Stark Pushes Surge Contracts to Turn Drone Production Into Deterrence

Sequoia Capital

True military deterrence requires contracting pre-paid factory surge capacity and continuous software updates rather than stockpiling small batches of static, exquisite weapons.

AI-assisted summary of European Defense Sovereignty Starts in the Factory: Stark’s Uwe Horstmann and Johannes Schaback — original source. Check the original for full context.

Source published: 2026-10-28

Key ideas

• European defense startup Stark scaled to 700 employees and secured billions in defense contracts within 18 months, deploying strike drones directly to the frontline in Ukraine. (0:00)
• Stark plans to deploy roughly €600 million into core R&D and manufacturing capacity over 12 to 36 months, matching the non-subsidized, company-funded R&D budgets of legacy European defense primes. (35:56)
• The company proposes a 'deterrence by contract' acquisition framework, urging governments to pre-pay for reserve manufacturing surge capacity and warm production lines like cloud infrastructure rather than relying strictly on static finished-goods stockpiles. (40:00)
• To align software agility with frontline feedback, Stark treats the staff sergeant operating a weapon as more critical than the procurement general, rapidly iterating autonomous systems designed for washing-machine-level manufacturing simplicity. (2:28)

Detailed notes

## Background Published on YouTube by Sequoia Capital on October 28, 2026, and hosted by Pat Grady and Luciana Lixandru, this episode features Stark co-founders Uwe Horstmann (CEO) and Johannes Schaback (CTO). According to the video description, the interview covers how Stark scaled to roughly 700 employees and deployed strike drones to Ukraine within two years, detailing why mass manufacturability, frontline user iteration, and contractually guaranteed surge capacity form the core of European defense sovereignty.

## TL;DR Stark co-founders Uwe Horstmann and Johannes Schaback outline their playbook as a European defense 'neo-prime' designing low-cost, software-driven autonomous strike munitions. They argue that credible deterrence relies on rapid industrial scale and frontline feedback from enlisted operators rather than slow procurement cycles and bespoke exquisite platforms. The discussion recounts Stark's crucible moments, including turning a public Bundeswehr test failure into validated sub-meter accuracy through round-the-clock field engineering, as well as embedding strict NATO safety guardrails into their AI autonomy stack. Looking ahead, they propose restructuring defense procurement like cloud computing contracts with pre-paid surge capacity to maintain continuous software readiness and industrial deterrence across Europe.

## Atmosphere & style The conversation carries a focused, high-urgency tone that blends Silicon Valley venture velocity with the sober operational realities of frontline combat. Horstmann and Schaback speak with grounded, self-effacing candor—recounting soldiers breaking down drone shipping crates for firewood and field engineers scrambling in freezing shipping containers rather than relying on polished defense-industry talking points. The dynamic with hosts Pat Grady and Luciana Lixandru feels like an intimate, strategic debrief, building an emotional arc from early trial setbacks to a steely, generational commitment to European rearmament.

## Bulletin - [00:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=0s) European defense startup Stark scaled to 700 employees and secured billions in defense contracts within 18 months, deploying strike drones directly to the frontline in Ukraine. - [35:56](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2156s) Stark plans to deploy roughly €600 million into core R&D and manufacturing capacity over 12 to 36 months, matching the non-subsidized, company-funded R&D budgets of legacy European defense primes. - [40:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2400s) The company proposes a 'deterrence by contract' acquisition framework, urging governments to pre-pay for reserve manufacturing surge capacity and warm production lines like cloud infrastructure rather than relying strictly on static finished-goods stockpiles. - [02:28](https://www.youtube.com/watch?v=qhIabMKHJlA&t=148s) To align software agility with frontline feedback, Stark treats the staff sergeant operating a weapon as more critical than the procurement general, rapidly iterating autonomous systems designed for washing-machine-level manufacturing simplicity.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=0s) Stark's four-part neo-prime playbook centers on direct frontline battlefield integration, venture capital-backed independent R&D, rigorous NATO qualification, and industrial scaling from day one. - [02:28](https://www.youtube.com/watch?v=qhIabMKHJlA&t=148s) Tactical operators in the trenches matter more than procurement brass; listening directly to enlisted soldiers and shipping iterations within 24 to 48 hours establishes frontline operational trust. - [02:28](https://www.youtube.com/watch?v=qhIabMKHJlA&t=148s) Frontline usability details often trump whiteboard engineering, as shown when Ukrainian units rejected lightweight plastic transit cases because they produce smoke when burned, demanding plain wooden crates that can double as fuel or trench reinforcement. - [10:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=600s) The fundamental lesson of the Ukraine war is not any singular weapon system like drones, but the underlying, self-sustaining innovation ecosystem capable of continuous adaptation. - [11:45](https://www.youtube.com/watch?v=qhIabMKHJlA&t=705s) Hardware systems are deliberately engineered to be 'as complex as a washing machine' so production can realistically scale from 1,000 to 10,000 units monthly without bespoke manufacturing bottlenecks. - [11:45](https://www.youtube.com/watch?v=qhIabMKHJlA&t=705s) Stark structures internal engineering like a venture portfolio, running 10 to 15 customer-facing programs supported by foundational capability platform teams solving cross-cutting problems like GNSS-denied navigation and edge computer vision. - [15:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=900s) A publicly scrutinized trial failure for the Bundeswehr served as a turning point, prompting an emergency deployment to a freezing test range where rotating teams worked round-the-clock until achieving three consecutive tank strikes. - [21:13](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1273s) European tech talent has experienced a cultural shift, removing historical stigmas around defense work and treating defense sovereignty as an urgent generational imperative. - [23:48](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1428s) Stark's strike drone mission flow relies on human-defined operational intent (corridors, no-fly zones, and kill zones), allowing systems to navigate autonomously through heavy electronic warfare while retaining modular modes for either autonomous or human-in-the-loop terminal engagement. - [27:22](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1642s) By shifting left on manufacturability, Stark prioritizes procurable, top-quartile commercial off-the-shelf components over fragile single-source proprietary ASICs, enabling international customers to integrate domestic supply chains. - [30:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1800s) While standard components can be sourced across allied European suppliers, deep structural Western defense dependencies remain unresolved in raw chemicals and base materials tied to China. - [32:53](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1973s) Stark adheres to NATO's principles for responsible AI and IEC 61508 functional safety standards, implementing layered safeguards and hard-coded constraints that prevent warhead arming over civilian zones. - [32:53](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1973s) The ground control system Minerva uses AI to automate tactical planning, calculating the exact threat allocation and selecting the most cost-effective drone sortie before launch. - [40:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2400s) Deterrence requires rewriting defense contracts to decouple 3-year hardware lifecycles from weekly software update loops, guaranteeing elastic manufacturing surge capacity so industrial readiness actively prevents conflict escalation. - [41:30](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2490s) There is no existing acquisition playbook for rapid-iteration defense hardware and software, forcing modern defense startups, military operators, and procurement ministries to co-develop new models in real time.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=0s) Billions in Contracts in 18 Months: Stark's Four-Part Playbook

