Helsing: A €12 Billion Unicorn in a Rearming Europe
5 min read
For years the question was: why doesn’t Europe have its own Palantir, its own Anduril? Now an answer exists, and it is called Helsing. Founded in Munich in 2021, the defence software and AI systems company closed a round a year ago — June 2025 — raising €600 million led by Prima Materia, the investment vehicle of Daniel Ek, Spotify’s founder. Valuation: around €12 billion, more than doubled in twelve months. It is Europe’s most valuable unlisted technology company, alongside a very small handful of others, and it has grown by selling a precise word: sovereignty. But between the valuation and the battlefield lies testing — and it is precisely there, in recent months, that the Helsing case has become interesting for anyone buying critical technology.
The facts, in order
- February 2025 — the alliance with Mistral AI: at the Paris AI summit, Helsing and France’s generative-model champion announce a strategic partnership to develop “vision-language-action” models for European defence systems.
- February 2025 — 6,000 attack drones for Ukraine: Helsing announces production of 6,000 HX-2 loitering munitions — X-wing configuration, electric propulsion, onboard AI claimed to be resistant to electronic warfare — following an initial batch of 4,000 HF-1 units. Production takes place at a “Resilience Factory” in southern Germany, with a stated capacity of over 1,000 units a month.
- May-June 2025 — expansion towards sea and sky: the company unveils Lura, an acoustic model for underwater surveillance, and the SG-1 Fathom underwater glider; it then acquires German aircraft manufacturer Grob and flies its own air-combat agent, Centaur, aboard a Saab Gripen E.
- June 2025 — the €600 million: the round, led by Prima Materia with Lightspeed, Accel, General Catalyst, Plural, Saab and others, brings the valuation to around €12 billion. The company states that the funds will be used to invest in “European technological sovereignty”.
- September 2025 — the CA-1 Europa: Helsing unveils an autonomous, uncrewed combat aircraft, with first flight planned for 2027.
- January 2026 — the cold shower: Bloomberg reports that Ukraine has suspended new orders for HX-2 after disappointing front-line tests — launch problems, vulnerability to electronic jamming, doubts over the cost-effectiveness of the HF-1 units. Helsing disputes the findings, citing documented hit rates described as “encouraging” and requests for more than a thousand additional drones from the units using them.
Neither version cancels out the other. But for outside observers, the contrast between the financial trajectory and the operational friction is the heart of the lesson.
Lesson 1: Europe is learning to build champions — and to depend on them
The Helsing case shows that European capital, talent and public demand can produce a world-scale player in four years. That is good news for anyone invoking strategic autonomy: technological sovereignty stops being a slogan once credible continental suppliers exist. But a national champion remains a supplier: dependence on a company in Munich is not, in itself, any healthier than dependence on one in Denver. What changes is the jurisdiction, not the logic of lock-in. For a ministry as for a company, the contractual question stays identical: where the data resides, who controls the models, how much it costs to switch. The flag on the contract does not replace the clauses.
Lesson 2: the boundary between deterrence and lethal automation is a design requirement
Swarms of drones managed by a single operator, AI agents fighting in the air, gliders patrolling the seabed for months: every Helsing announcement shifts more decisions from person to machine. The company has always stated that engagement decisions remain human; Europe’s debate on autonomous weapons has for years demanded that this principle — “meaningful human control” — be verifiable, not merely declared. For anyone adopting AI in critical settings, civilian ones included, the translation is direct: human oversight is not an ethics brochure, it is an architectural requirement. Who decides, with what information, in how much time, with what ability to stop the system: if the supplier cannot answer in writing, the requirement does not exist.
Lesson 3: what you buy is the testing, not the press release
Ukraine’s suspension of HX-2 orders — whatever the outcome of the dispute — is the most instructive part of the story. The world’s most demanding customer, the one that actually fights, has made new purchases conditional on results verified in the field. That is exactly the method every technology buyer should demand: acceptance tests defined before signature, metrics measured in the real environment, payments tied to performance. It holds for a loitering munition and it holds for an AI system in production: the gap between demo and operation is where budgets get lost. A serious supplier — and we say this as a supplier — accepts that verification, because it is the only way to build trust that outlasts a single funding cycle.
What to do if you are buying critical technology
- Separate valuation from value: the billions raised by a supplier measure investor expectations, not performance in your context.
- Write acceptance tests into the contract: real environment, agreed metrics, thresholds below which you renegotiate or walk away.
- Demand human control as a technical specification: who authorises the system’s actions, with what data, with what shutdown procedure.
- Treat sovereignty as architecture: exportable data, documented models, reversibility — the supplier’s nationality is not enough.
Are you evaluating an AI system for critical operations and want to define acceptance tests and human-control requirements before signature? Let’s talk.