Why Orchestration — Not Individual Platforms — Will Determine Who Wins the Next Decade of Defense
Over the past three years, warfare has crossed a structural threshold that most political systems, defense procurement agencies, and industrial planning frameworks have yet to internalize.
Autonomy is no longer a feature of modern weapons systems. It is becoming the operating logic of warfare itself.
What began as remotely piloted drones supporting surveillance missions in permissive environments has evolved into a rapidly expanding ecosystem of AI-enabled, networked, and increasingly independent systems capable of sensing, deciding, and acting across air, land, sea, space, and cyber domains, at machine speed. The Ukraine conflict has served as the first large-scale laboratory for integrating autonomous warfare, fundamentally altering NATO doctrine and European procurement priorities (IISS, The Military Balance 2026). The ongoing conflict in the Middle East, where Iranian-made drones number in the thousands and counter-UAS systems are tested daily, is accelerating the operational proof further. And the industrial response signals that this is not a temporary surge. It is a structural reordering: the U.S. Army awarded Anduril Industries a 10-year enterprise contract with a ceiling value of $20 billion (U.S. Army Contracting Command, March 2026); NATO allies formally committed to investing 5% of GDP annually in core defense and security-related spending by 2035 at The Hague Summit (NATO Summit Declaration, June 2025).
The consequence is structural. The center of gravity in military power is shifting from platforms to decision systems. From hardware to software. From individual system sophistication to network orchestration at scale.
The autonomous warfare systems market, valued at approximately $59 billion in 2025 and projected to reach $141 billion by 2035 across multiple estimates (Deloitte Global Defense Technology Outlook, 2024; McKinsey & Company Defense Practice, 2024; with growth rates ranging from 10–14% CAGR depending on segment and methodology), is not merely a growth story. It is a signal that global defense architectures are being reconfigured around software-defined, AI-enabled, and economically scalable combat systems. And the organizations that misread this signal, treating autonomy as a procurement category rather than a system architecture problem, will find themselves on the wrong side of the next decade.
Global Defense Spending Reaches $2.7 Trillion as Structural Forces Converge
Global military expenditure reached $2.718 trillion in 2024, a 9.4% increase in real terms from 2023, the steepest year-on-year rise since the end of the Cold War (SIPRI, April 2025). The top five spenders (the United States, China, Russia, Germany, and India) accounted for 60% of the global total. Three forces are converging to reshape this spending into something qualitatively different from previous defense cycles.
Geopolitical fragmentation has reintroduced high-intensity conflict scenarios into defense planning. NATO members collectively spent $1.506 trillion in 2024, representing 55% of global military expenditure (SIPRI, 2025). Eighteen of thirty-two allies now meet the 2% GDP threshold, up from eleven in 2023. The Hague Summit Declaration commits allies to 3.5% of GDP on core defense spending and up to 1.5% on defense-related infrastructure by 2035 (NATO, June 2025). The United States proposed a defense budget exceeding $1 trillion for FY2026, with $179 billion for RDT&E alone, a 27% year-over-year increase. EU NATO members spent €343 billion on defense in 2024 and plan to spend an additional €100 billion annually by 2027 (European Defence Agency, 2024). China’s reported defense budget reached $245 billion in 2025, though SIPRI estimates range considerably higher.
Capital and technology convergence is accelerating the transformation. Venture capital investments in defense technology reached $31 billion in 2024, up 33% year over year (McKinsey, Creating a Modernized Defense Technology Frontier, February 2025). This is not marginal capital. It is structural capital reshaping how defense capabilities are developed and fielded.
Inside the Five-Layer Autonomous Warfare Ecosystem
Autonomous warfare is not a single market. It is a system of five interdependent layers, and understanding where value concentrates, where margins sit, and where dependencies create fragility is the critical analytical task.
The Platform Layer is bifurcating into two economically distinct markets that should not be conflated. Expendable drones (@FPV systems priced between $500 and $5,000) are commoditizing rapidly. Ukraine has scaled domestic production into the hundreds of thousands annually (Ukrainian MoD; CSIS, 2025). This is a volume market with thin margins and low barriers to entry. But high-end platforms (Collaborative Combat Aircraft at $25 to $30 million per unit, classified payloads, EW systems, and precision sensors) remain oligopolistic and supply-constrained. The narrative that “hardware is commoditizing” is only half true. Low-end hardware is commoditizing. High-end hardware is becoming more scarce, more integrated, and more valuable.
The AI Decision and Software Layer is where the margin thesis concentrates, but the economics require scrutiny. Anduril reports gross margins of 40–45%, significantly above the 8–10% typical of traditional defense primes, driven by its commercial firm-fixed-price model rather than cost-plus contracting (industry analyst estimates; Acquinox Capital, January 2026). Anduril’s revenue approximately doubled to $1 billion in 2024 and is estimated at $2.1 billion for 2025, with internal projections targeting $4.3 billion in a near-term year, though management reportedly anticipates operating losses exceeding $1 billion during the hyper-growth phase (The Information; AviationOutlook investor deck reviews). Palantir provides the most transparent data point: FY2025 revenue reached $4.475 billion (up 56% year-over-year), with U.S. government revenue of $1.855 billion, GAAP operating margin of 41% in Q4 2025, and adjusted free cash flow margin of 51% (Palantir SEC filings, Q4 2025). Palantir’s Rule of 40 score of 127% is unprecedented at this scale, but its P/E ratio exceeding 200x means the market is pricing a decade of sustained defense software dominance that procurement-cycle volatility could disrupt.
The Network Layer contains a hidden constraint. The electromagnetic spectrum is finite and contested. In a peer conflict environment, spectrum congestion becomes the binding physical limit on how many autonomous platforms can coordinate simultaneously. Operational limits due to spectrum saturation are expected at scale, though exact thresholds remain scenario-dependent and largely classified (NATO NCIA, Operational Capacity Report, 2025; ITU spectrum allocation frameworks).