Pat Grady introduces co-founders Uwe Horstmann (CEO) and Johannes Schaback (CTO) of Stark, describing the company as a European "neo-prime." Right off the bat, Pat points out a milestone that is virtually unheard of in the traditional defense sector: landing billions of dollars in contracts within just 18 months of launching. While acknowledging the sheer scale of the success, Uwe reflects that the achievement reflects a broader, optimistic shift across the continent, showing that Europe is finally heading in the right direction rather than just celebrating one company's balance sheet.

```mermaid flowchart TD accTitle: Stark Four-Part Neo-Prime Playbook accDescr: The four core pillars that enabled Stark to scale rapidly and win major defense contracts.

subgraph RND["1. Development & Iteration"] A["Deep Battlefield Integration<br>(Ukrainian front lines & European forces)"] B["Venture-Backed Independence<br>(Self-funded R&D, not government grants)"] end

subgraph DEPLOY["2. Deployment & Scale"] C["End-to-End Usability & Certification<br>(Field-ready, safe, & NATO-compliant)"] D["Industrial Scaling from Day One<br>(Production is deterrence; factory as a weapon)"] end

A --> B B --> C C --> D ```

Uwe outlines the four foundational pillars behind Stark's execution: 1. **Innovation directly on the battlefield:** Embedding engineers and teams alongside frontline operational units who actually fire and operate the hardware. 2. **A venture-first model:** Funding product engineering and R&D with private venture capital (from firms like Sequoia) rather than waiting around for slow government development grants. 3. **Designing for operational usability and safety:** Moving beyond flashy prototypes to do the hard work of NATO validation, qualification, safety testing, and real-world compliance. 4. **Industrial scaling from day one:** Treating manufacturing capability itself as an instrument of national power. As Uwe emphasizes: “Like, how can we make sure that those are not only a few exquisite systems that, you know, stand around somewhere, but how can we make sure that, you know, factory is part of the weapon, you know, production is deterrence?”

[02:28](https://www.youtube.com/watch?v=qhIabMKHJlA&t=148s) Battlefield Proximity and Why the Staff Sergeant Matters More Than the General

When Pat asks how an early-stage startup managed to convince military organizations to let civilians walk into active units, Uwe points to their massive, direct presence on the ground in Ukraine. Stark embedded teams deeply with fighting units there, alongside expanding formats across European forces that allow engineers to join soldiers directly at training and testing grounds. Uwe distinguishes between the institutional buyer (a high-ranking civilian bureaucrat or general) and the actual customer: an operational soldier or staff sergeant in a trench. Gaining direct feedback from the tactical operator is Stark’s single most critical input factor.

At any given time, Stark manages a venture-style portfolio of 10 to 15 parallel programs. Their current flagship system is a software- and AI-driven strike munition designed to be orders of magnitude cheaper than legacy precision-guided weapons. Johannes and Uwe highlight that designing these weapons requires grappling with user psychology: what does operating an autonomous drone swarm actually do to an enlisted soldier's mental load?

``` Traditional Primes Stark's Neo-Prime Model ------------------ ----------------------- Buyer-First: Generals & high-level User-First: Staff sergeants & tactical officials dictating long cycles operators iterating on the ground

Vendor Push: "Dumping stuff on them Feedback Loop: "Tell us how exactly this and saying... you didn't understand it" is shit... we'll be a little bit less shit tomorrow"

Exquisite Assets: Small quantities of Scalable Mass: Inexpensive, software-defined, high-cost, bespoke machinery mass-manufactured autonomous munitions ```

This divide between clean engineering whiteboards and mud-soaked trenches is best illustrated by the packaging of their drones. Johannes recalls shipping their strike systems in sleek, lightweight plastic transit cases, convinced they had engineered the perfect transport solution. The soldiers at the front immediately rejected them and demanded plain wooden crates instead: - Plastic transit cases cannot be safely burned in a frontline position; they melt, smoke, and leave conspicuous trash that reveals a hidden team's location. - Wooden crates can be quietly broken apart, chopped up, and used for heating, cooking rations, or fortifying trenches.

