR007 |
| CR029 | No external board composition has been disclosed for Harmattan AI; investor board-seat rights are unconfirmed in any public source reviewed. | 中 | SR021, SR022 |
| CR030 | Harmattan AI has not made any public financial disclosure (audited accounts, revenue, burn rate, or EBITDA) since its founding in April 2024. | 中 | SR021, SR022 |
| CR031 | Open roles at Harmattan AI as of June 2026 include Chief Engineer, Electronic Warfare Team Lead, and System Architect — senior technical positions indicating gaps in the engineering leadership pipeline. | 中 | SR018, SR022 |
| CR032 | Defense News reported Harmattan has hired executives from Safran and Isar Aerospace, broadening its senior talent base, but no C-suite positions beyond co-founders have been publicly named. | 中 | SR007, SR027 |
| CR033 | Anduril raised $5 billion in its Series H at a $61 billion valuation in May 2026, having doubled revenue to $2.2 billion in 2025. | 高 | SR015, SR043 |
| CR034 | Shield AI raised $2 billion in aggregate in March 2026 at a $12.7 billion post-money valuation, led by Advent International and JPMorganChase. | 高 | SR016, SR015 |
| CR035 | Helsing is reportedly raising $1.2 billion at approximately $18 billion valuation led by Dragoneer and Lightspeed, per TechCrunch May 2026. | 中 | SR017, SR036 |
| CR036 | Both Harmattan Programs of Record were won under simplified or urgent procurement mechanisms, which may not apply under post-conflict normalized procurement cycles. | 中 | SR006, SR004 |
| CR037 | Quantum Systems raised a total of €310 million (including a €160 million Series C in May 2025) backed by Hensoldt, Airbus Defence and Space, and Balderton Capital. | 中 | SR025, SR017 |
| CR038 | Tekever raised €70 million from Baillie Gifford and the NATO Innovation Fund and is already profitable, directly contrasting with Harmattan's pre-profitability stage. | 中 | SR026, SR013 |
| CR039 | BCG's May 2026 C-UAS whitepaper projects European C-UAS market demand at $5–7 billion by 2029 at 50%+ annual growth but notes the landscape is fragmented with parallel national programs. | 高 | SR013, SR031 |
| CR040 | NATO's Latvia Innovation Range and LCI-X Beacon Project impose structured validation requirements before systems are eligible for Alliance-wide procurement, creating a certification lead-time risk. | 高 | SR009, SR011 |
| CR041 | Harmattan AI's total capital of ~$242 million represents approximately 4% of Anduril's $5 billion latest single round, indicating a material capitalization gap with US defense-AI peers. | 中 | SR021, SR015 |
| CR042 | Expert Market Research notes European UAV industry is 5–10 years behind US and Israeli equivalents in autonomy, AI integration, and production scale, requiring sustained investment. | 中 | SR032, SR031 |
| CR043 | As of June 2026, no public reports of Harmattan AI delivery failures, quality holds, or contract performance issues under the French MoD or UK MoD programs have been identified. | 中 | SR018, SR004 |
| CR044 | The ICRC has urged States to establish legally binding limits on autonomous weapon systems since 2015, citing that no-HITL systems raise humanitarian, legal, and ethical concerns including risks to civilians and challenges to IHL compliance. | 高 | SR037, SR040 |
| CR045 | Stop Killer Robots, a coalition of 250+ NGOs, advocates for a binding international ban on fully autonomous lethal weapons, representing the adverse advocacy environment that could accelerate legislative action against no-HITL systems like Gobi. | 中 | SR039, SR037 |
| CR046 | EU Regulation 2021/821 establishes the Union's export control regime for dual-use items; autonomous drone systems with SAR sensors and AI targeting capabilities are classified under dual-use categories requiring licensing for export outside the EU. | 高 | SR038, SR033 |
| CR047 | Anduril's official Series H announcement confirms doubling of revenue to $2.2 billion in 2025 and a $5 billion funding round — confirming the capitalization gap with Harmattan AI ($1.4B valuation, ~$242M raised). | 高 | SR043, SR015 |
| CR048 | Helsing's HX-2 drone is a direct European competitor to Harmattan AI's SORONA, designed for autonomous AI-enabled ISR with a vertically integrated hardware-software stack — Helsing is reportedly raising $1.2B at $18B, representing 12.9× Harmattan's valuation. | 中 | SR042, SR017 |
| CR049 | Tekever markets AR3 EVO as an expeditionary ISR platform, extending its overlap with Harmattan in European small-UAS reconnaissance tenders. | 中 | SR045 |
| CR050 | Tekever expanded its French industrial footprint through a 2026 partnership with Merio, increasing competitive pressure on Harmattan in France-adjacent ISR procurement. | 中 | SR046 |
| CR051 | Delair continues to market open-payload long-range drones for surveillance missions, showing that French ISR buyers retain domestic alternatives beyond Harmattan. | 中 | SR047 |
| CR052 | MyDefence opened a U.S. counter-drone production facility in 2026, highlighting how specialized C-UAS rivals are localizing manufacturing near buyers faster than many European startups. | 中 | SR048 |
| CR053 | MyDefence’s selection for Australia’s LAND 156 counter-drone program shows that portable EW-first C-UAS vendors can win national defense programs without offering Harmattan-style kinetic interceptors. | 中 | SR049 |
| CR054 | MyDefence’s Wingman wearable detector illustrates a softer-kill, lower-logistics alternative to hit-to-kill interception, increasing the burden on Harmattan to prove when kinetic systems outperform detection-and-jamming stacks. | 中 | SR050 |
| CV001 | The European C-UAS and autonomous systems market is projected to reach $5–7 billion by 2029, growing at over 58% per annum, according to BCG's May 2026 white paper on the European C-UAS opportunity. | 高 | SV008, SV023 |
| CV002 | European defense AI investments reached €946 million in the first half of 2025, a 26% increase year-on-year, according to DroneXL's January 2026 analysis of the Dassault-Harmattan partnership. | 中 | SV005 |
| CV003 | NATO's Technical Interoperability Exercise TIE 26, held in May 2026 in the Netherlands, tested more than 60 commercial C-UAS systems from 11 Allied Nations and explicitly validated the defense procurement pipeline for C-UAS suppliers. | 高 | SV031, SV030 |
| CV004 | Harmattan AI secured a Program of Record with the French Ministry of Defence in July 2025 for delivery of 1,000 SORONA drones by end 2025, constituting the company's first national MoD contract at approximately 15 months post-founding. | 高 | SV001, SV002, SV006 |
| CV005 | Harmattan AI secured a Program of Record with the UK Ministry of Defence in September 2025 for delivery of up to 3,000 autonomous systems for urgent operational needs, constituting the company's second national MoD contract within 18 months of founding. | 高 | SV002, SV004, SV006 |
| CV006 | Harmattan AI has disclosed no annual recurring revenue, contract value totals, gross margins, burn rate, or unit economics publicly as of June 2026; it is a private French SAS company with no mandatory financial filing obligation to public markets. | 高 | SV001, SV020 |
| CV007 | The Dassault Aviation Series B partnership targets AI development for Dassault's Rafale F5 standard and future UCAS programs, with both companies committing to "sovereign and scalable" embedded AI, according to their joint January 2026 press release. | 高 | SV015, SV014, SV002 |
| CV008 | Harmattan AI's board of directors composition has not been publicly disclosed as of June 2026; the four co-founders control executive decision-making without a publicly declared external board oversight structure. | 中 | SV020, SV001 |
