← Companies | defense

Anduril

Anduril Is Trying to Be the Operating System of American War

| defense
↓ .md Take this into your AI — the full analysis + graph as markdown, ready to paste into ChatGPT, Claude, Gemini or any AI.

Based on 64 related nodes across 3 research explorations in the defense sector.

What Anduril Actually Does

Most defense companies make weapons. Anduril makes the software that tells weapons what to do — and then builds some of the weapons too.

Think of it this way: your phone has a hardware layer (the chips, screen, and battery) and a software layer (the operating system that makes everything talk to each other). For decades, the US military’s “software layer” — the systems that connect drones, sensors, radar, missiles, and command centers — has been old, slow, and built by the same handful of giant contractors who have been doing it since the Cold War.

Anduril’s bet is that it can replace that software layer with something modern. Its product is called Lattice OS, and the goal is for Lattice to become the operating system that all American autonomous weapons run on — the way iOS runs iPhones or Windows runs PCs.

If that works, Anduril doesn’t have to win every weapons contract. It just has to be the layer everything else plugs into.

Why This Position Is Powerful

The most striking thing about Anduril’s position isn’t any single product. It’s that Lattice OS has already been embedded in multiple large, independent programs at the same time.

The US Army signed a contract worth up to $20 billion (the NGC2 contract) to use Lattice OS as its command software — consolidating more than 130 older contracts under one platform. The US missile defense program called Golden Dome, at an estimated $185 billion, uses Lattice OS as its core software. The AUKUS alliance (the US, UK, and Australia) is deploying an Anduril-built autonomous submarine called Ghost Shark, also running on Lattice, under a contract worth roughly $1.1 billion US dollars.

Three major independent programs, spanning land, air, and sea, all depending on the same software layer. This is how platform lock-in works: once you build your house on a foundation, you don’t tear up the foundation to lay a different one. Switching costs become prohibitive.

The comparison to a technology platform is not just an analogy — Anduril’s own research notes explicitly tag the company under a concept called “Neoprimes Enterprise Platform Lock-in,” referencing multi-decade contracts with $20 billion ceilings that structurally replicate the same sticky position that Lockheed Martin and Raytheon built over fifty years — just faster and with modern software economics.

The Political Network Factor

Anduril was co-founded by Palmer Luckey, who previously built the Oculus VR headset and sold it to Facebook. The company has backing and active support from Peter Thiel and Marc Andreessen, two of the most influential figures in Silicon Valley and, since 2024, in the Trump administration’s orbit.

This matters because defense procurement is not a pure technology meritocracy. Relationships, political alignment, and access shape which companies get contracts, how quickly, and under what terms. The research data captures this as a distinct structural advantage: the Thiel-Luckey-Andreessen network actively amplifies the entire “new defense tech” category, of which Anduril is the flagship.

The non-obvious part: this advantage is explicitly regime-dependent. It does not survive a hostile administration in its current form. The political network is both a real competitive advantage and a fragility — it is the same thing.

Strengths Worth Understanding

Doctrinal alignment that predates the current administration. The US military has been developing a war-fighting concept called Mosaic Warfare since roughly 2017 — the idea that future conflicts will be won by swarms of cheap, networked autonomous systems that can reconfigure faster than an opponent can respond, rather than by a small number of expensive, exquisite platforms. This doctrine was formalized before Trump, and it will likely survive after him. Lattice OS is structurally what Mosaic Warfare requires: a software layer that can orchestrate hundreds of autonomous systems across domains simultaneously. The doctrinal tailwind is bipartisan.

The Arsenal-1 factory. Anduril is building a 700,000 square foot manufacturing facility in Ohio designed to produce tens of thousands of autonomous systems per year. The strategic logic is that the US currently cannot build drones and autonomous weapons at scale — China can. Arsenal-1 is the attempt to close that gap domestically. It is real, it is under construction, and it directly addresses one of the military’s most urgent concerns.

International template. The Ghost Shark contract with Australia demonstrates that the Lattice OS model can be exported to close allies under existing legal frameworks. This matters because it means Anduril’s revenue and strategic position are not entirely dependent on US political cycles.

Vulnerabilities Worth Understanding

The Taiwan chip problem. Every sophisticated autonomous system Anduril builds — Lattice OS, the drone systems, the missile defense software — runs on advanced chips made almost exclusively at factories in Taiwan. The entire reason the US military is accelerating autonomous weapons development is fear of a potential conflict with China over Taiwan. But a conflict over Taiwan is also the scenario that would destroy the chip supply chain Anduril depends on to build its weapons. This is not a theoretical risk. It is a structural contradiction baked into the business model: the event that creates maximum demand is the same event that creates maximum supply disruption.