Johannes explains that traditional defense contractors rely on high-ranking brass, but Stark flips the paradigm: “So traditionally, you would say, the most important soldiers in the product development and the procurement are the generals, who tell you that's what a big budgets are and you should do this and that. And for us, we always like to say, you know, the staff sergeants, the enlisted personnel who then actually use those systems are maybe even more important to us because in a true startup spirit, you want to be close to the customer.”

[05:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=300s) Rapid Iteration in the Trenches: "A Little Bit Less Shit Tomorrow"

The tactical reality of drone warfare exposes countless usability details that traditional military specs ignore. Johannes points out that controlling a swarm raises practical human-interface questions: does an operator use a mouse, a rugged keyboard, or a touchscreen? How do those controls respond when a soldier is wearing heavy, wet winter combat gloves? How much sensor feeds and targeting data can a single operator absorb under fire before cognitive overload sets in?

Legacy defense contractors routinely dismiss frontline complaints by arguing that the soldier simply failed to understand the system. Stark embraces the startup ethos of rapid iteration and radical humility. When operators frankly tell them that a prototype is “absolute shit,” the team takes notes, works through the night, and brings back a refined version within 24 to 48 hours to prove they are listening.

> “And sometimes we show up and they say like, you know what? This is absolute shit. But you can really win trust by saying, tell us how exactly this shit. And we'll show up again tomorrow or the day after tomorrow. We create a new version and we'll be a little bit less shit tomorrow.” — Johannes Schaback

For Stark, standard commercial metrics like Net Promoter Score (NPS), retention, and ARR translate directly into frontline trust: “Listen to the soldiers, build with them, build for them, and hopefully good things will happen on the procurement side as well.”

[07:42](https://www.youtube.com/watch?v=qhIabMKHJlA&t=462s) The European Awakening: Deterrence and the Neo-Prime Window

Luciana Lixandru steps back to examine the macro climate in Europe, noting that Uwe—a former soldier himself—had recognized the defense opportunity and began investing in the space long before the wider European venture ecosystem deemed it acceptable. Uwe explains that Europe has experienced an intellectual and geopolitical sea change: a collective recognition that the continent can no longer outsource its security.

> “I think what came together is obviously an understanding in Europe that Europe needs to take more responsibility for its own defense. You know, that is, of course, something that the US have told us repeatedly. And they're right. You know, Europe has to step up and needs to be able to defend itself. Because that is our way of continuing to live in peace. It's deterrence, right?” — Uwe Horstmann

This geopolitical wake-up call arrived simultaneously with a technological revolution on modern battlefields: - Warfare has shifted decisively toward agile, low-cost, AI- and software-driven hardware that can be manufactured at vast scale. - European governments increasingly realize that buying a modern autonomous drone system cannot follow the multi-decade procurement frameworks created for naval battleships. - Gaps in existing military capability forced defense ministries to look outside the legacy primes for high-velocity software-first defense technology.

For Uwe, the convergence of his military and geopolitical background with venture-backed tech made founding Stark an unavoidable calling, prompting him to ask what his generation of builders will have contributed when looking back at this historic pivot.

[10:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=600s) European Procurement Reality and the Innovation "Wonder Weapon"

Uwe Horstmann reflects on the institutional transformation sweeping European defense, expressing confidence that reforms in procurement and certification are lasting. He highlights how cultural shifts occur in Europe: unlike environments driven by charismatic leaders rallying followers around grand narratives, change in Europe takes root when forced by external pressure. As Horstmann describes it, “Changes like that in Europe, they don't happen because there is like a charismatic leader who says, you know, we need to change. This is the way. Follow me. Here's a great narrative. That's not how Europe works. Europe works based on... well, I guess we kind of have to do it now. It's reality kicking in.”

Because these shifts emerge from structural necessity, Horstmann argues that reforms compressing multi-year procurement and certification cycles down to months will endure. More fundamentally, he contends that observers focusing on individual systems miss the deeper lesson of the Ukraine war. Rather than obsessing over hype about whether drones decide modern combat, the ultimate determinant is building a self-sustaining defense ecosystem capable of rapid, continuous iteration:

> “The real learning from the Ukraine-Russia war is probably not drones, no drones. It is about an innovation ecosystem, that is the actual wonder weapon.” — Uwe Horstmann

This broader ecosystem requires aligning several key inputs: direct tactical feedback from combat soldiers, attracting the smartest engineering talent into national security, and securing private and public capital.

[11:45](https://www.youtube.com/watch?v=qhIabMKHJlA&t=705s) Startup DNA, the "Washing Machine" Metric, and Venture-Style Programs