| CV009 | CEO Mouad M'Ghari confirmed in a February 2026 Le Grand Continent interview that Chinese rare-earth dependencies in brushless motor magnets persist, with sovereign motor magnets expected on the market by end 2026 and sovereign battery solutions 3–5 years away. | 中 | SV007 |
| CV010 | Harmattan AI claims to produce "thousands of systems per month" but this figure has not been independently verified by any third-party source and the DroneXL analysis cited plans for up to 10,000 drones monthly — also unverified. | 中 | SV005, SV007 |
| CV011 | No adverse press reporting, investor litigation, regulatory action, or governance dispute involving Harmattan AI has been identified in any source reviewed through June 2026. | 中 | SV026, SV001 |
| CV012 | Thales' cortAIx accelerator established a separate November 2025 AI partnership with Dassault Aviation, creating a competing relationship that could limit or displace Harmattan's scope within the Rafale F5 AI mandate. | 中 | SV005 |
| CV013 | Harmattan AI's January 2026 Series B established a $1.4 billion post-money valuation, confirmed by Defense News, TechCrunch, Tech.eu, AirForce-Technology.com, and Harmattan's own press release; French President Emmanuel Macron praised the deal on X. | 高 | SV001, SV002, SV004, SV014, SV016 |
| CV014 | Harmattan AI's total lifetime capital raised through January 2026 is estimated at approximately $242 million ($42 million pre-Series B as reported by TechCrunch, plus $200 million Series B), giving a capital-raised multiple of approximately 5.8× at the $1.4 billion valuation. | 中 | SV001, SV020 |
| CV015 | Harmattan AI's Series B was led by Dassault Aviation with Motier Ventures renewing its investment; the identity of other Series B co-investors and the size of Dassault's specific tranche were not disclosed at the time of the announcement. | 高 | SV001, SV002, SV015 |
| CV016 | The Series B valuation "significantly increases" from the Series A, but the Series A post-money valuation was not publicly disclosed, making the exact step-up ratio impossible to calculate from public evidence. | 中 | SV001 |
| CV017 | Helsing, the Germany-based defense AI and autonomous drone company, was reportedly close to raising a new $1.2 billion round at approximately $18 billion valuation in May 2026, according to TechCrunch citing the Financial Times. | 高 | SV009, SV025 |
| CV018 | Anduril Industries raised a $5 billion Series H round at a $61 billion post-money valuation in May 2026, after the company doubled revenue to $2.2 billion in 2025 — the clearest revenue-disclosed comparable for a vertically integrated autonomous defense platform. | 高 | SV010, SV015, SV039 |
| CV019 | Shield AI raised $1.5 billion in Series G funding at a $12.7 billion post-money valuation in March 2026, led by Advent International, with Hivemind AI pilot software deployed across 26 vehicle classes including F-16s. | 中 | SV011, SV038 |
| CV020 | Quantum Systems raised €160 million in Series C funding in May 2025 at approximately €1 billion valuation, and Reuters reported in November 2025 that the valuation tripled to more than €3 billion following disclosed 100% annual revenue growth. | 高 | SV012, SV019 |
| CV021 | TEKEVER, the Lisbon-based ISR drone company, raised £400 million at a valuation above £1 billion in 2025 and is already profitable — the only peer in the comparable set with explicit profitability disclosure. | 中 | SV013, SV040 |
| CV022 | TEKEVER's profitability proof provides a financial floor under its valuation that Harmattan AI cannot yet claim; TEKEVER operates an Intelligence-as-a-Service model while Harmattan sells systems, creating a different revenue profile and multiple basis. | 中 | SV013, SV040 |
| CV023 | Quantum Systems' valuation tripling from €1 billion to over €3 billion was explicitly anchored in disclosed 100% annual revenue growth, demonstrating that revenue disclosure is the primary catalyst for multiple re-rating in the European defense drone peer set. | 高 | SV019, SV012 |
| CV024 | Harmattan AI's $1.4 billion valuation implies a capital-raised multiple of approximately 5.8× ($1.4B / $242M), which is below Helsing's approximately 15× and Shield AI's approximately 9.8×, suggesting the current entry is at the low end of the peer band. | 中 | SV009, SV011, SV001 |
| CV025 | Anduril's capital-raised multiple at Series H is approximately 5.5× ($61B valuation / ~$11B total raised) with $2.2 billion in disclosed revenue, implying a revenue multiple of approximately 28× — the ceiling reference for Harmattan's fully-scaled comparable. | 中 | SV010, SV039 |
| CV026 | French President Emmanuel Macron endorsed Harmattan AI's Series B on X, calling it "excellent news for our strategic autonomy" — a political valuation floor that may provide downside protection in sovereign acquisition scenarios but is not fungible for financial return models. | 高 | SV001, SV002, SV015 |
| CV027 | The Dassault Aviation partnership creates a structural replication barrier for Harmattan AI because Dassault's systems architecture experience and mission-system integration expertise for the Rafale platform are unique industrial assets that cannot be replicated by other AI software suppliers in the near term. | 中 | SV015, SV007, SV005 |
| CV028 | The Rafale F5 standard is targeted for approximately 2030 and is designed to incorporate unmanned combat air systems; Harmattan's AI integration mandate for this program creates a 4–5 year development pipeline anchored to France's premier defense program. | 高 | SV002, SV015, SV021 |
| CV029 | Harmattan AI's vertically integrated platform spanning ISR, C-UAS, electronic warfare, strike, and C2 positions it to address four distinct defense segments simultaneously — a scope that narrow-focus peers like TEKEVER (ISR-only) or Quantum Systems (hardware-ISR) do not match. | 高 | SV024, SV014, SV004 |
| CV030 | Helsing has 5 years of operation vs. Harmattan's 2 years at broadly comparable strategic profiles (European sovereign AI + autonomous drones); Harmattan's faster path to dual national MoD Programs of Record suggests a positive age-adjusted trajectory if execution continues. | 中 | SV009, SV001 |
| CV031 | Reuters reported in November 2025 that French drone makers face financing uncertainty even as they eye Ukraine deals, providing adverse-context evidence that French defense startup valuations face scrutiny from investors concerned about undisclosed revenue. | 中 | SV035, SV026 |
| CV032 | Reuters reported in June 2026 on drone crashes and safety incidents at a $1.3 billion Silicon Valley defense startup, providing a cautionary precedent for how rapidly strategic premiums deflate when operational failures surface. | 中 | SV026 |
| CV033 | TEKEVER's profitability and operational deployment in Ukraine across thousands of combat hours distinguishes it from Harmattan and provides the clearest "what financial proof looks like" comparable for an early-stage European defense ISR company. | 中 | SV013 |
| CV034 | Harmattan AI's Skyeton partnership, announced October 2025, provides market access to Ukraine and NATO partners and demonstrates multi-country commercial development; it does not constitute an independent financial validation of Harmattan's revenues. | 中 | SV027, SV033 |
| CV035 | Quantum Systems' 100% annual revenue growth disclosed at Series C is the most directly comparable financial metric to what Harmattan AI would need to disclose to justify a comparable step-up in valuation from $1.4 billion to $3–5 billion at Series C. | 中 | SV012, SV019 |
