Anduril: Breaking Defense’s Cost-Plus Curse
In the early 1990s, the Pentagon initiated what would become the most expensive weapons program in history. The F-35 Joint Strike Fighter was conceived as a cost-saving measure: single airframe serving three branches, leveraging economies of scale that only a consolidated defense industry could deliver. Three decades and $1.5 trillion later, the program stands as a monument to the cost-plus model’s structural pathology: contractors reimbursed for every hour spent and every cost overrun, with no mechanism to penalize delay or reward speed. The industry consolidated from fifty significant contractors to five, innovation slowed to a generational crawl, and the Pentagon learned to accept that new capabilities arrive not in months but in decades.
Then, in 2017, Palmer Luckey and Brian Schimpf founded Anduril Industries with a premise that sounded either naive or radical: that defense procurement could operate on firm fixed-price contracts, that companies could invest their own capital in research and development without government reimbursement, and that software-defined systems could iterate faster than the threats they countered. The bet was not merely technological but economic, hat the inversion of the Pentagon’s incentive structures would create a durable competitive advantage that legacy contractors could not replicate without destroying their own business models.
This is the economic engine: Anduril captures value not by billing the government for time spent, but by selling complete hardware-software systems at prices that would bankrupt cost-plus competitors, then compounding returns through software licensing and cross-domain expansion. The model works only if the company can maintain a software platform that amortizes development costs across drones, submarines, and autonomous systems while sustaining the capital discipline to absorb R&D risk that public contractors refuse to bear. It is a machine built for speed, but its continued operation depends on a fragile assumption: that the Pentagon’s recent tolerance for rapid iteration and vendor concentration will survive the first major catastrophe involving an autonomous system, and that Silicon Valley engineering culture can persist within the gravity well of classified defense work.
The Contract Is the Product
The defense industry’s traditional cost-plus structure creates a straightforward perversity: the slower a contractor works, the more revenue it generates. When the government reimburses development costs plus a fixed fee calculated as a percentage of those costs, the rational economic actor maximizes billable hours, not outcomes. Brian Schimpf, Anduril’s chief executive, identified this mechanism precisely when he observed that legacy contractors spend minimal percentages of their budgets on internal research and development because they can “build that as a cost” reimbursed by the government. The result is not laziness but structural inevitability, ublic defense companies return capital to shareholders through dividends and buybacks rather than funding speculative capabilities that might fail, because failure under cost-plus merely extends the billing cycle.
Anduril’s inversion depends on refusing this premise entirely. The company invests its own capital in R&D, developing integrated hardware-software systems, drones, interceptors, autonomous submarines, that it then sells to the Department of Defense on firm fixed-price contracts. If Anduril underestimates the engineering hours required to deliver a counter-drone system or autonomous undersea platform, the company absorbs the loss. If it delivers efficiently, it captures the margin. This transforms the contractor-customer relationship from one of monitored effort to one of evaluated capability. The Pentagon no longer pays for time; it pays for outcomes.
This mechanism creates a brutal selection pressure. Legacy contractors cannot easily adopt fixed-price models because their cost structures, built to support the overhead of cost-plus administration and shareholder distributions, make accurate pricing impossible without catastrophic risk. Anduril’s advantage lies in its cost structure: by building a unified software platform called Lattice that abstracts sensor fusion, computer vision, and autonomous decision-making across multiple hardware form factors, the company amortizes development costs across domains. The software written for a Ghost drone intercepting hostile UAVs migrates with modification to undersea platforms surveilling adversary submarines. Legacy competitors, by contrast, maintain entirely separate software stacks for each platform, Boeing’s various aircraft programs operate as technological islands, meaning they cannot achieve the unit economics necessary to profitably bid fixed-price contracts without risking bankruptcy.