When asked about Stark's advantages as a two-year-old defense company competing against 70-year-old defense primes, Horstmann emphasizes their startup DNA in mass-scalable, software-driven, low-cost systems such as strike drones and unmanned surface vessels. Where traditional primes excel at crafting exquisitely engineered, highly expensive, and low-volume defense assets, Stark designs every system from the ground up for volume production.

```mermaid flowchart TD accTitle: Stark Organizational Architecture accDescr: Venture-style autonomous programs supported by foundational capability platform teams

subgraph Programs["Autonomous Frontline Programs (Customer-Facing)"] A["Aerial Systems / Strike Drones"] B["Unmanned Naval Vessels"] C["Electronic Warfare & Software"] end

subgraph Capabilities["Central Platform Layer (Capability Teams)"] D["GNSS-Denied Navigation"] E["Object Classification & Re-ID"] F["Communication Links & Flight Control"] end

D -->|"Compounding technology"| Programs E -->|"Compounding technology"| Programs F -->|"Compounding technology"| Programs ```

Horstmann shares an anecdote about welcoming operational manufacturing experts into their facilities: - The visiting experts remarked, “Oh, this is about as complex as a washing machine.” - Horstmann embraced the comparison: “I don't know if they tried to insult those, but it is exactly where I want to be.” - If an executive is tasked with scaling production from 1,000 washing machines a month to 10,000 a month or 1,000,000 a year, the goal feels operationally tractable, whereas scaling traditional military hardware at that speed is nearly impossible.

Johannes Schaback explains that Stark structures its engineering organization internally like a venture capital fund, creating small, cross-functional “startups within a startup.” Each independent program tackles a customer domain (aerial drones, autonomous naval vessels, electronic warfare, or core mission software). Underpinning these frontline programs is an engineering platform layer known as "capability teams." These capability teams build cross-cutting technologies—such as GNSS-denied navigation, edge object classification, target re-identification, flight control algorithms, and resilient communication links—that compound across all active defense systems.

[15:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=900s) Sub-Meter Strike Accuracy and the Virtus Trial Crucible

Schaback identifies terminal strike accuracy as Stark's proudest technical capability. Achieving sub-meter precision against moving targets traveling at high operational speeds requires deep co-design of hardware, actuation, and real-time edge computing. Simply executing a flight model in simulation or deploying an isolated neural network is insufficient; physical compute budgets, latency constraints, and memory allocations must match real-world aerodynamic forces and camera fields of view under tactical stress.

Schaback then details what he describes as a crucible event for the young company: a high-stakes, publicly scrutinized trial for the German Bundeswehr involving their Virtus system:

- **System Regression:** A major upgrade to both hardware and software created physical and sensor misalignments, shifting the center of gravity and altering the camera opening angle, which required retraining edge perception models. - **The Bundeswehr Challenge:** The German military invited Stark to test the unfinalized system on an aggressive schedule. Stark chose to participate despite knowing the system had not yet achieved its internal performance benchmarks. - **Media Backlash:** The experiment faced setbacks in an unoptimized test environment. While military stakeholders aimed to stress-test the system, external media outlets seized upon the stumble, amplifying the narrative of failure. - **The Emergency Deployment:** Stark convened an all-hands meeting, framing the situation as an existential test of their engineering capabilities. - **Improvised Cold-Weather Base:** Stark deployed engineers to a three-star spa hotel near a test range in November. Facing freezing temperatures, they purchased equipment from local DIY home-improvement stores, erected shipping containers on-site, deployed mobile power generators, and routed Starlink terminals to stream live flight telemetry.

[18:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1080s) The Round-the-Clock Comeback and Three Consecutive Hits

Working 20-hour shifts across three rotating teams, the engineering crew established a rapid feedback loop between freezing outdoor flight sorties and warm indoor data processing rooms. As test pilots gathered flight logs and target imagery, software engineers retrained models and calibrated flight control algorithms, while hardware engineers repaired and tuned physical flight surfaces and mechanical actuation systems.

The team established an uncompromising threshold before they would pack up and leave the range: - They set a rule that they would only depart after recording three consecutive direct hits on armored targets. - By the end of the two-week intensive push, the target tanks on the range were completely perforated from repeated strikes. - The company expended dozens of expensive test drones to conquer the edge cases, adopting a mindset where cost was completely subordinated to validating terminal accuracy.

Schaback concludes that through this rapid, continuous cycle of real-time testing and immediate hardware-software repair, Stark completely solved the regression and proved sub-meter terminal accuracy under harsh military testing conditions.

[20:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1200s) The Turning Point: Simulation, AI Flight Dynamics, and First Hits

Following their high-stakes failed trial, Stark’s engineering teams split into two intense parallel work streams on the ground and behind screens. At the freezing test range, field engineers scavenged debris from their own crashed drones, piecing together salvaged wings and motors to assemble enough workable airframes for extra test launches. Simultaneously, the simulation team worked remotely to overhaul the system's flight dynamics modeling. Stark had already integrated artificial intelligence into their engineering toolchain to model flight aerodynamics in simulated environments, using machine learning to boost the simulator's fidelity so the autonomous guidance models could train on realistic aerodynamic behaviors.

Within less than a month, the physical test data and the upgraded digital flight models converged. When the team launched the revised system, the drones struck their targets cleanly. That breakthrough marked a permanent turning point for Stark’s engineering cadence, transforming an embarrassing initial failure into a reliable baseline for all future field trials.

Johannes Schaback captures that transformative moment: “And after two weeks, as a matter of fact, it was less than four weeks. There was this one moment where we realized, okay, that's it. And from there on, we've always hit.”

[21:13](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1273s) Cross-Pollinating Talent: Startup Speed Meets Legacy Defense

Building a defense prime from scratch requires blending institutional defense knowledge with Silicon Valley-style software velocity. Johannes explains that Stark recruits across four distinct pools: traditional defense primes, universities, fast-moving tech startups, and governmental or military institutions. While legacy defense primes provide critical domain knowledge regarding qualification and defense standards, startup talent provides the bias toward rapid iteration. Stark integrates these divergent backgrounds through cross-functional program units. Each program operates autonomously to ship products, but functional heads oversee disciplines across mechanical engineering, electrical engineering, and software. These leads organize internal "communities of practice" that align engineering standards, tooling choices, and AI implementation across the entire company.

At the same time, a broader geopolitical shift has altered tech workers' attitudes toward defense across the European continent. Before the war in Ukraine, working on weapons systems was widely considered a career taboo for top software and hardware engineers. Today, engineers recognize defense as an urgent civil necessity. Uwe Horstmann emphasizes that Stark acts as an ideal landing pad for engineers who want to dedicate their skills to European security without getting bogged down in traditional defense bureaucracy.

Uwe highlights this shift across the European tech ecosystem: “We observe all over Europe, in all the countries basically, the people understand this is a generational challenge this is a generational task it's time to step up for it.” He underscores that Stark's door is wide open to nontraditional backgrounds: “You don't need to have worked in defense if you want to work at Stark.”

- **Talent Sources:** Legacy primes, tech startups, universities, and government bodies. - **Internal Matrix Structure:** Product-driven cross-functional teams paired with functional heads who manage horizontal communities of practice for mechanical, electrical, and software engineering. - **Cultural Reset:** Tech workers increasingly treat European defense sovereignty as a generational imperative rather than an off-limits industry.

[23:48](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1428s) Anatomy of an Autonomous Strike: From Wooden Crate to Kill Zone

Asked to deconstruct a single strike mission from launch to impact, Johannes begins not with flight controls, but with packaging and logistics. Wars are won and lost on the supply chain, meaning a weapon system must first be easily disassembled, stored in simple wooden boxes, transported under extreme shock and vibration, and withstand ambient pressure variations whether shipped by air or by sea. Systems must also maintain up to a ten-year shelf life in military depots before being deployed to the frontline, where operators unpack and mount them onto a catapult launcher.

Operationally, Stark’s flight stack relies on high-level operator intent rather than continuous manual piloting. Before launch, the human operator inputs the mission envelope—defining transit corridors, geo-fenced no-fly zones, and the terminal kill zone. Once launched, the drone navigates autonomously through GPS-denied and heavily jammed airspace, actively dodging anti-aircraft batteries. When flying as part of a swarm, the drones mesh-network their sensor feeds to build shared situational awareness, communicating targets across the group. Stark deliberately designs these networking interfaces to remain modular and open, allowing their systems to share data with third-party platforms rather than locking militaries into a closed vendor ecosystem.

Once inside the designated kill zone, the drone's edge computing payload takes over. High-speed computer vision systems process sensor streams from off-the-shelf electro-optical and infrared (EO/IR) cameras, feeding real-time target perception into an onboard reasoning engine. Depending on the customer’s specific rules of engagement and concept of operations, the strike profile executes in one of two ways: either the drone identifies and hits the target fully autonomously, or it loiters and holds for a final human-in-the-loop strike confirmation before executing its dive.

Johannes explains the underlying software philosophy: “The system always implements and follows the intent.”