| CV036 | The bull scenario assumption requires Harmattan AI to secure two or more additional NATO member Programs of Record by end 2027 and confirm Rafale F5 AI integration as a primary mandate, driving a valuation re-rating to $4–8 billion by 2028. | 低 | SV005, SV007, SV015 |
| CV037 | The base scenario assumes one additional European NATO Program of Record and stable delivery of existing French and UK contracts, supporting a valuation range of $2.5–4 billion by 2028–2029 with partial revenue disclosure as a potential catalyst. | 低 | SV001, SV019, SV012 |
| CV038 | The bear scenario assumes an operational failure or Rafale F5 AI mandate loss to a competitor (Thales cortAIx), leading to valuation compression toward the capital-raised anchor of $800 million to $1.2 billion. | 低 | SV026, SV005, SV035 |
| CV039 | Harmattan AI exhibited at UMEX 2026 in Abu Dhabi (January 2026), World Defense Show 2026 in Riyadh (February 2026), SOF Week 2026 in Tampa (May 2026), and will exhibit at Eurosatory 2026 in Paris (June 2026) — signaling international market development rather than near-term exit preparation. | 高 | SV028, SV029, SV032 |
| CV040 | Dassault Aviation's equity stake in Harmattan AI creates a logical strategic acquisition path if Harmattan's AI capabilities become integral to Dassault's F5 platform and UCAS program; this is the most probable structured exit in the near-to-medium term. | 低 | SV015, SV005 |
| CV041 | Harmattan AI's French SAS legal structure subjects the company to French commercial law and CMAP arbitration; no public filing obligation exists and the company has no stated IPO timeline, making a public market exit unlikely within 3 years. | 高 | SV024, SV001 |
| CV042 | The investment recommendation is Conditional Watch / Research More: the evidence supports a strategic premium at $1.4 billion but revenue opacity, governance gaps, and unverified production scale prevent a confident Buy recommendation as of June 2026. | 中 | SV001, SV007, SV026 |
| CV043 | The European Parliament Think Tank confirmed in a June 2025 study that military drone systems in the EU have shifted from experimental to operational baseline capability, validating the secular tailwind underpinning Harmattan AI's strategic premium. | 中 | SV034 |
| CV044 | Harmattan AI's US team expansion in Arlington, Virginia creates potential ITAR/EAR export control exposure if US-origin technology is integrated into systems sold to non-US NATO customers, representing a material legal risk not assessed in public sources. | 中 | SV024, SV002 |
| CV045 | Harmattan AI disclosed at Eurosatory 2026 (June 2026) and SOF Week 2026 (May 2026) that it is actively expanding its international customer footprint; the Eurosatory exhibit in Paris is the company's home-turf flagship European land defense show debut. | 高 | SV028, SV029 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Harmattan AI | Harmattan AI — Homepage | Harmattan AI is redefining how forces are built and deployed, shaping a future where autonomous robots fight as one across the full spectrum of operations. |
| SO002 | Harmattan AI | Autonomy — Harmattan AI | |
| SO003 | Harmattan AI | Capabilities — Harmattan AI | |
| SO004 | Harmattan AI | VSHORAD — Harmattan AI | |
| SO005 | Harmattan AI | Sonora — Harmattan AI | Flight Time > 40 min; Ready for take-off < 1 min; Range 2 km; Weight (incl. batteries) < 1.8 kg. |
| SO006 | Harmattan AI | Sahara SAR Payload — Harmattan AI | Spatial Resolution 0.25 m at 2 km; Weight < 3.5 kg. |
| SO007 | Harmattan AI | Kalahari C2 Platform — Harmattan AI | Interoperability: Built on open and defense-grade standards: Asterix, CoT, Lattice, Sapient, NATO state-of-the-art standards. |
| SO008 | Harmattan AI | Blog — Harmattan AI | |
| SO009 | Harmattan AI | Harmattan AI's $200 million Series B led by Dassault Aviation | Founded in 2024, the company has rapidly become one of the fastest-growing players in the sector, delivering thousands of systems each month. |
| SO010 | Harmattan AI | Careers — Harmattan AI | Where we work: United Kingdom, France, Switzerland, Morocco, United Arab Emirates, United States of America. |
| SO011 | Harmattan AI | Open Roles — Harmattan AI | |
| SO012 | Harmattan AI | Privacy Policy — Harmattan AI | We are committed to protecting your personal information and your right to privacy in accordance with GDPR, CCPA/CPRA, and CalOPPA. |
| SO013 | Harmattan AI | Terms of Use — Harmattan AI | Harmattan AI, a simplified joint-stock company registered in France with its registered office at 1 rue du Mail, 75002 Paris, France and VAT number FR39978035392. |
| SO014 | Harmattan AI | Harmattan AI & Skyeton Announce Strategic Partnership to Advance UAS Capabilities | Founded in 2024, the company has rapidly become one of the fastest-growing players in the sector, delivering thousands of systems each month. |
| SO015 | Dassault Aviation | Harmattan AI's $200 Million Series B Led by Dassault Aviation — Press Kit | Dassault Aviation has always placed technological excellence and sovereignty at the heart of its values. This partnership with Harmattan AI reflects our commitment to integrating high-value autonomy into the next generation of combat air systems. |
| SO016 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn | Harmattan AI, which once described itself as 'a European Anduril,' is now also ready to partner with [defense primes] — even if it means no longer calling itself 'a next-generation defense prime.' |
| SO017 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion value | Harmattan currently has more than 130 employees, with a median experience of 15 years, according to the company, which in the past year hired executives from companies including Safran and Isar Aerospace. |
| SO018 | TechFundingNews | France's Answer to Helsing: Harmattan AI Secures $200M from Dassault Aviation | |
| SO019 | AirForce Technology | Harmattan AI raises $200M in Series B funding | In September last year, the UK Ministry of Defence awarded Harmattan AI a contract for autonomous systems under a multi-million pound programme that includes 3,000 units scheduled for near-term deployment. |
| SO020 | Tech.eu | French startup Harmattan AI hits unicorn status as it strikes deal with Dassault | Harmattan AI is led by Mouad M'ghari, alongside co-founders Martin de Gourcuff and Edouard Rosset. |
| SO021 | DroneXL | Dassault's Harmattan: France's AI Defense | The $1.4 billion valuation might seem aggressive for a company barely 18 months old, but it reflects reality: Harmattan has actual military contracts, actual battlefield-proven technology, and now actual backing from the company building France's next-generation fighter jets. |
| SO022 | Army Recognition | France's Massive 1,000 AI Drone Deal to Transform Military Training and Deterrence | Established in April 2024, the company positioned itself as a disruptor by outbidding larger competitors across Europe. |
| SO023 | OpexNews.fr | Harmattan AI — Commande UK 3 000 Systèmes Autonomes | Reste l'épreuve des prochains mois: tenir la cadence, sécuriser la chaîne d'approvisionnement (capteurs, calculateurs, liaisons) et prouver la fiabilité en conditions réelles. |
| SO024 | Le Grand Continent | Entretien — Harmattan AI avec Mouad M'Ghari | C'est notamment sur le sujet des chaînes d'approvisionnement qu'il nous reste du travail à faire: les moteurs électriques brushless et les batteries restent dépendants de terres rares chinoises. |