Yet the fixed-price model carries its own fatal vulnerability: it requires accurate cost prediction in an environment of inherent uncertainty. Defense problems evolve. The counter-drone tactics employed by Iranian-backed forces in Syria today differ from those of tomorrow. If Anduril underbids to win a contract, a temptation given the aggressive pricing necessary to displace incumbent vendors, it faces the Boeing trap of compressed margins and forced cost-cutting that undermines the very speed advantage that justifies its existence. Overbid, and the Pentagon reverts to legacy players willing to absorb losses to maintain market position. The contract structure that enables Anduril’s speed also exposes it to the volatility that cost-plus was designed to eliminate.
Lattice as Capital Structure
Analysts typically describe Anduril as a defense contractor with superior software, but this misses the architectural significance of its platform. Lattice functions not merely as a product but as a capital structure: mechanism for deploying fixed R&D investment across variable revenue streams. By abstracting the common functions of autonomous systems (perception, decision-making, command-and-control) from the specific physics of individual platforms, Anduril transforms hardware from a bespoke cost center into a commodity instance of software capability.
This abstraction layer explains why Anduril can profitably manufacture hardware while legacy contractors cannot. When the company develops a new autonomous capability, say, swarm coordination for intercepting hypersonic missiles, the engineering cost is borne once but captured across multiple deployments: on aerial drones, undersea vehicles, and ground sensors. The marginal cost of extending Lattice to a new domain approaches zero, while legacy contractors must rebuild software architecture from scratch for each new platform, paying the full capital cost of capability development every time they bid a new program.
The durability of this advantage rests on a technical moat that is both deep and narrow. Rebuilding Lattice would require five or more years of concentrated R&D investment combined with expertise in both artificial intelligence and the arcane specifics of military systems integration, hermal management in desert environments, hardened communications in electronic warfare, certification for munitions safety. Legacy contractors possess the latter but not the former; commercial tech companies possess the former but lack the security clearances and institutional knowledge to navigate Pentagon procurement. Anduril occupies a narrow band where these competencies overlap, and its first-mover advantage creates network effects: as more Defense Department units adopt Lattice, the integration friction between different commands decreases, making the platform more valuable with each deployment.
However, this architecture creates a structural tension that threatens the model’s viability. To optimize Lattice’s performance, Anduril must control the hardware it runs on, pecifying sensor payloads, processing architectures, and communication protocols. This creates customer dependency. If the Pentagon perceives this as vendor lock-in, a historical anathema to military procurement officers trained to avoid single-source vulnerabilities, it may mandate open hardware standards or multi-vendor competition. Yet if Anduril opens its hardware specifications to satisfy these concerns, competitors can replicate the physical platforms while free-riding on the software development costs that Anduril alone bears. The company must navigate a narrow path: sufficient hardware openness to satisfy procurement regulations, sufficient control to prevent commoditization of the software layer that generates returns.
The Urgency Arbitrage
Anduril’s initial market entry exploited a gap between the speed of emerging threats and the institutional velocity of legacy defense procurement. The company’s first major beachhead came not from competing for established fighter jet contracts, but from solving urgent, unsolved problems: specifically, the threat of Iranian-backed drone swarms targeting special operations forces in the Middle East. Traditional procurement timelines, measured in years for requirements definition, then decades for development, were unacceptable when soldiers faced evolving drone tactics measured in weeks.
This beachhead strategy reveals the specific conditions under which Anduril’s economic engine functions. The model requires threat environments characterized by rapid evolution, where the Pentagon’s risk tolerance exceeds its institutional aversion to new vendors. Special operations commands, operating with delegated acquisition authority and immediate operational needs, provided the ideal initial customer: they possessed budget autonomy, technical sophistication, and organizational willingness to accept unproven systems if they delivered immediate capability.
The danger lies in the gap between these early adopters and the mainstream Defense Department procurement apparatus. Special operations units prioritize speed and adaptability; mainstream program offices prioritize stability, documentation, and compliance with acquisition regulations designed to prevent fraud and ensure interoperability with legacy systems. Success with the former does not guarantee acceptance by the latter. As Anduril attempts to scale from urgent tactical problems to strategic systems, long-range missiles, manned-unmanned teaming architectures, integrated air defense, it encounters procurement officers who operate on different timelines and risk profiles. Mainstream customers may demand cost-plus contract terms to ensure government visibility into development, or insist on hardware specifications that fragment the software platform, or require support infrastructures that replicate the bureaucratic overhead Anduril was founded to eliminate.
