Anduril: Building autonomous defense systems at startup speed

Executive overview

Anduril builds software-first defense systems that operate at Silicon Valley pace—delivering in months what traditionally takes years. Founded by Paul Schimpf (ex-Oculus) and Brennan Schimpf, the company directly challenges the legacy defense industry's cost-plus model, which removes incentives for efficiency and innovation. The core insight: modern warfare demands autonomous systems that can scale cheaply and intelligently, not expensive, specialized hardware operated by dozens of people.

How the defense industry got stuck

Throughout the Cold War, defense drove innovation at incredible pace (Pentagon built in 13 months, ICBMs in years). By the late 1970s, DoD shifted to cost-plus contracts prioritizing predictability and process over speed. The 1990s consolidation left just five major players controlling most contracts, killing incentives to innovate. Legacy primes now spend minimally on R&D—they finance dividends instead—and design products that require massive teams, long timelines, and high costs. A single F-35 program costs $1.5 trillion over its lifecycle and entered service decades after conception.

Why Anduril can succeed where others can't

The company operates on firm fixed-price contracts and invests heavily in its own R&D. This aligns incentives: they win by shipping faster and cheaper, not by extending timelines. Their Lattice software platform (not a drone, not a submarine) is the foundation—a general-purpose sensor fusion and decision-support stack that can power multiple hardware applications. This separates them from traditional primes that build completely independent software stacks for each aircraft or system.

Finding the first beachhead

Early attempts to sell to Marine Corps base protection failed due to budget cycles—they proved the concept worked, but funding was allocated years before the technology even existed. Success came with counter-drone systems during urgent operations. Special forces units facing an evolving aerial threat needed someone willing to iterate continuously rather than write five-year specs upfront. This became Anduril's playbook: find customers with urgent missions, real risk tolerance, and willingness to work with a new approach.

The autonomy-at-scale principle

Traditional Predator/Reaper drones require a dozen people operating each aircraft—pilots, sensor operators, exploitation teams. Anduril builds systems designed for single-operator or team deployment. Ghost drones autonomously scan areas, plot courses, and deliver footage; humans task them with mission objectives rather than flying them like joysticks. Larger systems follow the same rule: minimize manpower overhead, maximize what software handles. This unlocks deployment at scale and reduces cost per system dramatically.

Modern warfare favors disaggregated defense

Day-one conflicts (Gulf War model) still matter for overwhelming initial advantage. But extended conflicts—Ukraine, Armenia-Azerbaijan—prove that distributed, low-cost systems dominate. Tanks and armor become vulnerabilities; unmanned drones and precision munitions from small units win. Countries need defensive capability to disaggregate, disperse, and absorb early strikes. Anduril targets this asymmetric fight: cheaper, smarter systems that change the cost calculus. One tank cannot hide from many drones; many cheap systems overwhelm expensive air defense.

Products in production

• Base protection & border security: Automated towers with AI-driven threat detection; deployed at scale within three years.
• Counter-drone systems: Fast racing quadcopters that autonomously detect and intercept threats; program-of-record in ~2.5 years.
• Ghost drone: Autonomous helicopter in rugged case, deployable by small units for squad-level intelligence; now in production with adoption by UK Royal Marines.
• Tube-launch drones (via Area acquisition): Launch from helicopters, planes, ground, ships; all operate within Lattice ecosystem.
• Underwater vehicles (via Dive acquisition): 20-foot unmanned submarines for hundreds-of-kilometer missions; bringing existing teams into broader solution.

The cost-plus trap

Cost-plus contracts (7–12% fixed fee) reimburse all costs regardless of efficiency, removing risk from contractors and incentive to ship fast. Boeing's recent underbidding on firm-fixed-price contracts shows the corrosive effect: a company tooled for cost-plus can't operate efficiently elsewhere. NASA director publicly called cost-plus contracts "a huge anchor." Anduril invests own capital in R&D, bets on firm-fixed-price deals, and aligns with customer success rather than process compliance.

Ethical guardrails

Anduril refuses to build truly autonomous kill systems that decide targeting without human accountability. The company invests in humans having better information and control, not removing them from decisions. They've declined facial recognition projects they believe would be ineffective or irresponsible—choosing not to build something is as important as what they ship. Policy frameworks and democratic processes set boundaries; technologists' job is to inform what's possible and responsible, not circumvent ethics to show technical capability.

Hardware lessons for founders

Dozens of specialists are required: electrical, mechanical, manufacturing, supply chain. The cost to rework is not zero—it's expensive. The key is cheaply validating product-market fit before committing to expensive production. Prototype quickly and visibly, pull customers along the journey, prove you're solving real problems before building at scale. Blended strategies (co-investment with customers, shared risk) accelerate timelines and skip years of development.

The political reality

Anduril is bipartisan—Schimpf is a lifelong Democrat; the company attracts talent across the spectrum. Defense is massively popular and bipartisan in America; the company operates in regions where that's understood. Internally, zero political chat. The mission is the focus. The company is transparent: we build weapons, we take it seriously, we try to be ethical and thoughtful. Certainty and clarity build confidence.

Lessons from Palantir and SpaceX

Palantir taught the importance of controlling your destiny and understanding DoD's buying behavior: the government buys systems and capabilities, not tools. Subcontracting to larger primes is death—they minimize your revenue and maximize their unique positioning. You need something they cannot replicate or you'll lose control of your roadmap.

SpaceX proved that entrenched positions in the "hardest" sectors are not defensible. They went from zero to largest satellite constellation operator in two years. Speed, cost, and innovation shame legacy players. No sector is uniquely locked to incumbent players.

Both companies had to sue the government to succeed because DoD's institutional incentives punish admitting prior approaches were wrong. The smarter play: target novel problems with no existing approach, not spaces where you require institutional failure to be acknowledged. Avoid fights requiring admission of defeat; find markets where new technology enables something genuinely new.

Why openness works

Anduril openly shares its playbook for penetrating DoD. Selfishly, it makes DoD a better buyer—more competitors, more traction, stronger models for everyone. Unselfishly, national security is a mission problem. The company can't solve everything. Fresh competitors are needed. If the space improves, everyone wins.

Hardest decisions

Picking what to work on is the recurring hard call. Everything is expensive, requires years of conviction and investment. The company must achieve high batting average on success. Unlike typical startups, there's no single billion-dollar product; instead, multiple hardware products each capturing a market segment. That forces constant reevaluation and ruthless prioritization.

Greatest source of pride

Mission wins—seeing deployed systems solve consequential problems in the real world. Not titles, not contracts, but the fact that work changed outcomes.