| SO025 | DefenceJobs.org | Harmattan AI — Company Profile | Founded in Paris by Mouad M'Ghari and co-founders including Edouard Rosset, Marc Grelet and Martin de Gourcuff. Backed by Dassault Aviation, FirstMark Capital, Atlantic VC, Motier Ventures, Tholus Capital. |
| SM001 | NATO Communications and Information Agency (NCIA) | Allies and industry test the latest counter-drone technology during NATO exercise TIE 26 | "TIE 26 demonstrates that interoperability in modern multi-layered C-UAS operations is no longer optional, it is essential!" |
| SM002 | NATO | New NATO Innovation Range starts counter-drone technology testing in Latvia (March 2026 TEVV campaign) | "NATO's Innovation Range for uncrewed systems in Latvia is one of the five pilot ranges established under NATO's Rapid Adoption Action Plan (RAAP)." |
| SM003 | European Commission Joint Research Centre (JRC) / EUROCONTROL | Drone, Counter drone and autonomous systems—JRC presentation at EUROCONTROL High-Level Workshop on C-UAS (November 2024) | |
| SM004 | NATO Allied Command Transformation (ACT) | LCI-X Beacon Project builds layered C-UAS approach to fast-moving threat (Crucible 1-26, Romania, 2026) | "LCI-X moved from framework development into practical experimentation through the launch of its Crucible Series … around 500 personnel and roughly 215 technical systems." |
| SM005 | NATO | Autonomous drones take flight at NATO-backed SAPIENCE competition (2025 Huntsville competition) | |
| SM006 | European Parliamentary Research Service (EPRS) | Drones and new systems of warfare: Adapting the EU to today's security challenges (EPRS AT-A-GLANCE January 2026) | "Europe cannot afford reliance on non-EU suppliers for technologies that shape modern conflict, stressing the need for robust domestic production … to preserve real strategic independence." |
| SM007 | StartUs Insights | Military Drones Report 2026: Multi-Million Unit Demand & Alliance Re-Armament | "The global military drone market size is projected to reach USD 98.24 billion by 2033, growing at a CAGR of 8.9% from 2026 to 2033." |
| SM008 | MarketsandMarkets | Europe UAV (Drone) Market by Application—Forecast to 2030 (USD 7.98 BN, CAGR 9.8%) | |
| SM009 | Expert Market Research | Europe Unmanned Aerial Vehicle (UAV) Market Report 2026–2035 (USD 9.91B base, 6.32% CAGR) | |
| SM010 | IMARC Group | Europe Drones Market Report 2026–2034 (USD 275.6M base, 9.03% CAGR) | |
| SM011 | Grand View Research | Europe Drone Market Outlook 2025–2030 (USD 17.05B base, 13.4% CAGR) | |
| SM012 | European Defence Agency (EDA) | EDA publications and data—autonomous systems and unmanned aerial systems | |
| SM013 | European Parliament Think Tank | Military drone systems in the EU and global context: Types, capabilities and regulatory frameworks (EPRS briefing, June 2025) | "In May 2024, Ukraine's military leadership indicated that drones were the leading cause of casualties on both sides, stating that 'drones kill more soldiers than any other weapon'." |
| SM014 | Market Research Future (MRFR) | Europe Military Drone Market Size, Share, Industry Trend & Analysis—Forecast to 2035 (USD 4,650M, CAGR 7.26%) | "The Europe military drone market is projected to grow from 2306.09 USD Million in 2025 to 4650.0 USD Million by 2035, exhibiting a compound annual growth rate (CAGR) of 7.2% during the forecast period 2025–2035." |
| SM015 | MarketsandMarkets | Counter-Unmanned Aircraft Systems (C-UAS) Market by Solution, End User, Deployment, Range, Technology, and Region—Forecast to 2030 | |
| SM016 | Boston Consulting Group (BCG) | Countering the drone threat: Europe's C-UAS opportunity (White Paper, May 2026) | "In the near term, the C-UAS demand is projected to rise significantly above announced programs, reaching $5-7 billion by 2029 with an annual growth rate of over 50%." |
| SM017 | Stockholm International Peace Research Institute (SIPRI) | SIPRI Arms Transfers Database (updated March 2026, data 1950–2025) | |
| SM018 | Harmattan AI | Harmattan AI homepage—coordinated saturation and autonomous defence systems | |
| SM019 | Harmattan AI | Harmattan AI—VSHORAD capability page (Gobi, Gobi Tempest) | |
| SM020 | Harmattan AI | Harmattan AI—Persistent ISR capability page (Sahara SAR) | |
| SM021 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defence unicorn (January 2026) | |
| SM022 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion valuation (January 2026) | |
| SM023 | Army Recognition | France's massive 1,000 AI-drone deal to transform military training and deterrence (2025) | |
| SM024 | Harmattan AI | Harmattan AI $200M Series B blog announcement—Dassault Aviation partnership | |
| SM025 | Harmattan AI | Harmattan AI capabilities overview—Kalahari C2 command and control system | |
| SP001 | Anduril Industries | Transforming U.S. Defense Capabilities with Advanced Technology | |
| SP002 | Anduril Industries | Counter UAS | |
| SP003 | Anduril Industries | Roadrunner | |
| SP004 | Shield AI | V-BAT: Next-Generation UAS | V-BAT was competitively selected for the U.S. Coast Guard Maritime Unmanned Aircraft System Services program. It has deployed on nearly every class of U.S. Navy ship and with all seven Marine Expeditionary Units (MEUs). |
| SP005 | Shield AI | Hivemind EdgeOS: Next-Generation Robotics Middleware | In published benchmarks, EdgeOS has sustained millions of messages per second with sub-millisecond local latency while keeping the Central Processing Unit (CPU) and memory usage efficient. |
| SP006 | TEKEVER | TEKEVER — Unmanned Aerial Systems | |
| SP007 | TEKEVER | AR5 — Long-Endurance Fixed-Wing UAS | |
| SP008 | TEKEVER | TEKEVER Raises €70 Million and Brings Strategic Investors On Board | TEKEVER, which is already profitable, will accelerate investment in R&D to support product innovation. |
| SP009 | Quantum Systems | Vector AI — Smart Mid-Range eVTOL sUAS | Equipped with onboard computing power and AI capabilities, Vector delivers reliable performance under any circumstances while reducing the cognitive load of the operation. |
| SP010 | Quantum Systems GmbH | Quantum Systems raises €160M as it targets global leadership in aerial intelligence solutions | The new funding follows several years of exceeding 100% year-over-year revenue growth for the company, which now has 550 people across sites in Germany, Australia, Ukraine and Romania. |
| SP011 | Delair | Security and Defense Solutions | |
| SP012 | Delair | DT26X Surveillance — Commercial Surveillance Drone | |
| SP013 | MyDefence | MyDefence — Counter-UAS Solutions | 5000+ Systems deployed in the world; 100+ Customers worldwide. |
| SP014 | TechCrunch | Daniel Ek-backed defense tech Helsing to raise $1.2B at $18B valuation | Five-year-old European military drone startup Helsing is reportedly close to raising a new $1.2 billion round at about an $18 billion valuation. |
| SP015 | TechCrunch | Germany's Helsing doubles down on drones for Ukraine, scales up manufacturing | Helsing, the German defense tech startup backed by Spotify's Daniel Ek and others, is producing 6,000 HX-2 strike drones in addition to the 4,000 HF-1 strike drones financed by Germany. |
| SP016 | TechCrunch | Anduril raises $5B, doubles valuation to $61B | Anduril has raised a $5 billion Series H round at a $61 billion valuation, led by returning investors Thrive Capital and Andreessen Horowitz. |
| SP017 | Shield AI | Shield AI to Acquire Aechelon and Raise $2B at $12.7B Valuation | Today, Hivemind software has already piloted 26 classes of vehicles including F-16s, jet-powered UAVs, helicopters, drone boats, and ground vehicles. |
