Harmattan AI
The French military drone startup that became a unicorn in under 18 months: three autonomous systems (Sonora, Sahara, Gobi), contracts with DGA, UK MoD and NATO, and a $200M Series B led by Dassault Aviation.
Europe's most interesting startup stories. Every week.
No spam. Unsubscribe anytime.
Podcast Episode
๐๏ธ Deep Dive
EP35 - Harmattan AI: controlled autonomy for European defense
Listen to the episodeAbout Harmattan AI
Harmattan AI is the most remarkable speed story in European defense-tech in recent years. Founded in Paris in April 2024 by Mouad M'Ghari and five co-founders, all trained at France's grandes รฉcoles and MIT, the company went from startup to $1.4 billion unicorn in under eighteen months. This is not a coincidence: it is the perfect convergence of a favorable geopolitical context, a radical technological approach, and a go-to-market strategy that ignored the unwritten rules of traditional defense procurement. The context was 2024โ2025: Europe rearming rapidly in the wake of Russia's invasion of Ukraine, governments seeking reliable and immediately deployable drones, and a VC ecosystem that, for the first time, was treating European defense-tech as a serious asset class. Harmattan read this window better than anyone else. The core product is a three-system platform. SONORA is an "attritable" micro-ISR drone โ small, cost-effective, expendable โ designed for training and real operations with on-edge AI that works in GNSS-denied environments, without GPS and without cloud connectivity. It weighs 1.8 kg, operates for 40 minutes, and is ready in under a minute. SAHARA is a SAR radar payload that reconstructs images and detects changes directly on-board, eliminating cloud latency in high-intensity operational scenarios. GOBI is the C-UAS interceptor: it exceeds 230 km/h and neutralises an enemy drone in under a minute from launch, using IR and radar computer vision to identify the optimal impact point. The AI approach is deliberately "frugal": distilled models, minimal latency, fully on-device inference. There is no cloud infrastructure dependency in real operational environments โ a choice that is not technical but strategic, because high-intensity conflict theatres lack reliable connectivity, and because data sovereignty is a non-negotiable requirement for government customers. The go-to-market bypassed the typical slowness of defense procurement. In July 2025, the French DGA ordered 1,000 Sonora drones โ delivered by January 2026. In September 2025, the UK Ministry of Defence signed a Program of Record for 3,000 systems. NATO awarded a PoR contract that the company describes as "the fastest ever given to a company this young." No lengthy POCs, no years of preliminary certifications: real operational urgency, a product that worked, and relationships built with the right decision-makers. The $200 million round in January 2026, led by Dassault Aviation, is not just financing: it is an industrial partnership. Dassault brings Harmattan into the heart of the French military aviation system โ AI integration in the Rafale F5 and future UCAS drones of the SCAF programme. For Harmattan, this means access to international government relationships and a defense ecosystem that is otherwise impenetrable. For Dassault, it means bringing embedded AI and agile mission software to platforms built with traditional industrial logic. In the European competitive landscape, Harmattan competes with Helsing (defense AI, software-focused), Quantum Systems (German ISR drones) and, indirectly, with America's Anduril as the reference model. The declared moat is full vertical integration โ hardware, payload, AI, mission software all designed in-house โ eliminating dependencies on critical third-party suppliers and accelerating iteration cycles. The real test will be industrial scalability: the stated goal is 10,000 drones per month, with a pilot site in Paris and a second facility at Orly. Harmattan AI teaches a precise playbook: choose a market with real urgency, build the minimum product that solves the concrete operational problem (not the perfect product), go directly to the Program of Record skipping endless POCs, and find the industrial partner that transforms the funding round into a distribution multiplier. In eighteen months, with $242 million raised and thousands of drones delivered, the playbook worked.
The Story
Harmattan AI was founded in April 2024 at a time when Europe was rediscovering the urgency of strategic autonomy in defense. The founding team, with backgrounds in AI and embedded systems, chose not to build generic hardware but to focus on the software layer of autonomy โ the differentiator that enables existing drones to operate in coordinated swarms.
How Harmattan AI works
Business Model
B2G (Business-to-Government): Program of Record contracts with NATO governments for the supply of autonomous drone systems (Sonora, Sahara, Gobi) and associated training and operational support services. Additional revenue from the R&D partnership with Dassault Aviation for AI integration in next-generation defense systems (Rafale F5, UCAS). Model based on scalable industrial production in France targeting 10,000 units/month.
Market & Clients
Key Customers
NATO governments and armed forces: French DGA (1,000 Sonora drones delivered), UK Ministry of Defence (3,000 systems, PoR September 2025), NATO (multi-million dollar PoR). Industrial partner: Dassault Aviation for AI integration in Rafale F5 and UCAS.
Funding & Investors
Funding Rounds
- 2024 - Early stage: $42M
Lead: FirstMark, Atlantic VC - 2026 - Series B: $200M
Lead: Dassault Aviation - 2026 - Funding round: $200M
Key Investors
Founding Team
Founders
Key Metrics
Lessons from Harmattan AI
- Defense tech has become attractive for European VC: 0 to unicorn speed in 2 years is extraordinary.
- Focusing on autonomy software instead of hardware lowers production barriers and increases margins.
- Government contracts as a 'Program of Record' are the ultimate moat: they create institutional dependency.
Similar Startups
AMI Labs
Next-generation AI research lab founded by Yann LeCun focused on artificial general intelligence development and innovation.
Read AnalysisGenesis AI
French startup that unveiled GENE-26.5, a foundation model for industrial robotics compatible with multiple robot types, and a humanoid robotic hand for delicate manipulation. Raised $105M in seed funding with industrial clients in France, Germany and Italy.
Read AnalysisAlta Ares
French defence tech startup developing air defence systems and AI software to intercept drones, missiles and hostile aircraft.
Read Analysis