This tension represents the defining structural constraint on the company’s growth. If Anduril must rebuild its business model, its contract structures, its support organizations, its compliance procedures, for each new customer segment, it loses the economies of scale that justify the fixed-price risk. The company remains perpetually vulnerable to the moment when the Pentagon decides that the urgency has passed, that the threat has stabilized sufficiently to return to traditional procurement safeguards.
Private Ownership and the Cost of Capital
The economic viability of Anduril’s model depends on capital allocation decisions that public markets would not tolerate. Legacy defense contractors operate under shareholder pressure to maximize quarterly earnings and return excess cash through dividends and buybacks. This creates a structural disadvantage in emerging technology areas: public companies cannot justify sustained investment in unproven autonomous systems or artificial intelligence capabilities that may not generate revenue for years, particularly when they can bill the government for time spent on existing programs.
Anduril’s private ownership structure, with founders Palmer Luckey and Brian Schimpf deploying their own capital and reputational risk, functions as an economic variable that directly affects returns. Without quarterly earnings pressure, the company can fund speculative R&D in undersea autonomy or counter-drone interception that public competitors must abandon. This patient capital allows Anduril to absorb the fixed-price contract losses that inevitably accompany market entry, betting that scale and software amortization will eventually yield margins that cost-plus contractors cannot match.
Yet this advantage carries a fragility related to human capital rather than financial capital. Anduril’s competitive position depends on attracting Silicon Valley engineering talent, software architects, machine learning researchers, autonomy specialists, into defense work. Defense contracting imposes frictions that repel this demographic: security clearances require invasive background investigations and limit career mobility; compartmentalization prevents engineers from discussing their work publicly; classified environments lack the technical infrastructure (internet access, open-source tools) that commercial software development requires. If Anduril cannot hire and retain talent faster than attrition to commercial tech companies or burnout from security overhead, the iteration speed that justifies fixed-price contracts evaporates. The company’s economic engine stalls not from lack of capital, but from the inability to maintain the engineering velocity necessary to outpace legacy competitors.
The Assumption of Permanence
All of Anduril’s strategic advantages rest on a single fragile assumption: that the Department of Defense’s shift toward software-defined, rapidly-evolving systems represents a permanent institutional transformation rather than a temporary aberration. This assumption faces two distinct threats, either of which could collapse the economic model.
First, the regulatory threat. Military autonomous systems currently operate in a gray zone of certification and oversight. If a high-profile failure occurs: friendly fire incident involving an autonomous interceptor, a civilian casualty caused by algorithmic misidentification, or a system hijacking by adversary electronic warfare, ongressional oversight would likely impose strict testing, validation, and approval requirements on AI-driven systems. Such regulations would eliminate Anduril’s speed advantage by forcing the company to adopt the same bureaucratic development processes, extensive documentation, change control boards, formal verification, that slow legacy contractors. The fixed-price model becomes untenable when regulatory compliance consumes unpredictable engineering hours.
Second, the market maturity threat. Anduril’s value proposition assumes a continuously evolving threat environment that justifies ongoing software updates and platform evolution. If geopolitical conflicts stabilize or threat tactics become predictable, military customers will likely demand fixed, certified systems rather than continuously-evolving platforms. The Pentagon prefers vendor diversity and hardware standardization when urgency fades; it returns to multi-vendor procurement strategies that prevent single-source dependency. In such an environment, Anduril’s hardware-software integration becomes a liability rather than an advantage, as customers demand open systems compatible with multiple vendors rather than optimized performance locked to a single platform.
The bear case is straightforward: Anduril succeeds in urgent niches, scales toward mainstream adoption, encounters a catastrophic autonomous system failure or threat stabilization, and faces regulatory retrenchment that forces it to adopt the cost-plus, hardware-centric model it sought to displace. It becomes, in effect, a smaller, faster version of the legacy contractors, urdened by compliance overhead, unable to sustain Silicon Valley talent retention, and trapped in fixed-price contracts that destroy margins when development timelines extend beyond predictions.
