How Anduril makes money: Building weapons the way Detroit made cars

Somewhere along the US southern border, a 33-foot steel tower scans the desert on its own. No operator watches a screen. Cameras, radar, and software do the looking and flag a human from miles away without anyone pressing a button. Anduril built that tower, and it also builds attack drones, jet-powered missiles, and autonomous submarines.

Anduril takes its name from Andúril, the sword reforged for Aragorn in The Lord of the Rings. Nine years ago, the company did not exist. In 2024, it made about $1 billion in revenue, and by 2025, that had more than doubled to $2.2 billion. Investors now value it at $61 billion, more than some of the century-old defense contractors it is trying to beat.

In this breakdown, we'll unpack exactly how Anduril makes money: how it charges for its weapons and software, what it spends to build them, who it competes with, and whether a company that loses money on purpose can grow into its $61 billion price tag.

Anduril was founded in 2017 by Palmer Luckey, the entrepreneur who built the Oculus virtual-reality headset and sold it to Facebook. He started Anduril with four co-founders, including Brian Schimpf, now Anduril's CEO. Its stated mission is to "rebuild the arsenal of democracy."

Anduril builds machines that fight or surveil without a person controlling them directly, and the software that runs them. It sells almost entirely to governments: the US Department of Defense first, and allies like Australia and the UK second.

The way Anduril makes money starts with a bet most defense companies would never take. The big legacy defense manufacturers, known in the industry as the primes, wait for the Pentagon to define what it wants. Then they win a contract that covers their costs plus a profit and start building.

Anduril does the opposite. It funds its own research, builds finished products before a customer has agreed to buy them, and then sells them at a fixed price. The primes' slow, cost-covered model is the inefficiency Anduril is built to capitalize on.

Tying all of it together is software called Lattice, an operating system for the battlefield. Lattice takes the feeds from every drone, sensor, and tower, fuses them into a single live picture, and lets one operator direct many machines at once. It is the glue and the part of the business that earns recurring revenue through subscriptions.

The hardware spans nearly every domain of warfare:

Behind those products sits real industrial weight. Anduril employs roughly 8,000 people as of 2026. Investors have put more than $11 billion into the company. The company is also building factories in Ohio, California, Mississippi, Rhode Island, and Australia.

Anduril makes money by selling autonomous systems and the software that runs them to governments. Its revenue has roughly doubled two years running: ~$1 billion in 2024, ~$2.2 billion in 2025, and a projected $4.3 billion in 2026.

Anduril is private and does not break out revenue by product, so its business is best understood by how it charges and the mission areas it sells into. The contract figures below are a rough guide to scale, not a record of exact sales.

Revenue is climbing fast, losses are widening, and the valuation is running far ahead of both.

Counter-drone and border surveillance is the company's oldest and steadiest source of cash. Sentry, Anduril’s autonomous towers, has been a US Customs and Border Protection program since 2020, with more than 350 fielded.

The border work has grown into contracts worth $363 million, plus a later task order of about $42 million. These systems are already deployed and running, so they behave more like a dependable revenue base than a speculative bet.

Stopping small drones is the other half of this line. It may be the larger one. Anduril holds a Special Operations counter-drone agreement worth up to roughly $1.9 billion, a $642 million Marine Corps counter-drone deal, and a nearly $2 billion approved foreign sale to Kuwait. The volume shows how central drone defense has become to the business.

Drones are a cheap, widespread threat. A $500 quadcopter can put a base at risk. Countering that at scale is exactly what Anduril's low-cost hardware and software are built for.

Anduril's boldest products are aircraft. Its centerpiece is Fury, an autonomous fighter jet (officially the YFQ-44A) built to fly alongside crewed jets as a robot wingman. Fury first flew in October 2025, and in June 2026, the Air Force picked it for production, one of two winners.

The Fury’s low cost is the pitch. Anduril targets a unit price near $30 million for Fury, about a third of an F-35. The bet is that the military needs cheap, expendable mass, not a handful of expensive jets. Barracuda follows the same logic. Anduril aims to build these cruise missiles for as little as $150,000 to $216,000 each.

The Roadrunner rounds out the group. It’s a reusable, jet-powered interceptor that can launch, chase down a threat, and land again to be used another day. One reported deal put it at around $250 million for more than 500 units, or roughly $500,000 apiece. 

These products only make financial sense if Anduril can build them by the thousand, which is the whole purpose of its giant new factory in Ohio.