```mermaid flowchart TD accTitle: Anatomy of a Stark Autonomous Strike Mission accDescr: An end-to-end operational flow of a Stark drone strike from crated logistics to terminal engagement.

subgraph Logistics["Logistics & Deployment"] A["Flat-pack Wooden Crate<br/>(10-Year Shelf Life)"] --> B["Transport via Sea / Air<br/>(Shock & Pressure Proof)"] B --> C["Field Catapult Launch"] end

subgraph AutonomousTransit["Autonomous Transit & Swarm"] D["Operator Defines Intent<br/>(Corridors, No-Fly, Kill Zone)"] --> C C --> E["Autonomous Ingress<br/>(EW & Anti-Air Evasion)"] E <--> F["Mesh Swarm Comms<br/>(Shared Awareness)"] end

subgraph TerminalEngagement["Terminal Kill Zone"] E --> G["Edge Vision & Reasoning<br/>(EO/IR Sensor Fusion)"] G --> H{"Customer CONOPS"} H -->|"Autonomous Mode"| I["Direct Terminal Strike"] H -->|"Human-in-the-Loop"| J["Wait for Operator OK"] J --> I end ```

[27:22](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1642s) Vertical Integration vs. Local Sourcing, and Cross-Domain Expansion

Stark approaches the classic hardware question of vertical integration through the lens of customer need, balancing production volume against geopolitical resilience. While vertical integration offers direct control over high-volume output, it conflicts with allied defense requirements: purchasing nations want their own domestic defense suppliers integrated into the production chain to root manufacturing locally. To reconcile these forces, Stark builds around standard, procurable commercial components—such as off-the-shelf infrared and electro-optical sensors—and fuses them with proprietary edge software. This modular architecture allows Stark to adapt rapidly as tactics and target profiles change daily on the Ukrainian battlefield, avoiding brittle reliance on custom silicon ASICs that cannot easily be swapped.

This core software architecture also drives Stark’s decision to pursue a multi-product portfolio across multiple domains rather than remaining a single-product drone startup. When militaries publish explicit capability gaps, Stark assesses whether an existing market solution exists or if Stark has the technical authority to build one. Because the foundational building blocks—such as mission intent planning, navigation in electronically jammed environments, computer vision target classification, and terminal homing—are structurally identical across dimensions, software developed for aerial drones translates directly to maritime vessels. As Stark expands into manufacturing at massive scale, every program balances rapid time-to-market against long-term, high-volume production efficiency.

Uwe Horstmann explains the delicate trade-off between proprietary manufacturing and supplier flexibility: “The governing principle for us is how do we solve the customer issue? And the scalability of output is a big customer issue. So that would be my first question towards vertical integration.” He summarizes their twin imperatives: “Number one principle, can we solve the scaling issue? Number two, how can we plug in different suppliers to make everything more resilient?”