The company’s survival depends on maintaining the narrow conditions of its birth: threat environments too dynamic for traditional procurement, institutional tolerance for vendor concentration, and engineering cultures willing to accept the frictions of classified defense work. These conditions are not historical constants but contingent arrangements. Anduril has built a formidable economic engine, but it runs on fuel that the Pentagon could decide to stop refining.
The Hardware-Software Binding Problem
To understand why Anduril cannot easily become the “Apple of defense”, a locked ecosystem of premium hardware running proprietary software, one must examine the specific physics of military procurement. The Pentagon’s institutional memory carries deep scar tissue from vendor lock-in. Maintenance of legacy systems, B-52 bombers running on 1960s avionics, nuclear command-and-control systems dependent on obsolete programming languages, has created a bureaucratic allergy to dependency on single vendors for critical capabilities. This institutional memory manifests in acquisition regulations that mandate open architectures, modular open systems approaches (MOSA), and competitive sourcing for major subsystems.
Anduril faces a binding constraint: to achieve the unit economics necessary for fixed-price profitability, it must maximize software reuse across hardware platforms, yet to achieve the performance metrics that win contracts against legacy systems, it must optimize software for specific hardware configurations. A generic autonomous navigation stack runs poorly on generic sensors; tight integration between Lattice and Anduril’s specific sensor payloads, compute modules, and communication systems delivers the performance advantages, lower latency, higher accuracy, reduced bandwidth, that justify selection over Boeing or Lockheed Martin.
This creates the abstraction trap. If Anduril over-abstracts Lattice to satisfy Pentagon demands for hardware-agnostic software, it becomes a commodity middleware provider competing with defense giants who can subsidize software development through cost-plus overhead on hardware manufacturing. The performance edge blurs; the economic advantage dissipates. If Anduril refuses abstraction, maintaining tight hardware-software binding to preserve performance advantages, it triggers procurement antibodies. Program offices at Air Force Material Command or Naval Air Systems Command, stewarding billion-dollar budgets and careers measured by program stability, view single-source dependency as an unacceptable risk. They demand technical data rights, source code escrow, and interface specifications that enable “second sources”, lternative vendors who could theoretically build compatible hardware.
Yet releasing interface specifications vitiates Anduril’s moat. Competitors could replicate the physical platforms, drones, submarines, interceptors, while avoiding the R&D costs of software development. The firm fixed-price model collapses if competitors can bid hardware at marginal cost while free-riding on Anduril’s software architecture. Anduril must therefore perform a delicate institutional negotiation: demonstrating sufficient “openness” to satisfy procurement regulations while maintaining sufficient opacity to prevent architectural theft.
The Ghost drone exemplifies this tension. Designed as a “modular flying platform” with APIs for payload integration, it presents an open face to customers who need to mount specific sensors or munitions. Yet the real-time power and performance monitoring systems, the specific implementations of computer vision on the edge-compute modules, and the low-latency mesh networking protocols remain proprietary. Anduril sells the Ghost at margins that assume recurring revenue from Lattice licensing and software updates, not merely hardware markup. If a competitor reverse-engineers the Ghost’s physical airframe and underbids on hardware while running generic autopilot software, Anduril faces a squeeze: match the hardware price and lose money on each unit, or maintain price and lose the contract.
This dynamic intensifies as Anduril scales from urgent tactical systems to strategic platforms. A special operations unit buying counter-drone systems measures success in immediate operational effect, id the system stop the threat yesterday? A program office managing a $10 billion autonomous wingman program for the Air Force measures success in program stability, contractor health, and congressional oversight survival. The former tolerates vendor concentration; the latter demands industrial base diversity. As Anduril attempts to migrate from the former customer to the latter, it encounters procurement officers who view its integrated model not as efficiency but as fragility.
The Estimation Problem
Fixed-price contracting transforms uncertainty from a billable asset into a balance-sheet liability. In cost-plus arrangements, uncertainty generates revenue: requirements that evolve, technical challenges that emerge, scope that creeps, ll extend the development timeline and increase reimbursable costs. In Anduril’s model, uncertainty destroys capital.