China still makes the parts. Arsenal-1 domesticates the assembly of drones and autonomous systems. It does not domesticate the components. Roughly 85-90% of the rare-earth magnets used in drone motors come from China. Much of the battery chemistry comes from China. Many electronics components come from China. Building the factory domestically while still importing the critical inputs shifts some risk but does not eliminate the dependency. Arsenal-1 is a necessary step, not a solution.

Europe is building a wall. The European Union is in the middle of the largest defense spending increase in its history. It is also, deliberately and structurally, spending that money in ways that exclude American companies. The EU’s SAFE procurement framework essentially requires European defense money to go to European companies. The EU is actively funding a European competitor to Anduril called Helsing, a German AI defense company. This is not a preference for a different vendor. It is the construction of a procurement barrier. Anduril’s success in Australia does not transfer to France, Germany, or Poland.

Program execution risk. Before DAWG (the current $54.6 billion autonomous drone procurement program), there was a predecessor called Replicator. Replicator’s goals were to field “thousands” of autonomous drones quickly. It delivered “hundreds,” and they were reportedly unreliable in contested environments. The defense startup ecosystem has a history of promising more than it delivers. If DAWG faces similar problems, the entire procurement narrative that supports Anduril’s enterprise contracts faces political and budgetary pressure.

The Non-Obvious Finding

The most structurally interesting thing in the data is not a product or a contract. It is a relationship.

Anduril and Palantir — the two most prominent technology companies in US defense AI — are simultaneously partners and potential competitors. They are both inside the NGC2 Army contract and both inside the Golden Dome missile defense program. But they occupy different layers: Anduril owns the command-and-control layer (what orchestrates the weapons), and Palantir owns the intelligence-and-targeting layer (what decides where to aim them).

This co-deployment in the same programs looks like collaboration. But it also means each company has intimate knowledge of the other’s architecture. The question the data does not answer — and which may determine both companies’ long-term positions — is whether the next major program replicates this partnership or forces one layer to absorb the other.

Bull Case

The case for Anduril’s future, in plain terms: it has already done the hardest part.

Platform lock-in in defense procurement is extraordinarily difficult to achieve. The legacy companies — Lockheed, Raytheon, Northrop Grumman — took fifty years to build it. Anduril has established simultaneous platform dependencies across three major independent programs in roughly a decade. The Army contract, the Golden Dome contract, and the AUKUS contract are not the same customer saying yes three times. They are different programs, different services, and different countries all building on the same foundation.

The doctrinal demand for what Anduril makes — networked, software-defined autonomous systems — was established before Anduril was founded and reflects a genuine military consensus. It will survive political transitions.

If Arsenal-1 reaches production scale before a Taiwan conflict materializes, and if Lattice OS expands beyond the Army into Air Force and Navy command infrastructure, Anduril becomes infrastructure for American warfare in the same way that Amazon Web Services became infrastructure for American commerce. You don’t switch off AWS because it’s too embedded to replace. Lattice OS is being built to achieve the same irreplaceability.

Bear Case

The case against Anduril’s future, in plain terms: the company’s position depends on conditions that may not hold.

The political network that accelerated Anduril’s procurement access does not survive a hostile administration. The chip supply chain that every Anduril product depends on could be severed by the exact geopolitical event the products are designed to address. The component supply chain for drone manufacturing is still overwhelmingly Chinese, regardless of where assembly happens.

The largest defense market outside the US — Europe — is actively being closed off by procurement law. The predecessor to the current drone procurement program failed badly. And the company’s most capable current partner, Scale AI, has the capability and competitive incentive to eventually become a platform competitor rather than an infrastructure provider.

The negative scenario is not implausible: supply chain disruptions, program execution failures, a 2028 election reversal, and European market exclusion could collectively confine Anduril to a smaller domestic market under contract renegotiation pressure, without the growth trajectory needed to justify its current valuation.

Bottom Line

Anduril has achieved something genuinely unusual in defense technology: it has embedded a single software platform across multiple large, independent programs in a short time. That is real structural leverage, and it compounds.

The vulnerabilities are equally real and are not fully within Anduril’s control. The chip dependency problem has no near-term solution. The European market is structurally closed. The political tailwind is regime-dependent.

What separates the bull and bear cases is largely timing. If Anduril can deepen its platform lock-in, scale Arsenal-1, and replicate the AUKUS model before the supply chain and political risks materialize, the position becomes durable. If those risks arrive first, the platform moat that looks so strong today is sitting on a more fragile foundation than the contracts suggest.

The company is best understood not as a defense contractor but as a platform infrastructure play that happens to operate in the most politically and geophysically volatile market on earth.