Under the water, Anduril's flagship is Ghost Shark, a large autonomous submarine developed with the Royal Australian Navy. In September 2025, Australia signed a program of record worth about 1.7 billion Australian dollars, roughly $1.1 billion, over five years.

A dedicated Sydney factory opened the following month, and the first vehicle was delivered ahead of schedule. It is a rare case of Anduril winning a full national program rather than a prototype, showing the model works end-to-end.

Anduril is extending the same approach to undersea weapons with Copperhead, a family of high-speed autonomous vehicles that function as torpedo-class munitions. The undersea line is newer than the border business, but it is a clear example of Anduril selling an entire capability to an allied government.

In early 2025, Anduril took over one of the military's most troubled programs. Microsoft had spent years trying to build IVAS, an augmented-reality headset for the US Army. However, in 2025, the Army handed the effort to Anduril with a ceiling of up to $22 billion.

That $22 billion number gets quoted a lot, but it is an inherited ceiling, not money Anduril has earned or is guaranteed to build. When the program was recompeted later in 2025 as Soldier Borne Mission Command, Anduril's actual award was $159 million, with first scaled deliveries expected around 2027.

The gap between $22 billion and $159 million shows how to read Anduril's numbers: big ceilings show ambition, the smaller active awards show today's reality, and only one is revenue.

Lattice, the battlefield operating system, is sold as software: licenses and subscriptions plus continuous updates Anduril calls "phasing as a service." It does not disclose how much Lattice earns, so any figure is an estimate, but the strategic logic is clear.

Software carries far higher margins than steel, and it creates lock-in. Once a customer runs its drones, towers, and sensors through Lattice, replacing it means rebuilding how the whole force operates. That is the same recurring, sticky economics that makes commercial software so profitable.

Lattice also anchors Anduril's largest framework deal: a ten-year US Army enterprise contract awarded in March 2026 worth up to $20 billion, consolidating more than 120 separate buying actions around Anduril's open software. The deal’s size shows where the company wants to sit: underneath everything.

Anduril spends more than it earns, on purpose. The company posted an estimated operating loss of around $800 million in 2025 and expects to lose even more in 2026. 

Anduril pays to invent and manufacture products before it has contracts to sell them. The highest cost is not a factory or payroll. It is research, because the whole model depends on building first and selling later.

Anduril spends more than 60% of its revenue on research and development. That’s an unthinkable share at a traditional prime, where the figure is usually in the single digits. On 2025 revenue, that implies well over a billion dollars poured into designing products, much of it before a customer has committed a cent.

This is the cost that defines the company. The primes let the government fund their development and bill it back. Anduril fronts the money itself, betting a finished product wins contracts a slide deck never could. The trade-off is that every year of heavy self-funded research is a year of losses, which only pays off if the products sell in volume.

Anduril's largest physical bet is Arsenal-1, a "hyperscale" weapons factory in Ohio. It has committed at least $910 million, a figure the press rounds to $1 billion. The plant spans 5 million square feet and is expected to create more than 4,000 jobs by 2035, churning out tens of thousands of autonomous systems a year.

The state is helping foot the bill. Ohio agreed to a $310 million JobsOhio grant and a tax credit that could be worth more than $450 million over three decades.

Arsenal-1 is not the only site. Anduril also runs a $75 million solid rocket motor plant in Mississippi, a component the US badly lacks. It is building a roughly $1 billion campus in California. All three sites are funded with investor equity rather than debt, which is why the fundraising never stops.

Building all this takes people. Anduril has hired aggressively, employing roughly 8,000 after nearly doubling its workforce in 2025. The senior engineers it wants, in AI, robotics, and manufacturing, are expensive and in high demand.

Anduril competes for that talent with the biggest names in technology, including SpaceX and the large software firms, and pays accordingly. At Arsenal-1 alone, the average salary is projected to be around $132,000. Payroll rises every time the company takes on a new program.

Anduril also buys capabilities rather than building every one from scratch. It bought Dive Technologies for undersea vehicles, Adranos for rocket motors, Blue Force Technologies for the Fury airframe, Numerica for radar and command software, and more recently Klas and ExoAnalytic.

The prices are not disclosed, so the exact spend is unknown. The pattern, however, is deliberate. Anduril buys a specific technology, folds it into Lattice, and turns a niche product into part of an integrated system. It is growing by acquisition like a prime, only faster.