| SP018 | Harmattan AI | Harmattan AI's $200 Million Series B Led by Dassault Aviation | Harmattan AI has been awarded multiple Programs of Record by the French and UK Ministries of Defence for its autonomous systems portfolio and is now scaling globally. |
| SP019 | Dassault Aviation | Harmattan AI's $200 Million Series B Led by Dassault Aviation — Press Kit | This partnership will support the development of embedded AI capabilities by Harmattan AI within Dassault Aviation's future air combat systems (Rafale F5 and UCAS). |
| SP020 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion value | |
| SP021 | Airforce Technology | Harmattan AI secures $200M funding | |
| SP022 | Reuters | Drone crashes, severed fingers: inside the $13 billion Silicon Valley military startup | Drone crashes, severed fingers: inside the $13 billion Silicon Valley military startup [Reuters June 2026; full text behind paywall; title confirms safety incident reporting at a major US autonomous defense startup]. |
| SP023 | Reuters | French drone makers eye Ukraine deals amid funding uncertainty | |
| SP024 | Helsing | Helsing — European Defense AI | |
| SP025 | Helsing | Helsing raises €600M to invest in European technological sovereignty | |
| SP026 | Boston Consulting Group | Countering the Drone Threat: Europe's C-UAS Opportunity | |
| SP027 | Tech.eu | French startup Harmattan AI hits unicorn status as it strikes deal with French aerospace group | |
| SP028 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn | |
| SI001 | Harmattan AI | Harmattan AI's $200 million Series B led by Dassault Aviation — Official Blog Post | The proceeds from this investment round will be used to expand the deployment of AI-enabled missions across new operational theaters, extend Harmattan AI's product offering into new domains and scale industrial manufacturing of its ISR, drone interception, and electronic warfare platforms. |
| SI002 | Dassault Aviation | Harmattan AI's $200 million Series B led by Dassault Aviation — Press Kit | Dassault Aviation is leading Harmattan AI's $200 million Series B funding round. |
| SI003 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn | the new funding comes in addition to the $42 million Harmattan AI had raised to date, including a seed round led by Atlantic and a Series A led by FirstMark |
| SI004 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion value | Harmattan currently has more than 130 employees, with a median experience of 15 years, according to the company |
| SI005 | Airforce Technology | Harmattan AI receives $200m in Series B funding | |
| SI006 | Tech.eu | French startup Harmattan AI hits unicorn status as it strikes deal with French aerospace group | |
| SI007 | DroneXL | Dassault's Harmattan: France's AI Defense Bet | Companies like Harmattan are planning facilities capable of producing 10,000 drones monthly. That scale-up requires supply chains, skilled workers, and testing infrastructure. |
| SI008 | Le Grand Continent | Entretien — Harmattan AI: Mouad M'Ghari | Deux composants clefs de nos systèmes sont concernés: les moteurs électriques brushless, notamment les aimants présents dans ces derniers, et les batteries. Ces deux sujets sont très complexes. Les résultats devraient être appréciés d'ici trois à cinq ans. |
| SI009 | Opex News | Harmattan AI — commande UK 3000 systèmes autonomes | Reste l'épreuve des prochains mois: tenir la cadence, sécuriser la chaîne d'approvisionnement (capteurs, calculateurs, liaisons) et prouver la fiabilité en conditions réelles. |
| SI010 | TechFundingNews | France's Answer to Helsing: Harmattan AI Secures $200M from Dassault Aviation | |
| SI011 | DefenceJobs | Harmattan AI — Company Profile | Backed by Dassault Aviation, FirstMark Capital, Atlantic VC, Motier Ventures, Tholus Capital |
| SI012 | Harmattan AI | Terms of Use — Harmattan AI Website | No e-commerce functionality is provided: you cannot place an order, make a payment, or conclude any sales contract on the Site. |
| SI013 | Harmattan AI | Open Roles — Harmattan AI Careers | |
| SI014 | Harmattan AI | Careers at Harmattan AI | |
| SI015 | Harmattan AI | Harmattan AI and Skyeton Announce Strategic Partnership | |
| SI016 | StartUs Insights | Military Drones Report 2026: Multi-Million Unit Demand & Alliance Re-Armament | |
| SI017 | Boston Consulting Group | Countering the Drone Threat: Europe's C-UAS Opportunity (White Paper) | |
| SI018 | Army Recognition | France's Massive 1,000 AI Drone Deal to Transform Military Training and Deterrence | the budget for these 1,000 drones is described as a multi-million-euro commitment, with deliveries scheduled between October and December this year |
| SI019 | The Defense Post | Dassault and Harmattan AI Sign Strategic Partnership | |
| SI020 | Skyeton | Skyeton and Harmattan AI Announce Strategic Partnership | |
| SI021 | Avions Légendaires | Dassault Aviation fait embarquer Harmattan AI à bord du Rafale F5 | |
| SI022 | Harmattan AI | About Harmattan AI — Company Overview | Harmattan AI integrates scalable manufacturing, rapid iteration, and operational deployment into the core of its defense systems. Our internal integration enables accelerated production timelines and high volume deployment. |
| SI023 | Harmattan AI | Force Readiness — Harmattan AI Capabilities | |
| SI024 | Militarnyi | Ukrainian drones to be equipped with French SAR radars capable of operating through clouds and fog | |
| SI025 | Drone Actu | Harmattan AI va améliorer le drone ukrainien Raybird aux côtés de Skyeton | |
| SI026 | The Defense Post | France to Equip Ukrainian Drones with AI Eyes | |
| SI027 | Harmattan AI | Gobi — Counter-UAS System | |
| SI028 | Defence Industry EU | Skyeton and Harmattan AI partner to advance Raybird UAS capabilities in French and global markets | |
| SE001 | Harmattan AI | Autonomy — Two-Layer Autonomy Model | |
| SE002 | Harmattan AI | Gobi — Autonomous Interception of Group 1-2 UAS | |
| SE003 | Harmattan AI | Gobi Tempest — Autonomous Interception of Group 2-3 UAS | |
| SE004 | Harmattan AI | Barkhan — Autonomous Precision Engagement | |
| SE005 | Harmattan AI | Capabilities — Force Readiness | |
| SE006 | Harmattan AI | Capabilities — Precision Strike | |
| SE007 | Harmattan AI | Sonora Praxis — Training Ready ISR | |
| SE008 | Harmattan AI | Sahara — Synthetic Aperture Radar Imagery on Demand | |
| SE009 | Harmattan AI | Kalahari — C2 with Embedded Autonomy | |
| SE010 | Harmattan AI | Capabilities — Mission Capabilities Overview | |
| SE011 | Harmattan AI | Harmattan AI — Homepage | |
| SE012 | Harmattan AI | Capabilities — VSHORAD | |
| SE013 | Harmattan AI | Capabilities — Persistent ISR | |
| SE014 | Harmattan AI | Blog — Harmattan AI and Skyeton Announce Strategic Partnership | |
| SE015 | Harmattan AI | Blog — Harmattan AI is Exhibiting at Eurosatory 2026 | |
| SE016 | Harmattan AI | Blog — Join Harmattan AI at SOF Week 2026 | |
| SE017 | Harmattan AI | Blog — Harmattan AI $200M Series B led by Dassault Aviation | |
| SE018 | Harmattan AI | Careers — Building Autonomous Defense Systems | |
| SE019 | Harmattan AI | Open Roles — Engineering and Technical Positions | |
| SE020 | Militarnyi | Ukrainian UAVs Will Be Equipped with French SAR Radars from Harmattan AI | |
| SE021 | Drone Actu | Technologie française — Harmattan AI va améliorer le drone ukrainien Raybird | |
| SE022 | Skyeton | Skyeton and Harmattan AI Strategic Partnership Announcement | |
| SE023 | Defence-Industry.eu | Skyeton and Harmattan AI Partner to Advance Raybird UAS Capabilities | |
| SE024 | Le Grand Continent | Entretien — Harmattan AI, Mouad M'Ghari (February 2026) | |