- **Modular Edge Hardware:** Stark avoids single-purpose proprietary ASICs, opting instead to procure high-volume, off-the-shelf EO/IR sensors and unify them via custom edge sensor-fusion models. - **Dual-Track Sourcing:** Scales core production internally while enabling international military customers to integrate domestic sub-tier suppliers. - **Cross-Domain Software Reuse:** Navigation under electronic jamming, sensor fusion, computer vision, and operator mission-intent interfaces apply equally to aerial and maritime systems.

[30:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1800s) Shifting Left on Manufacturability and Supply Chain Resilience

Speed to market and production readiness frequently pull in opposing directions. If an engineering team optimizes purely for speed, they can deliver a functioning prototype for a demonstration remarkably fast, but manufacturing it at scale becomes a nightmare. Johannes Schaback uses a motorsport analogy to illustrate the trap: a Formula One team builds an exceptionally fast car on a tight deadline, but they only ever build one or two of them, not tens of thousands. To avoid that pitfall, Stark deliberately “shifts left” on manufacturability during early design. As Johannes explains:

> “We shift left on manufacturability. That's extremely important because otherwise you accrue a debt that you have to fix later on.”

This engineering philosophy directly dictates component procurement. Early in the design cycle, supply chain resilience takes precedence over raw technological novelty. Returning to the washing machine comparison, Stark does not seek out the most bleeding-edge camera sensor on the market if it cannot be sourced reliably under wartime conditions; instead, they select proven, top-quartile components that can be procured in massive volumes.

When asked whether fully localized European defense supply chains are realistic, Uwe Horstmann notes that for many standard components, diversification across multiple allied suppliers and countries is completely achievable. While conventional business wisdom might question whether that redundancy hurts bottom-line margins, it is essential when the entire business model is built on resilience. At the same time, Uwe acknowledges that Western defense manufacturing as a whole still harbors critical dependencies on Chinese supply chains, particularly for raw chemical precursors and base materials. Solving those deep vulnerabilities exceeds what any single company can execute on its own, requiring an industry-wide coordination effort between commercial manufacturers, ministries of defense, and political leadership.

- **Manufacturability first:** Stark shifts manufacturing considerations into the earliest design phases to prevent downstream production debt. - **Top-quartile component strategy:** Sourcing dependable, high-volume components over experimental, single-source bleeding-edge parts. - **Supply chain sovereignty:** Simple components can be diversified across European and allied suppliers today, but fundamental bottlenecks—especially in raw chemicals—require coordinated industrial policy.

---

[32:53](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1973s) After the Hugging Face Hack: Guardrails on Autonomous Weapons and the AI Mission Stack

Addressing the wider shift from exquisite, human-operated platforms toward autonomous and attritable systems, the discussion turns to the safety risks of military artificial intelligence—especially in light of recent incidents like the Hugging Face hack, where autonomous agents broke out of their software sandboxes. Johannes emphasizes that Stark has strictly anchored its software architecture in NATO’s principles for the responsible use of AI since day one. The autonomy stack is constrained by layered, human-audited guardrails that force the software to operate within a strictly defined corridor reflecting human command intent:

> “Stark has since the beginning implemented NATO's principle of responsible use of AI. And that entails that we're auditing our own systems, not only by humans, but also by safeguard or layers that safeguard the AI in a corridor to be restricted to implement the operator's intent.”

