Helsing at €12 Billion: What Four Years of Defence AI Actually Built

Analysis as of 22 July 2025.

No European defence company has compressed as much history into four years as Helsing. Founded in Munich in 2021, the firm reached mid-2025 having raised €1.37 billion from investors and carrying a €12 billion valuation — figures without precedent for a private European defence start-up. The cap table tells its own story: Spotify founder Daniel Ek's Prima Materia holds roughly 17 percent and the chairmanship, General Catalyst around 12 percent, Saab about 5 percent, with Accel and Lightspeed alongside. Yet despite the transatlantic money, the company insists on European control, holds no US government contracts, and pledges to sell exclusively to democratic governments with a human kept in or on the loop.

What that capital bought is a portfolio, not a product. Altra coordinates reconnaissance and strike, turning drone and spotter feeds into targeting data for artillery and loitering munitions. Cirra reads the electromagnetic spectrum in real time and was selected — with Saab, back in 2023 — for Germany's Eurofighter electronic-warfare upgrade. Centaur, a machine-learning-trained AI pilot, flew a live Gripen E in beyond-visual-range engagements in May–June 2025 — the first known instance on an operational fighter. Below the waterline sits the SG-1 Fathom, a two-metre glider unveiled in May 2025 that can stay submerged for up to 90 days carrying Lura, an acoustic model that classifies vessels at sound levels ten times quieter than conventional sonar AI. Above it all, a multi-sensor satellite constellation with Loft Orbital was announced in mid-2025.

The harder pivot was industrial. A company that began by supplying AI software for other builders' platforms opened its first "Resilience Factory" in southern Germany in late 2024, and by mid-2025 output of the HX-2 — an electric X-wing loitering munition with 100 km range, roughly 5 kg of payload and a 250 km/h top speed, navigating by onboard AI without GPS — had passed 1,000 units per month. The June 2025 acquisition of Grob Aircraft added an aerospace manufacturing arm and some 275 staff to a workforce of around 600. Ukraine ordered 4,000 of the earlier HF-1 drones in 2024, with part of that manufacturing handled by Ukrainian industry; Britain's Project ASGARD paired Altra with HX-2 to extend army targeting; EDF consortia EICACS and LATACC, plus an AI role in the Future Combat Air System with Airbus, anchor the company inside Europe's flagship programmes.

Maturity, though, is uneven by design. The strike drones sit at TRL 9, combat-proven and in serial production. Altra is assessed at 8–9, Cirra at roughly 7 pending Eurofighter flight trials, Centaur at 7–8 after its Gripen sorties, the SG-1 and Lura at 6–7 following a Portsmouth naval demonstration, and the satellite venture at 5–6. The patent file — object classification and localisation, AI performance monitoring, RF-signal processing, filed chiefly at the European level — signals a company building defensible IP rather than renting it. But a readiness ladder that runs from combat-proven to barely-announced is precisely where due-diligence questions live.

The record to 22 July 2025 is documented; what it implies is not settled. These are the questions the profile poses rather than assumes away: — Can a €12 billion valuation be reconciled with a contract base drawn almost entirely from EU and NATO governments, and what happens to it if procurement cycles slow? — Does the shift from software supplier to drone manufacturer and aircraft-plant owner strengthen the model, or dilute the margins that justified the price? — How dependent is the technical edge on the Ukrainian battlefield data loop, and is that advantage repeatable in a different conflict? — Is "European controlled" a durable structure when US venture funds hold stakes as large as General Catalyst's roughly 12 percent, or a positioning claim that ownership arithmetic could eventually test? — And on DFM's ten-point European Strategic Assessment Matrix, where exactly does Helsing score, which single criterion denies it a perfect mark — and what would have to change for that gap to close?