Consider the counter-drone mission. Iranian-backed forces in Syria and Yemen continuously evolve their tactics: switching from commercial quadcopters to fixed-wing munitions, employing swarm behaviors, utilizing frequency-hopping communications, deploying drone-dropped munitions rather than kamikaze attacks. Each evolution requires Anduril to update detection algorithms, modify interception kinematics, and potentially redesign hardware components. Under cost-plus, these changes generate engineering change proposals (ECPs), ormal mechanisms for increasing contract value. Under fixed-price, Anduril absorbs these costs unless it can negotiate contract modifications, a process that introduces the very bureaucratic friction the company promises to eliminate.
The estimation risk compounds across domains. When Anduril bid on autonomous undersea systems, submersible platforms for seabed warfare and surveillance, it transferred risks from the Navy to its own balance sheet. Undersea autonomy presents technical challenges distinct from aerial systems: acoustic rather than radio-frequency communications, pressure-hull engineering, energy density constraints for submerged endurance, navigation without GPS. If Anduril underestimated the complexity of sustained autonomous operation in contested undersea environments, the firm faces cost overruns that cannot be passed to the customer. A single major contract loss, absorbing, say, $200 million in unanticipated development costs to deliver a fixed-price submarine system, could consume the capital reserves necessary to fund the next generation of R&D.
This creates a selection bias toward conservative bidding that threatens the strategic beachhead. If Anduril prices contracts to ensure against technical uncertainty, its bids approach the cost levels of legacy contractors, eliminating the price advantage that justifies Pentagon tolerance for vendor risk. If it prices aggressively to win, to demonstrate the speed and cost differential that validates its existence, it accepts existential balance-sheet risk. The company walks a tightrope where a single misstep in technical estimation, a harder-than-anticipated computer vision problem, a certification requirement more stringent than anticipated, transforms from a manageable engineering challenge into a company-threatening loss.
The irony is that Anduril’s speed advantage depends on exactly the kind of aggressive estimation that makes such losses probable. To deliver systems in months rather than years, the company must compress design cycles, rely on rapid prototyping rather than exhaustive analysis, and accept technical debt that may surface unexpectedly during operational deployment. Silicon Valley engineering culture prioritizes minimum viable products and iterative refinement; defense procurement demands upfront certainty about performance specifications and failure modes. These cultures collide in the contract negotiation, where Anduril’s bid assumes engineering velocity that may not survive first contact with the Pentagon’s test and evaluation bureaucracy.
From Beachhead to Mainland
Anduril’s growth narrative assumes a smooth translation from special operations customers to mainstream Department of Defense adoption. This assumption ignores the fundamental structural discontinuity between how SOCOM (Special Operations Command) buys capabilities and how the military services acquire major weapon systems.
SOCOM operates under Title 10 authorities that allow rapid procurement of “urgent operational needs”: bureaucratic exception that compresses timelines from years to months. SOCOM program officers are evaluated on operational outcomes in active theaters; they possess technical sophistication and risk tolerance born of combat exigency. They accept 80% solutions delivered immediately over 100% solutions delivered after the conflict ends. This customer profile matches Anduril’s capabilities perfectly.
The Air Force, by contrast, acquires systems through Program Executive Offices (PEOs) that manage portfolios over decades. An F-35 or B-21 program involves thousands of subcontractors, congressional oversight committees, international partners, and certification regimes that span safety, interoperability, and nuclear surety. Program officers rotate every few years, creating institutional memory gaps filled by process documentation. Success is measured not by immediate fielding but by compliance with acquisition regulations and avoidance of audit findings.
When Anduril attempts to transition from SOCOM to, say, the Navy’s PEO for Unmanned and Small Combatants, it encounters customers who require different contract vehicles. These customers may demand cost-sharing arrangements where the government funds development in exchange for technical data rights. They may require ” Contractor Logistics Support” (CLS) arrangements that mandate decades of maintenance infrastructure incompatible with Anduril’s software-defined upgrade model. They may insist on “government purpose rights” to software source code, effectively nationalizing the intellectual property that constitutes Anduril’s primary asset.