Anduril fights on two fronts at once. It’s trying to take work from the entrenched primes that have run the industry for decades, while racing a wave of well-funded defense-technology startups chasing the same money.

Anduril was built to challenge giants like RTX, Lockheed Martin, and Northrop Grumman. Each earns far more than Anduril. Lockheed Martin and RTX earned $75 billion and $88.6 billion, respectively, in 2025, and Northrop earned about $42 billion. Anduril’s ~$2.2 billion is a small fraction of any one of them.

The primes’ strength is entrenchment. They own the biggest platforms, and decades of relationships mean the Pentagon rarely buys a major weapon without them, on cost-covered contracts that guarantee steady, low-risk profit.

Their weakness is the flip side of that comfort. They are slow. They build to detailed government requirements rather than racing ahead of them, and their business is not built for churning out cheap systems by the thousand. In newer programs like the Golden Dome missile-defense effort, the primes and Anduril now compete for the same work.

Palantir is Anduril's closest rival in software, and at times its partner. It earned about $4.5 billion in revenue in 2025, growing more than 50% a year. Its Gotham platform competes with Lattice to be the software brain that militaries run on.

The two are not purely enemies. They have teamed up on TITAN, an Army sensor-and-command program. This shows how tangled the new defense-tech world is: a competitor on one deal can be a collaborator on the next.

They differ in starting point. Palantir began as pure software and is edging toward hardware. Anduril began with hardware and built its software outward. They are closing in on the same prize from opposite directions.

Shield AI is the most direct threat to Anduril's autonomy ambitions in the air. Its Hivemind software is designed to pilot aircraft and drones without human control or GPS. That’s the same broad problem Anduril is solving with Fury and Lattice.

The company is well-funded and growing fast. Investors valued it at $12.7 billion in early 2026, more than double a year earlier, and it has projected more than $540 million in revenue for 2026. That is smaller than Anduril, but squarely in the same fight.

The difference is focus. Shield AI concentrates on autonomy software, an AI pilot for existing aircraft, and partners for airframes, while Anduril builds both the aircraft and the brain. The market hasn't decided yet whether integration or specialization wins.

Saronic is Anduril's mirror image at sea. It builds autonomous naval vessels, chasing the same money Anduril pursues with Ghost Shark and Dive.

The company has raised enormous sums quickly. Its valuation reached $9.25 billion in early 2026 on the back of a $1.75 billion funding round, up from $4 billion the year before. That’s a sign of how hot investor appetite for defense hardware has become.

Saronic is narrower, focused on unmanned ships rather than a whole catalog across air, land, and sea. That focus could make it faster in its lane, but it lacks the cross-domain software platform central to Anduril's pitch.

Anduril is making two enormous bets on the next decade. The first: manufacture autonomous weapons at a scale no startup has attempted, turning Arsenal-1 into a factory that builds drones and aircraft by the thousand. The second: make Lattice the software layer the entire US military runs on, earning recurring, high-margin revenue across every branch and mission.

The growth opportunities are large. Anduril leads an industry team for Golden Dome, the proposed US missile-defense shield estimated at around $185 billion. It also sits in a pool of contracts worth up to $3.2 billion for space-based interceptors. Abroad, it is delivering Ghost Shark to Australia and pushing into Europe, where defense budgets are climbing.

An eventual public listing hangs over all of it. Luckey has said Anduril will "definitely" go public, but not before it’s profitable. Analysts expect an IPO could come in 2026 or 2027, though the fresh $5 billion raised in 2026 takes the pressure off.

The financial reality is where the tension lives. Anduril is valued at $61 billion, roughly 28 times its 2025 revenue, a multiple the established primes trade nowhere near. Yet it is projected to lose around $1.2 billion in 2026. Outside estimates do not expect sustained profitability until close to 2030. For that price to make sense, Anduril has to keep roughly doubling its revenue for years, which is why every factory, hire, and acquisition is aimed at capacity rather than near-term profit.

Three things that Anduril does not fully control will decide whether that confidence is justified. First, can it manufacture at the scale it has promised, not just targeted? Second, will US and allied defense budgets hold long enough to buy what it builds? Third, can it turn profitable before the defense-technology boom, which some investors already call a bubble, reprices companies like it? 

The first answers will come as Arsenal-1 ramps through 2027. Until then, Anduril is growing faster than any defense company in memory and losing more money than most, betting that its revenue will finally catch up with its spending.