These safety layers include hard-coded rules derived from functional safety standards like IEC 61508 and practices borrowed from autonomous automotive driving. For instance, an airborne strike drone has strict geofenced software locks that categorically prevent it from arming its warhead if its flight path passes over an inhabited civilian zone.

```mermaid flowchart TD accTitle: Stark Autonomy Stack and Mission Flow accDescr: Diagram illustrating how operator intent and hard-coded safeguards govern target evaluation and mission execution through Minerva.

subgraph GroundPlanning ["Ground Control & Mission Planning"] OP["Human Operator Intent"] MIN["Minerva Ground Control System"] ALLOC["Threat Evaluation & System Allocation"] OP --> MIN MIN --> ALLOC end

subgraph Safeguards ["Layered Safety Corridors"] GEO["Hard-Coded Constraints (e.g., Inhabited Zones)"] IEC["IEC 61508 Functional Safety Standards"] AUDIT["Continuous Human & Automated Auditing"] end

subgraph AirborneExecution ["In-Flight Autonomy Stack"] SENS["Onboard Perception & Environment Interaction"] ENG["Terminal Guidance & Target Engagement"] end

ALLOC -->|"Optimized Flight & Engagement Plan"| AirborneExecution Safeguards -.->|"Enforces Operating Corridor"| AirborneExecution SENS --> ENG ```

Johannes sketches out where AI operates across the lifecycle of a mission, noting that it touches perception, environmental navigation, and pre-flight ground planning. Historically, military mission planning was slow, manual, and cumbersome. Under Stark's setup, the ground control station—named Minerva—automates that tactical algebra. If an incoming threat appears and an operator has three strike drones on standby, Minerva evaluates the threat profile, determines how many drones are needed to defeat it, selects the most cost-effective solution, and optimizes the flight path before the operator launches the system.

- **Governing framework:** Built around NATO’s principles of responsible AI, combining automated safety layers with mandatory human auditing. - **Hard-coded constraints:** Built-in negative triggers (borrowed from IEC 61508 automotive and industrial safety practices) prevent arming over civilian or restricted zones. - **Minerva mission planning:** AI calculates resource allocation on the ground, helping the operator decide the exact number of assets needed to neutralize incoming threats at the lowest operational cost.

---

[35:56](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2156s) Matching Primes on R&D and Scaling to 700 People

When asked why traditional defense primes failed to seize the market for low-cost strike drones and maritime autonomous systems, Uwe explains that legacy contractors are structurally misaligned with mass-manufactured, low-cost hardware. While a few legacy players have attempted to adapt, several primes have explicitly told Stark leadership that building cheap, attritable systems is fundamentally outside their organizational DNA. Stark welcomes competition from legacy primes, arguing that a competitive procurement race ultimately sharpens European defense capabilities across the board.

Addressing whether legacy primes hold an insurmountable advantage in capital, Uwe pushes back by comparing genuine R&D investments rather than topline revenue. Over a 12-to-36-month horizon, Stark plans to deploy roughly €600 million (well over $650 million) into core research, product development, and factory production capacity. Once government-funded, pass-through research contracts are subtracted from the primes' balance sheets to isolate their true internal, equity-funded R&D, Stark's deployment is directly comparable:

> “And if you subtract all the R&D spending that is government funded, but the actual core R&D spending that those companies do, we're not way off.”

Managing the hypergrowth of expanding from zero to roughly 700 employees within two years requires strict organizational blueprints. While rapid expansion creates friction, Uwe insists that urgency is a moral and strategic imperative because the security crisis facing Europe is unfolding right now:

> “Being slow is not an option. And so allocating these 700, not all engineers, but the majority are engineers, to extremely important and valuable programs is key.”

Looking ahead at technical roadmaps, Johannes highlights key focus areas: engineering systems that fly significantly faster, bringing low-latency edge compute onboard to handle dynamic multi-domain targeting across maritime and aerial domains, and exploiting low-cost software-defined radios (SDRs) alongside accessible compute to radically reduce the cost of electromagnetic spectrum sensors. Meanwhile, Uwe points to what he considers the next structural frontier for deterrence: redesigning defense procurement contracts themselves to provide elastic, software-like surge capacity.

- **Capital parity in core R&D:** Stark’s planned €600M+ capital deployment into internal R&D and manufacturing matches the non-subsidized, company-funded R&D budgets of major European primes. - **Workforce composition:** The company scaled to approximately 700 people in 24 months, with engineers making up the vast majority of the staff. - **Next technical frontiers:** Faster flight speeds, adaptive multi-domain target tracking, low-cost electromagnetic sensing via software-defined radios, and novel contract structures that treat production capacity as an active deterrent.

[40:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2400s) Deterrence by Contract: Surge Capacity, Stockpiles, and Continuous Software Updates

Uwe Horstmann outlines a fundamental shift in how defense contracts must be conceived and structured. While selling hardware in increments of hundreds of units represents normal business, the more transformative challenge lies in designing contracts that actively prepare for massive surge and rapid manufacturing scaling in the event of high-intensity conflict. Real deterrence, Uwe argues, does not stem merely from the isolated technical capabilities of an individual product deployed here or there. True deterrence comes from industrial readiness—having a baseline physical stockpile paired with contractual agreements where governments pre-pay for reserve manufacturing capacity that can spin up instantly when a crisis strikes.

This contractual evolution also requires solving the diverging lifecycles of modern defense hardware and software. Johannes Schaback highlights that while mechanical and physical components—such as electric motors or airframes—may remain viable and sit in storage for three years, autonomous software stacks cannot sit idle. Software requires near-continuous refinement, ingesting fresh training data, model updates, and electronic warfare countermeasures on a weekly cadence. Structuring long-term procurement frameworks to accommodate this disparity is critical to keeping stockpiled systems operationally relevant.

As Johannes Schaback observes regarding the mismatch between hardware durability and software cycles: “Probably if you look at a drone at the moment, probably the motors, they're good for three years, but a software probably should be updated every week with new training data and so on and so forth.” Uwe emphasizes that solving this contractual puzzle is vital, noting: “That's something that's very interesting to me, because that is probably where the real deterrence comes from, I think.”

```mermaid flowchart TD accTitle: Contractual Surge and Deterrence Model accDescr: Stark's framework for defense contracts balancing baseline hardware stockpiles, continuous software iteration, and rapid manufacturing surge capacity. subgraph Readiness["Baseline Readiness Layer"] Hardware["Hardware Stockpiles<br>(Motors & airframes rated for ~3-year shelf life)"] Software["Continuous Software Stack<br>(Weekly updates with training data & EW patches)"] end subgraph Contracting["Contractual Architecture"] Prepaid["Pre-paid Manufacturing Capacity"] SurgeProtocol["Surge & Rapid Scaling Commitments"] end subgraph Impact["Strategic Outcome"] Deterrence["Credible Industrial Deterrence<br>(Production readiness prevents conflict escalation)"] end

Hardware --> Prepaid Software --> SurgeProtocol Prepaid --> Deterrence SurgeProtocol --> Deterrence ```

- Traditional defense procurement focuses strictly on purchasing static quantities of finished goods, which leaves production lines cold between orders. - Under Stark’s proposed framework, governments contractually secure warm factory lines and reserve tooling, treating manufacturing bandwidth like cloud capacity with on-demand surge elasticity. - Maintaining operational effectiveness requires decoupling hardware storage warranties from software update schedules, ensuring that field algorithms reflect the latest operational data.

---

[41:30](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2490s) No Existing Blueprint: A Shared Learning Curve Across Industry and Government

When asked whether any Western nation or defense contractor has already solved this model of contractual surge capacity and rapid technological refresh, the Stark founders acknowledge that no established playbook exists. Across both procurement mechanisms and hardware-software product development, there is no seasoned cohort of hundreds of veterans who have already navigated this exact transition and can hand down an instruction manual. Instead, modern defense tech companies, armed forces, and procurement ministries are figuring out the new paradigm together in real time.

