Torsten Reil, co-founder and CEO of Helsing, explains how AI-driven autonomous drones and fighter jets are making traditional tanks, artillery, and crewed aircraft obsolete, and why Europe must build its own mass-producible, software-defined defense industrial base to deter aggression and protect democracies.
The battlefield has already shifted to autonomous drones
Tanks and artillery are no longer survivable; Russian front-line soldier life expectancy is roughly 20–30 minutes because ISR drones spot targets and FPV or larger strike drones destroy them within minutes.
The kill chain is now: ISR drone detects → strike drone (FPV or long-range autonomous) engages; humans in vehicles or on foot cannot survive this environment.
Jamming is defeated either by fiber-optic tether (FPVs) or by onboard AI that navigates visually without GPS and performs terminal guidance autonomously.
Once strike drones become fully autonomous (no radio link to jam), the lethality and scale increase further; Helsing’s HX2 strike drone already flies 100 km with 2–4 kg payload using GPS-free visual navigation.
Reil argues 100,000 such autonomous strike drones at a border could destroy an invading armored column of 3,000 tanks in hours, making large-scale land invasion effectively impossible — the core of his deterrence thesis.
Helsing was founded to give Europe a sovereign, AI-first defense industrial base
Reil (ex-NaturalMotion, sold to Zynga for ~$500M) did not plan another company until meeting co-founders Gundbert Scherf (ex-McKinsey aerospace/defense) and Nicholas Gross (AI for legacy primes) in late 2020/early 2021.
All three independently concluded Europe had no Anduril/Palantir equivalent: no venture-backed, software-first defense company capable of mass-producing autonomous systems.
Founded March 2021; extremely unpopular — banks treated defense like porn/gambling, VCs refused meetings; raised €8M seed from angels/family offices, then Daniel Ek (Prima Materia) invested €100M three months later, pre-Ukraine invasion.
Mission: “protect our democracies.” This clarity replaces generic corporate values; deployed systems in Ukraine make the mission concrete daily.
Explicit commitment: never sell the company; build a generational European champion like Spotify.
Winning programs of record — the Eurofighter electronic-warfare upgrade
Strategy: use the €100M war chest to bid directly on large government “programs of record” (prime contracts in national budgets) instead of subcontracting — high risk, long cycles, but the only path to scale fast.
First target: German Eurofighter electronic-warfare upgrade. Legacy radar-warning receivers failed against software-defined, shape-shifting radars; Helsing’s deep-learning approach won against established consortia in 2 years 4 months — fastest founding-to-program-of-record on record.
Nicholas’s AI intuition + Gundbert’s program knowledge + Reil’s fundraising/ops created the repeatable playbook: find large programs where AI gives a differentiated edge, bet big, deliver.
Centaur: a superhuman AI fighter pilot for beyond-visual-range combat
Beyond-visual-range (BVR) air combat is “chess in the sky” — radar management, missile kinematics, energy bleeding, offensive/defensive timing.
Active pilots called it the last bastion of human intuition; Helsing trained a neural net via reinforcement learning in a simulator running 20,000–25,000× real-time (AlphaGo/AlphaStar approach).
Result: AI consistently beats active Eurofighter, Rafale, and Gripen pilots in simulation and, in 2023, flew a Saab Gripen (Centaur) in live air-combat maneuvers over the Baltic Sea against a human-piloted Gripen — won all three engagements.
Centaur now scales to swarms of autonomous fighters; Reil concludes crewed fighters will become untenable against autonomous swarms — same logic as tanks: immoral to send humans when machines are superior and survival is near zero.
Deterrence, autonomy, and the morality of defense
Ultimate goal: zero lives lost by preventing war through credible deterrence — make the adversary’s cost calculus prohibitively high.
If deterrence fails, minimize own casualties by replacing humans with autonomous systems on land, air, and sea; this is not aggression but moral necessity.
Society must decide whether machines may make life/death decisions when they are more accurate than humans; Helsing runs internal ethics workshops, deployment guides, and proliferation controls, but public/journalist scrutiny has faded since Ukraine.
Reil acknowledges pacifist pushback (especially in Germany) but argues it ignores the reality of adversaries building capabilities; containment via strength is the only stable peace.
Mass manufacturing: simple hardware, complex software, vertical integration where it matters
Thesis: design hardware “good enough” for mass production, absorb complexity into AI software (sensor fusion, recalibration, coordination) — enables saturation attacks and rapid scaling.
HX2 strike drone: scaled from dozens/month to 1,000 baseline, 3,000/month in a small factory; supply chain, QC, and traceability had to be reinvented.
Current bottleneck is not manufacturing speed but government procurement volume; €1B framework contract in Germany signed, more needed across NATO/Allies.
“Resilience factories”: distributed, potentially underground micro-factories on allied soil so partners control supply chain and production survives strikes; sensitivity varies by country (e.g., UK maritime data sovereignty).
Airbase vulnerability drives design: CA1 autonomous fighter takes off from highways/rough strips, not fixed airbases — cannot bomb all roads.
Talent density as the single lever of company performance
Reil’s biggest NaturalMotion regret: under-investing in hiring, performance management, and leadership development; at Helsing, talent density is the explicit #1 priority.
System: external recruiting agency from day one; founders reviewed pipeline KPIs weekly for first year; built internal talent team; quarterly calibration reviews where managers must justify every rating — forces shared standard of “what good looks like” and makes moving underperformers low-opportunity-cost because pipeline is strong.
Hiring signal: “effortlessness” — deep knowledge without posturing, combined with curiosity and high agency; detectable in ~10 minutes for senior roles.
Founders still run end-to-end searches for transformational roles; Daniel Ek participates in key hires.
Culture emerged bottom-up: “Whatever it takes” and “The truth is in the field” (systems must work in combat, not just tests) — born from deployed engineers’ experiences.
Deploying in Ukraine changed the company’s gravity
First combat use was sobering; light-hearted startup culture gave way to seriousness — building for war is fundamentally sad, even when necessary.
Best outcome for Helsing: nothing happens (deterrence works); war is a failure of deterrence, not a business goal.
From biology simulations to gaming to defense — an outsider’s first-principles path
1999 PhD in complex systems: evolved recurrent neural nets (genetic algorithms) to walk simulated creatures — early precursor to reinforcement-learning motor control.
Spun out NaturalMotion → “Endorphin” procedural animation (Lord of the Rings, GTA IV) → pivoted to mobile games (CSR Racing, #1 grossing) → sold to Zynga.
Never felt part of the “in-group” (biology, gaming, now defense) — outsider status enables unconventional moves: 99¢ pricing when industry said $5, free-to-play 3D first, AI-first defense when primes said impossible.
North stars: “#1 grossing game” then “company that protects democracies” — ambitious, measurable, orienting.
Commentary is worthless; building is the only lever that matters
Social platforms incentivize negative commentary; net negative for society, drags down ambition (especially in Europe).
Founders have the highest leverage to change society — politicians, investors, creators follow; commentary creates zero value.
Advice: ignore the noise, focus entirely on building.