| SE025 | The Defense Post | France Equips Ukrainian Drones with AI Eyes | |
| SE026 | Shield AI | V-BAT — Maritime and Tactical ISR UAS | |
| SE027 | Quantum Systems | Vector AI — Smart Mid-Range eVTOL sUAS | |
| SE028 | Tekever | AR5 — Long-Endurance ISR UAS | |
| SE029 | Delair | Security and Defense Solutions | |
| SE030 | Reuters | Drone crashes, severed fingers — inside a $13 billion Silicon Valley military startup | |
| SE031 | Harmattan AI | Blog — Harmattan AI is Exhibiting at World Defense Show 2026 | |
| SE032 | Harmattan AI | Blog — Join Harmattan AI at UMEX 2026 | |
| SE033 | NATO Support and Procurement Agency | C-UAS Capabilities 2025 — NSPA Reference Document | |
| SU001 | Harmattan AI | Harmattan AI will exhibit at Eurosatory 2026 — Autonomous systems built for modern warfare | At Eurosatory 2026, Harmattan AI will present a portfolio of autonomous military technologies across airborne intelligence, tactical sensing, counter-UAS defense, and autonomous mission management. |
| SU002 | Harmattan AI | Join Harmattan AI at SOF Week 2026 — Advancing operational edge for U.S. and allied forces | Harmattan AI will exhibit at SOF Week 2026 in Tampa, Florida, engaging with special operations forces, U.S. defense stakeholders, and allied partners. |
| SU003 | Harmattan AI | Harmattan AI will exhibit at World Defense Show 2026 — Industrial maturity. Mission readiness. Regional sovereignty. | |
| SU004 | Harmattan AI | Join Harmattan AI at UMEX 2026 — Accelerating autonomy. Advancing defense innovation. | |
| SU005 | Army Recognition | France's massive AI drone deal to transform military training and deterrence | Paris-based startup Harmattan AI has been awarded a multi-million-euro contract to deliver 1,000 AI-enabled micro-drones for the French forces. These drones are expected to enhance operational training ahead of the Orion 2026 multinational exercise. |
| SU006 | Skyeton | Skyeton and Harmattan AI Announce Strategic Partnership | This collaboration demonstrates our commitment to working hand-in-hand with trusted allies to address today's security challenges. By joining forces with Harmattan AI, we are reinforcing Europe's defense capabilities and accelerating the Raybird's adoption in key NATO markets such as France. |
| SU007 | Defence Industry (defence-industry.eu) | Skyeton and Harmattan AI partner to advance Raybird UAS capabilities in French and global markets | |
| SU008 | Militarnyi | Ukrainian drones to be equipped with French SAR radars capable of operating through clouds and fog | Ukrainian drones will be equipped with reconnaissance SAR radars from the French company Harmattan AI, which will enable reconnaissance even in the presence of clouds or fog. |
| SU009 | The Defense Post | Dassault Aviation Leads $200M Funding Round in Harmattan AI | |
| SU010 | Avions Légendaires | Dassault Aviation fait embarquer Harmattan AI à bord du Rafale F5 | On lui doit notamment le drone quadricoptère d'entraînement Sonora qui équipera bientôt l'Armée de Terre, le drone d'interception Gobi acquis par la British Army. |
| SU011 | Drone Actu | Technologie française: Harmattan AI va améliorer le drone ukrainien Raybird aux côtés de Skyeton | |
| SU012 | Crunchbase | Harmattan AI — Company Profile | |
| SU013 | Harmattan AI | Harmattan AI's $200 million Series B led by Dassault Aviation | Harmattan AI has been awarded multiple Programs of Record by the French and UK Ministries of Defence for its autonomous systems portfolio and is now scaling globally in response to growing demand. |
| SU014 | Harmattan AI | Harmattan AI and Skyeton announce strategic partnership | |
| SU015 | Dassault Aviation | Harmattan AI's $200 million Series B led by Dassault Aviation (Press Kit) | Harmattan AI has been awarded multiple Programs of Record by the French and UK Ministries of Defence for its autonomous systems portfolio and is now scaling globally in response to growing demand. The proceeds from this investment round will be used to expand the deployment of AI-enabled missions across new operational theaters. |
| SU016 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion value | The startup in September won an order from the United Kingdom Ministry of Defence to provide as many as 3,000 autonomous drones, following a June order from France's Armed Forces Ministry for delivery of 1,000 combat drones by the end of 2025. |
| SU017 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn | Harmattan AI, which builds autonomy and mission-system software for defense aircraft, had already received strong validation signals from the French and British ministries of defense in its less than two years of existence. |
| SU018 | OpexNews | Harmattan AI — commande UK, 3 000 systèmes autonomes | Ce succès outre-Manche fait écho à la dynamique française. Fin 2024, l'armée de Terre et la DGA ont formulé un besoin simple. Harmattan AI l'a emporté avec un modèle conçu et assemblé en France, baptisé SORONA, commandé à 1000 exemplaires pour des livraisons d'ici fin 2025. |
| SU019 | Le Grand Continent | Entretien — Harmattan AI (Mouad M'Ghari) | La supériorité technologique ne fait pas forcément la supériorité opérationnelle. |
| SU020 | TechFundingNews | France's answer to Helsing: Harmattan AI secures $200M from Dassault Aviation | |
| SU021 | DefenceJobs | Harmattan AI — Company Profile | Customers and Partners: French Ministry of Defence, UK Ministry of Defence, Dassault Aviation, Skyeton. |
| SU022 | The Defense Post | France Gives Ukrainian Drones AI-Powered Eyes That Penetrate Fog | |
| SU023 | AirForce Technology | Harmattan AI secures $200M funding to scale autonomous defence systems | The company supplies its systems to several NATO members and allied countries. In September last year, the UK Ministry of Defence awarded Harmattan AI a contract for autonomous systems under a multi-million pound programme that includes 3,000 units scheduled for near-term deployment. |
| SU024 | DroneXL | Dassault's Harmattan — France's AI Defense | Harmattan has actual military contracts, actual battlefield-proven technology, and now actual backing from the company building France's next-generation fighter jets and their robot wingmen. |
| SU025 | Tech.eu | French startup Harmattan AI hits unicorn status as it strikes deal with French aerospace group | |
| SU026 | Reuters | French drone makers eye Ukraine deals amid funding uncertainty | French drone makers face the challenge of scaling production while managing input cost volatility and procurement budget timing risks from their government customers. |
| SU027 | Reuters | Drone crashes, severed fingers: inside the $13 billion Silicon Valley military startup | Defense tech startups face operational and customer-relationship risks when scaling autonomous systems under compressed timelines; program-of-record contracts provide revenue certainty but not insulation from deployment failures. |
| SU028 | Harmattan AI | Harmattan AI — Company | |
| SU029 | Harmattan AI | Harmattan AI — Contact Us | |
| SU030 | NATO | Autonomous drones take flight at NATO-backed competition (SAPIENCE, July 2025) | NATO-backed competition for autonomous drones demonstrates growing NATO institutional interest in autonomous systems procurement and allied experimentation as a pathway to capability adoption. |
| SU031 | NATO | New NATO Innovation Range starts counter-drone technology testing in Latvia (March 2026) | |
| SU032 | NATO Communications and Information Agency (NCIA) | Allies and industry test the latest counterdrone technology during NATO exercise | |