Despite the steep learning curve, there is a tangible, optimistic spirit bridging the traditional divides between government agencies, military end users, and commercial industry. Stakeholders recognize that obsolete peacetime acquisition rules cannot meet current geopolitical threats, creating an urgent willingness to experiment and iterate collaboratively. The objective is to align what is technologically feasible with what is militarily required to prevail in modern warfare—ultimately establishing a sustained industrial posture capable of preserving peace.

Reflecting on the absence of established models across the ecosystem, Uwe Horstmann explains: “There is not a blueprint; that is also for all the things we do in procurement and all the things we do in product development.” He points out the collaborative effort underway across sectors, adding: “There are not hundreds of people who have done it and can now just teach us; I think we are all learning it together.”

- The transition toward software-defined, mass-manufacturable defense systems requires rewriting acquisition frameworks across military, governmental, and industrial branches simultaneously. - Traditional prime contractors historically optimized for lengthy development cycles and proprietary lock-in, leaving Western defense ministries without turnkey precedents for rapid-iteration programs. - The shared recognition of technological reality has fostered unprecedented openness among military leadership to adopt startup-style development velocity and iterative contracting.

---

[42:45](https://www.youtube.com/watch?v=qhIabMKHJlA&t=2565s) Stark's Core Mission: Becoming an Anchor Pillar for European and Western Defense

Addressing the overarching mission of Stark, Uwe Horstmann makes it clear that the ambition is to serve as one of the definitive industrial pillars of the future European and Western defense architecture. Achieving this status does not mean operating as an isolated, closed-off conglomerate; rather, Stark intends to be an open collaborator that enables a thriving defense innovation ecosystem. This includes actively partnering with established defense primes, specialized component suppliers, and emerging software startups alike to ensure the Western alliance maintains unmatched manufacturing and technological depth.

The ultimate measure of success for Stark is not merely corporate scale or contract value, but the geopolitical outcome their systems produce. The founders reflect on the long-term vision: sitting down five or ten years into the future and seeing that major conflicts around the world were deterred and never flared up in the first place because Western industrial readiness was too formidable to challenge.

Articulating the company's defining objective, Uwe Horstmann states: “The mission is quite clearly: we want to be one of the absolute pillars of this EU and Western defense architecture.” Looking ahead to the ultimate test of their work, he adds: “And if we succeed, and we sit here 5 or 10 years from now, and the wars and situations around the world and others have never flared up.”

- Stark aims to combine the mass-production capabilities of a prime contractor with the agile software development practices of an innovation ecosystem anchor. - Collaborative integration across large defense incumbents and small cutting-edge suppliers is viewed as essential to building resilient, sovereign European defense manufacturing. - The company evaluates its long-term deterrent value by whether credible mass-manufacturing capacity successfully dissuades adversaries from initiating conflict.

## Notable quotes > [10:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=600s) The real learning from the Ukraine-Russia war is probably not drones, no drones. It is about an innovation ecosystem, that is the actual wonder weapon. — Uwe Horstmann

> [30:00](https://www.youtube.com/watch?v=qhIabMKHJlA&t=1800s) We shift left on manufacturability. That's extremely important because otherwise you accrue a debt that you have to fix later on. — Johannes Schaback

> [07:42](https://www.youtube.com/watch?v=qhIabMKHJlA&t=462s) Europe has to step up and needs to be able to defend itself. Because that is our way of continuing to live in peace. It's deterrence, right? — Uwe Horstmann

## Mentioned Stark, Uwe Horstmann, Johannes Schaback, Pat Grady, Luciana Lixandru, Sequoia Capital, Bundeswehr, Virtus, Minerva, NATO, Hugging Face, Starlink, IEC 61508, Ukraine, Europe

## Video description Two years after its founding, Stark has roughly 700 employees and strike drones in the field in Ukraine. CEO Uwe Horstmann and CTO Johannes Schaback argue that the real lesson of that war is not any single weapon: the innovation ecosystem behind it is the wonder weapon. They explain why Stark designs every system for mass manufacturing from day one, runs 10 to 15 programs like a venture portfolio, and views the staff sergeant who operates a system as just as important as the general who buys it. Johannes recounts the publicly reported failed trial that became a forging moment, as engineers worked around the clock at a freezing test range until they landed three consecutive hits. They discuss NATO’s principles of responsible AI use and the hard-coded constraints in Stark’s autonomy stack. And Uwe lays out his next big idea: defense contracts structured like cloud contracts with surge capacity built in, so production itself becomes deterrence.

Hosted by Pat Grady and Luciana Lixandru, Sequoia Capital

Chapters

0:00 – Introduction 1:03 – Billions in contracts in 18 months: Stark's four-part playbook 2:28 – The staff sergeant matters as much as the general, and wood beats plastic 7:42 – Why Europe is ready for a neo prime, and why procurement reform will last 11:45 – Startup DNA, venture-style programs, and "as complex as a washing machine" 16:27 – Born out of fire: the failed trial and the round-the-clock comeback 21:13 – Where Stark finds talent, and why defense is no longer off-limits 23:48 – Anatomy of a strike, from the wooden crate to the kill zone 27:22 – Vertical integration, multiple domains, and shifting left on manufacturing 32:53 – After the Hugging Face hack: guardrails on autonomous weapons 35:56 – Matching the primes on R&D and scaling to 700 people in two years 39:24 – Deterrence by contract, and Europe's next chapter with the US

[Source](https://www.youtube.com/watch?v=qhIabMKHJlA)