Each mainstream procurement relationship requires rebuilding the business model. The fixed-price, product-sale model that works for urgent counter-drone needs may not satisfy the Navy’s preference for “power by the hour” service contracts or the Army’s “other transaction authority” agreements that morph into traditional procurement mid-stream. If Anduril must maintain parallel business processes, one for urgent tactical sales, another for strategic platform development, it sacrifices the operational leverage that justifies its valuation. The company becomes not a scalable platform but a collection of bespoke consulting engagements, each requiring custom compliance architectures and support infrastructures.
Moreover, mainstream customers demand “color of money” compliance: he intricate federal accounting rules that dictate whether funds may be spent on research, development, procurement, or maintenance. Anduril’s model blurs these categories: software updates delivered continuously constitute both development and sustainment. Traditional defense accounting systems cannot easily categorize Lattice licensing fees that enable hardware capabilities years after initial fielding. Anduril must either adopt the financial infrastructure of legacy contractors, hiring hundreds of compliance officers and accountants, or risk audit findings that disqualify it from major competitions.
The Talent Constraint as Economic Variable
The analysis of Anduril’s competitive position typically focuses on capital structure and contract mechanics, but the binding constraint may prove to be human capital. The company’s economic engine requires engineering velocity, rapid iteration, aggressive abstraction, continuous deployment, that depends on specific talent densities available only in Silicon Valley and similar technology clusters.
Defense contracting traditionally attracts engineers oriented toward systems engineering, requirements management, and process compliance, kills valuable for managing the complexity of aircraft carriers or satellite constellations. Anduril requires software engineers trained in machine learning, computer vision, and autonomous systems, skills valued at premium compensation levels by commercial technology companies. The wage compression between commercial tech and defense, historically a 30-50% gap, reates a recruitment headwind that Anduril must overcome through mission appeal and equity participation.
However, security clearances impose frictions that equity cannot fully resolve. The TS/SCI (Top Secret/Sensitive Compartmented Information) clearance process takes 12-24 months, during which candidates cannot work on classified projects and may accept commercial offers. Once cleared, engineers face “need to know” restrictions that prevent them from understanding the full context of their work, compartmentalization that isolates them from the collaborative development practices of commercial tech, and facility restrictions that prevent remote work or access to modern development tools. The clearance infrastructure effectively imposes a productivity tax on Anduril’s engineering workforce.
If attrition rates exceed recruitment capacity, a real risk given the tight labor market for AI/ML talent, the company’s iteration speed slows. Slower iteration transforms fixed-price contracts from calculated risks into probabilistic losses. The margin compression from delayed programs cascades through the financial model, reducing the capital available for speculative R&D and forcing the company toward the cost-plus behaviors it was founded to disrupt.
The culture clash extends beyond recruitment to retention. Silicon Valley engineering culture values public technical recognition, conference presentations, open-source contributions, academic publications, activities largely prohibited in classified defense work. Anduril must manufacture internal status systems and technical challenges sufficient to retain talent despite the inability to build public reputations. This requires investment in “skunk works” environments, aggressive compensation, and mission narrative that convinces engineers to accept career constraints. If these investments exceed the cost savings from software reuse, the economic advantage evaporates.
Second-Order Collapse Scenarios
The bear case for Anduril is not merely that growth slows, but that the specific conditions enabling its existence reverse simultaneously. Consider a scenario unfolding over eighteen months: a Ghost drone malfunctions during a demonstration, resulting in a high-visibility incident that injures friendly forces. Congressional hearings follow, revealing that the autonomous decision-making software lacked the formal verification standards applied to manned aircraft. The Pentagon responds with a new directive mandating exhaustive testing for all autonomous systems, effectively banning the rapid iteration model. Simultaneously, threat stabilization in the Middle East reduces the urgency driving SOCOM procurement, while a recession in commercial tech increases competition for engineering talent as laid-off software engineers return to defense contractors rather than startups.