| SU033 | NATO Support and Procurement Agency (NSPA) | C-UAS Capabilities 2025 | NATO procurement framework for counter-UAS capabilities provides the institutional buying context for allied governments acquiring systems like Harmattan AI's Gobi and Kalahari products. |
| SU034 | NATO | NATO — Autonomous weapons and AI in defense: topic hub | |
| SU035 | International Committee of the Red Cross (ICRC) | ICRC position on autonomous weapon systems | The ICRC calls for new international legal rules on autonomous weapon systems, creating regulatory and reputational risk for government customers procuring such systems without IHL compliance guarantees. |
| SU036 | Stop Killer Robots Campaign | About — Stop Killer Robots Campaign | Growing civil society pressure and international diplomatic activity against fully autonomous weapons systems creates procurement and reputational risk for government customers of AI-enabled autonomous defense systems. |
| SU037 | Stockholm International Peace Research Institute (SIPRI) | Autonomous weapon systems — SIPRI research hub | SIPRI research on autonomous weapon systems maps the growing sovereign government procurement interest in AI-enabled defense systems and the associated governance and humanitarian law concerns. |
| SU038 | EU Publications Office (EUR-Lex) | EU Regulation 2021/821 — dual-use goods export controls | EU dual-use export regulation governs cross-border sales of defense-related items including autonomous drone systems; Harmattan AI customer expansion into non-EU markets requires regulatory compliance review under Regulation 2021/821. |
| SU039 | Helsing AI | HX-2 — Helsing AI product page | Helsing AI's HX-2 product demonstrates that the European sovereign government market actively procures AI-enabled autonomous defense systems, providing comparative context for Harmattan AI's customer acquisition trajectory. |
| SR001 | Harmattan AI | Terms of Use & Conditions — Harmattan AI | Use the Site or any Content for competitive analysis or to build a competing service. |
| SR002 | Harmattan AI | Privacy Policy — Harmattan AI | Your data may be processed in countries outside the European Economic Area (EEA). In such cases, we rely on Standard Contractual Clauses. |
| SR003 | DroneXL | Dassault's Harmattan: France's AI Defense Play | Companies like Harmattan are planning facilities capable of producing 10,000 drones monthly. That scale-up requires supply chains, skilled workers, and testing infrastructure. |
| SR004 | OpexNews | Harmattan AI — commande UK 3000 systèmes autonomes | Reste l'épreuve des prochains mois : tenir la cadence, sécuriser la chaîne d'approvisionnement (capteurs, calculateurs, liaisons) et prouver la fiabilité en conditions réelles. |
| SR005 | Le Grand Continent | Entretien — Harmattan AI, Mouad M'Ghari | Il faut être assez indépendant de certaines licences d'export, de certains composants, de certaines chaînes d'approvisionnement, notamment lorsqu'on cherche à produire en grande quantité. |
| SR006 | Army Recognition | France's massive 1,000 AI drone deal to transform military training and deterrence | |
| SR007 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion value | The deal comes as France, Germany and Spain are struggling to move forward with plans for a joint Future Combat Air System. |
| SR008 | NATO Communications and Information Agency (NCIA) | Allies and industry test the latest counter-drone technology during NATO exercise (TIE 26) | Interoperability in modern multi-layered C-UAS operations is no longer optional — it is essential! |
| SR009 | NATO | New NATO Innovation Range starts counter-drone technology testing in Latvia | |
| SR010 | NATO | Autonomous drones take flight at NATO-backed competition (SAPIENCE) | |
| SR011 | NATO Allied Command Transformation | LCI-X builds approach to fast-moving counter-UAS threat | LCI-X provides a structured environment to understand how those national and commercial systems can be integrated into a more coherent NATO approach. |
| SR012 | Dassault Aviation | Harmattan AI's $200 million Series B led by Dassault Aviation (press kit) | |
| SR013 | Boston Consulting Group | Countering the Drone Threat: Europe's C-UAS Opportunity (May 2026 Whitepaper) | C-UAS systems remain expensive, difficult to scale, and dependent on supply chains not yet built for sustained attrition. |
| SR014 | Harmattan AI | Terms of Use — Governing Law and Dispute Resolution (French law, CMAP) | |
| SR015 | TechCrunch | Anduril raises $5B, doubles valuation to $61B | When we founded Anduril in 2017, defense was not a category that attracted significant venture investment. That has changed meaningfully. |
| SR016 | Shield AI | Shield AI to acquire Aechelon and raise $2B at $12.7B valuation | |
| SR017 | TechCrunch | Daniel Ek-backed defense tech Helsing to raise $1.2B at $18B valuation | |
| SR018 | DefenceJobs.org | Harmattan AI — Company Profile (DefenceJobs) | Gobi is Harmattan's counter-UAS system... without requiring a human in the loop for each intercept. |
| SR019 | Defence Industry EU | Skyeton and Harmattan AI partner to advance Raybird UAS in French and global markets | |
| SR020 | Militarnyi | Ukrainian drones to be equipped with French SAR radars — Harmattan AI Sahara | |
| SR021 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn | The company was formerly described as a European Anduril before the Dassault partnership prompted a rebranding. |
| SR022 | TechFundingNews | France's answer to Helsing: Harmattan AI secures $200M from Dassault Aviation | |
| SR023 | The Defense Post | Dassault and Harmattan AI announce partnership and Series B funding | |
| SR024 | Tech.eu | French startup Harmattan AI hits unicorn status in deal with Dassault | |
| SR025 | Quantum Systems GmbH | Quantum Systems raises €160M targeting global leadership in aerial intelligence | |
| SR026 | Tekever | Tekever raises 70 million euros and brings strategic investors on board | TEKEVER, which is already profitable, will accelerate investment in R&D to support product innovation. |
| SR027 | AirForce Technology | Harmattan AI funding: $200M round and Dassault partnership details | |
| SR028 | TechCrunch | Germany's Helsing doubles down on drones for Ukraine, scales up manufacturing | |
| SR029 | The Defense Post | France, Ukrainian drones and AI eyes — Harmattan AI Sahara SAR | |
| SR030 | Reuters | Drone crashes, severed fingers: safety risks at a defense drone startup | Drone crashes and safety incidents illustrate that rapid scaling in defense drones carries real operational safety risk sector-wide. |
| SR031 | MarketsandMarkets | Europe UAV (Drone) Market — Forecast to 2030 | |
| SR032 | Expert Market Research | Europe Unmanned Aerial Vehicle (UAV) Market — export control and supply chain | European UAV exports require coordination between national export licensing and EU common position frameworks, creating regulatory complexity for European manufacturers targeting global markets. |
| SR033 | European Commission Joint Research Centre / Eurocontrol | Drone, Counter-drone and Autonomous Systems — Eurocontrol C-UAS Workshop | UAS legal base: at the early stage — drone rules, drone airspace rules, U-Space — governed under EASA basic regulation. |
| SR034 | European Parliament Research Service (EPRS) | Military drone systems in the EU and global context — EPRS At a Glance 2026 | |
| SR035 | European Parliament Think Tank | Military drone systems in the EU and global context: types, capabilities and regulatory frameworks | |
| SR036 | Helsing | Helsing raises €600M to invest in European technological sovereignty | |
| SR037 | International Committee of the Red Cross (ICRC) | ICRC position on autonomous weapon systems | Autonomous weapon systems select and apply force to targets without human intervention... This loss of human control and judgement in the use of force raises serious concerns from humanitarian, legal and ethical perspectives. |
| SR038 | Official Journal of the European Union | Regulation (EU) 2021/821 — EU dual-use items export control regime | |
| SR039 | Stop Killer Robots | About Stop Killer Robots — Campaign Against Lethal Autonomous Weapons | |
| SR040 | SIPRI | Autonomous weapon systems research — Stockholm International Peace Research Institute | |
| SR041 | NATO | Artificial Intelligence topics — NATO policy and principles | |
| SR042 | Helsing | Helsing HX-2 — AI-enabled drone for European defense | |
| SR043 | Anduril Industries | Anduril raises Series H funding — $5B at $61B valuation | Defense was not a category that attracted significant venture investment. That has changed meaningfully over the last several years. |
| SR044 | Tekever | Tekever solutions — UAS Intelligence as a Service | |
| SR045 | Tekever | AR3 – EVO – Tekever | |
| SR046 | Tekever | TEKEVER and MERIO Partner to Advance ISR Innovation and Strengthen Industrial Growth in France | |
| SR047 | Delair | DT26 Open Payload Drone : onboard your own preferred sensors | |
| SR048 | MyDefence | MyDefence Opens U.S. Counter-Drone Production Facility in Oklahoma City | |
| SR049 | MyDefence | MyDefence Technology Selected for Australia’s LAND 156 Counter-Drone Program | |
| SR050 | MyDefence | Wearable Drone Detector for Reliable UAS Detection - MyDefence | |
| SV001 | TechCrunch | Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn | "Harmattan AI raises $200M Series B led by Dassault Aviation, becomes defense unicorn." |
| SV002 | Defense News | Dassault Aviation invests in Harmattan AI at $1.4 billion value | "Harmattan currently has more than 130 employees, with a median experience of 15 years." |
| SV003 | Tech Funding News | France's answer to Helsing — Harmattan AI secures $200M from Dassault Aviation | |
| SV004 | AirForce Technology | Harmattan AI Funding: $200M | |
| SV005 | DroneXL | Dassault's Harmattan — France's AI Defense | "The $1.4 billion valuation might seem aggressive for a company barely 18 months old, but it reflects reality: Harmattan has actual military contracts, actual battlefield-proven technology, and now actual backing from the company building France's next-generation fighter jets." |
| SV006 | opexnews.fr | Harmattan AI — commande UK, 3 000 systèmes autonomes | |
| SV007 | Le Grand Continent | Entretien avec Mouad M'Ghari, CEO de Harmattan AI | "La livraison à des armées partenaires s'effectuent en quatorze mois pour l'armée française, quinze mois pour l'armée britannique." |
| SV008 | Boston Consulting Group | Countering the Drone Threat: Europe's C-UAS Opportunity | "Demand is projected to exceed $5–7 billion by 2029, driven by the need to protect thousands of military and civil assets." |
| SV009 | TechCrunch | Daniel Ek-backed defense tech Helsing to raise $1.2B at $18B valuation | "Five-year-old European military drone startup Helsing is reportedly close to raising a new $1.2 billion round at about an $18 billion valuation." |
| SV010 | TechCrunch | Anduril raises $5B, doubles valuation to $61B | "Anduril has raised a $5 billion Series H round at a $61 billion valuation... the nine-year-old defense tech company doubled revenue in 2025 to $2.2 billion." |
| SV011 | Shield AI | Shield AI to Acquire Aechelon and Raise $2B at $12.7B Valuation | |
| SV012 | Quantum Systems GmbH | Quantum Systems raises €160M as it targets global leadership in aerial intelligence solutions | "The new funding follows several years of exceeding 100% year-over-year revenue growth for the company, which now has 550 people." |
| SV013 | TEKEVER | TEKEVER Raises €70 Million and Brings Strategic Investors On Board | TEKEVER, which is already profitable, will accelerate investment in R&D. |
| SV014 | Harmattan AI | Harmattan AI $200 Million Series B Led by Dassault Aviation | |
| SV015 | Dassault Aviation | Harmattan AI's $200 Million Series B Led by Dassault Aviation (Press Kit) | "This partnership with Harmattan AI reflects our commitment to integrating high-value autonomy into the next generation of combat air systems." — Eric Trappier, CEO Dassault Aviation |
| SV016 | Tech.eu | French startup Harmattan AI hits unicorn status as it strikes deal with French aerospace group | |
| SV017 | SIPRI | SIPRI Arms Transfers Database | |
| SV018 | The Defense Post | Dassault and Harmattan AI forge strategic defense AI partnership | |
| SV019 | Reuters | German drone maker Quantum Systems triples valuation after new funding round | |
| SV020 | Crunchbase | Harmattan AI — Funding, Investors, and Team | |
| SV021 | Avions Légendaires | Dassault Aviation embarque Harmattan AI à bord du Rafale F5 | |
| SV022 | MarketsandMarkets | Europe UAV Drone Market Size, Share and Trends Analysis | |
| SV023 | MarketsandMarkets | Counter-UAS Systems Market Size, Share and Trends Analysis | |
| SV024 | Harmattan AI | Harmattan AI — Company Overview | |
| SV025 | Helsing | Helsing raises €600M to invest in European technological sovereignty | |
| SV026 | Reuters | Drone crashes, severed fingers: the $1.3 billion Silicon Valley military startup | "Drone crashes, severed fingers: the $1.3 billion Silicon Valley military startup" — Reuters June 2026, cautionary precedent for rapid valuation deflation following operational reliability failures |
| SV027 | Skyeton | Skyeton and Harmattan AI — Strategic Partnership | |
| SV028 | Harmattan AI | Harmattan AI is exhibiting at Eurosatory 2026 | |
| SV029 | Harmattan AI | Join Harmattan AI at SOF Week 2026 | |
| SV030 | NATO | New NATO innovation range starts counter-drone technology testing in Latvia | |
| SV031 | NATO Communications and Information Agency | Allies and industry test the latest counter-drone technology during NATO exercise | |
| SV032 | Harmattan AI | Join Harmattan AI at World Defense Show 2026 | |
| SV033 | defence-industry.eu | Skyeton and Harmattan AI partner to advance Raybird UAS capabilities in French and global markets | |
| SV034 | European Parliament Think Tank | Military drone systems in the EU and global context: types, capabilities and regulatory frameworks | |
| SV035 | Reuters | French drone makers eye Ukraine deals amid financing uncertainty | Reuters November 2025 article on French drone makers and financing uncertainty provides adverse context on revenue opacity risk in the French defense startup sector |
| SV036 | Helsing | HX-2 — Helsing Autonomous Strike Drone for Ukraine | Helsing HX-2 autonomous strike drone operational in Ukraine; primary product of the $18B-valued European defense AI peer. |
| SV037 | Helsing | Helsing — About | |
| SV038 | Shield AI | Shield AI — About Us | What does the military of 2030 look like, and what role does autonomy play? The answer: AI pilots powering every military asset. |
| SV039 | Anduril Industries | Anduril News | |
| SV040 | TEKEVER | TEKEVER — About | |
| SV041 | Delair | Delair Defense Solutions | |
| SV042 | NATO Support and Procurement Agency | C-UAS Capabilities 2025 | NSPA C-UAS capabilities report 2025 documents the NATO procurement landscape for counter-drone systems used as procurement pipeline reference. |
| SV043 | Shield AI | Shield AI News | |
| SV044 | Delair | Delair — Solutions for Defense and Security |