In this scenario, Anduril faces simultaneous pressures: regulatory compliance costs that demand cost-plus accounting, reduced demand from early adopters, and talent attrition to legacy contractors now hiring aggressively. The company cannot pivot to cost-plus without destroying its equity value premised on margin expansion; it cannot maintain fixed-price without absorbing unsustainable compliance costs; it cannot retain talent without cash reserves depleted by contract overruns.
The structural fragility lies in the monoculture of the business model. Legacy contractors survive precisely because they maintain multiple business lines, cost-plus development, fixed-price production, services, maintenance, allowing cross-subsidization during market shifts. Anduril’s pure-play fixed-price, software-centric model offers no diversification. When the Pentagon’s appetite for vendor risk disappears, when software-defined warfare encounters its first catastrophic failure, the company lacks the diversified revenue streams to survive the transition.
Furthermore, the consolidation of the defense industry has created a political economy that resists disruption. Lockheed Martin, Raytheon, and Northrop Grumman maintain lobbying infrastructures and congressional relationships built over decades. They employ hundreds of thousands of workers in strategically important congressional districts. If Anduril threatens significant market share in major weapon systems, say, a $10 billion autonomous wingman program, these incumbents will activate political mechanisms to restrict competition. This may take the form of “buy American” provisions that advantage domestic manufacturing footprints Anduril lacks, or requirements for past performance on contracts Anduril has not yet performed, or security certification standards written to favor established processes.
Anduril’s survival depends on the Pentagon maintaining the political will to protect novel vendors against incumbent opposition: will that historically weakens after the first procurement scandal or operational failure. The company exists in a narrow window of institutional tolerance, a window that closes when the urgency of current conflicts fades or when the risks of autonomous systems manifest in visible catastrophe.
The Network Effects Paradox
Anduril’s strategy assumes positive network effects: as more Defense Department units adopt Lattice, the platform becomes more valuable through data network effects (improved algorithms trained on broader operational data) and integration network effects (reduced friction when joint forces operate common systems). However, military procurement generates negative network effects in vendor relationships.
When the Army adopts Lattice for counter-drone operations while the Navy adopts a competing platform for maritime autonomy, Anduril bears the cost of this fragmentation, aintaining multiple integration teams, customizing software for incompatible military networks, navigating different security architectures. If Lattice achieves true ubiquity across services, Anduril faces a different risk: the Pentagon designates the platform a “program of record,” subjecting it to the full weight of acquisition oversight, configuration control boards, and requirements creep that slowed the F-35.
The transition from experimental system to program of record transforms the customer relationship. Experimental systems operate on “other transaction authority” (OTA) agreements with minimal oversight. Programs of record operate under the Defense Federal Acquisition Regulation Supplement (DFARS), with earned value management systems, cost accounting standards, and congressional reporting requirements. The economic model that thrives under OTA suffocates under DFARS. Anduril must therefore grow large enough to achieve sustainable scale, yet remain small enough to avoid the regulatory gravity of major defense acquisition programs: structural paradox with no clear resolution.
The company’s trajectory suggests an attempt to thread this needle by maintaining multiple semi-autonomous product lines, Ghost, Dive-LD, Lattice, each small enough to avoid major program status while collectively achieving scale. Yet this fragmentation risks recreating the siloed software architectures of legacy contractors, eliminating the cross-domain amortization that enables fixed-price profitability.
Anduril has built a machine for winning contracts that legacy contractors cannot profitably bid, but the machine runs on assumptions about institutional tolerance, threat urgency, and talent retention that remain historically anomalous. The economics work only while the Pentagon accepts vendor concentration, while threats evolve faster than bureaucratic processes, while software engineers accept security clearances over stock options. When these conditions shift, and in defense procurement, they always shift, the fixed-price model becomes a trap rather than a weapon. The company’s ultimate test is not technical but temporal: whether it can establish institutional entrenchment before the window of opportunity closes, transforming from an urgent vendor into an indispensable infrastructure before the Pentagon remembers why it stopped buying this way.