The Counter-System Layer is the fastest-growing segment and the most commercially overlooked. The counter-UAS market is projected to grow from $2.08 billion in 2025 to $19–20 billion by 2035 (PwC, 2025; MarketsandMarkets™ - Aerospace & Defense Research ). Fiber-optic-controlled drones and fully autonomous systems with no jammable control link are rendering conventional electronic countermeasures obsolete. Every offensive capability creates defensive demand, a self-reinforcing cycle that outlasts any single conflict.
Value Capture: How Margins Distribute Across the Autonomous Warfare Stack
Both investment committee critiques converged on one gap: the article described where value is migrating, but never quantified where the margin concentrates. Here is the structural picture.
Traditional defense primes (Lockheed Martin, RTX, Northrop Grumman, BAE Systems, Inc.) operate at 10–12% EBITDA margins on platform hardware under cost-plus and firm-fixed-price contracting. Their competitive moat is certification authority, security clearances, and decades of procurement relationships. They are not losing control today, but they are losing deal flow at the margin.
The neoprime software layer (Palantir, Anduril Lattice, Shield AI Hivemind) targets a structurally different margin profile. Palantir’s adjusted operating margin of 57% (Q4 2025, SEC filing) demonstrates that defense software can achieve commercial-grade margins when sold as a platform rather than a bespoke integration. Anduril’s 40–45% gross margin reflects its commercial sales model, but the company remains unprofitable during its scaling phase, with projected operating losses exceeding $1 billion. Shield AI reported that Hivemind software accounted for approximately 30% of revenue in FY2025 (ending March 2025), targeting 50% by 2028, but 70% of current revenue remains hardware.
The sustainment and lifecycle layer (AI model retraining, data pipeline maintenance, software validation, spectrum management) is where the OPEX inversion matters most. Consider an illustrative lifecycle cost comparison for a 1,000-unit expendable drone fleet over a five-year operational cycle: procurement cost is approximately $3–5 million (at $3,000–$5,000 per unit). But five-year sustainment (AI training data refresh, continuous software updates, spectrum allocation and management, battery logistics, maintenance labor for hardware-software integration) adds an estimated $4–7 million in operational cost (derived from U.S. GAO sustainment data, McKinsey computing gap analysis, and first-principles modeling of data pipeline and spectrum management requirements). Hardware accounts for roughly 40% of the total lifecycle value. OPEX and software consume 60%. This ratio inverts the cost narrative that frames autonomous systems as inherently cheaper than traditional platforms. At a sustained conflict scale, with fleet sizes in the tens of thousands, these costs scale super-linearly.
The implication for capital allocators: the highest EBITDA pools are concentrating in sustainment, software licensing, and integration services, not in platform hardware. The investment analogy is less “buying the next Lockheed” and more “the TransDigm model applied to autonomous warfare aftermarket,” capturing recurring, high-margin, mission-critical spend on systems already fielded.
Europe as a Distinct Market: Fragmentation, Sovereignty, and Untapped Scale
Europe’s autonomous warfare challenge is qualitatively different from the U.S. model, creating a specific commercial opportunity.
North America commands 42% of the autonomous systems market (PwC, 2025). Europe holds 28% but faces three structural constraints that the U.S. does not. First, procurement fragmentation: the European Defence Fund, EDIRPA, and national procurement agencies operate in parallel, with limited interoperability between programs. Second, regulatory drag: over 60% of defense leaders surveyed cite ethical and regulatory barriers as the top AI adoption hurdle (RAND, 2025), and European time-to-field for autonomous systems runs significantly longer than in the United States. Third, industrial base competition: the commercial ramp-up at Airbus and Germany's defense rearmament (planned 8,300 military drone systems by 2029) compete for the same industrial base: engineering talent, precision manufacturing capacity, and supply chain access.
But Europe also has distinct assets. Helsing, valued at approximately €5.5 billion after its €450 million raise, is positioning itself as the European orchestration layer. FCAS/SCAF, despite its well-documented Franco-German governance challenges, represents the largest autonomous combat air program outside the United States. The European Defence Fund has allocated specific AI and autonomy budget lines. And the Hague Summit’s 5% GDP commitment creates a fiscal forcing function that will drive hundreds of billions in European procurement over the next decade. Procurement that European industrial policy will fight to keep sovereign.
For investors and industrial leaders, the European autonomous warfare market is not a smaller version of the U.S. market. It is a structurally different market with its own chokepoints, consolidation opportunities, and policy-driven demand curves.
Five Structural Constraints Forcing the Transition
The acceleration of autonomous warfare is not driven by technology enthusiasm. It is driven by five structural constraints.
AI integration at scale: 7.1% of global defense budgets in 2025 are directed toward autonomous and AI capabilities, rising toward 11% by 2030 (IMF/World Bank, 2024). The defense AI systems integration market is growing at 14.3% CAGR (EY, 2024).
Cost asymmetry: A $500 FPV drone can neutralize a $10 million armored vehicle. But the SaaS-in-defense model remains unproven at portfolio scale. Defense software margins are structurally capped by procurement frameworks, classification requirements, and the absence of true self-service adoption cycles that drive commercial SaaS economics.
The constraint triad: Defense-grade AI chips face a 23% supply-demand gap (Reuters, 2024). Secure satellite bandwidth faces an 18% shortfall (NATO NCIA, 2025). UAV maintenance incurs a 14% turnaround delay due to software validation (GAO, 2025). These are not technology problems. They are infrastructure problems.
The human-on-the-loop transition: HITL systems hold 42% market share but are declining. HOTL systems are the fastest-growing segment, with a 11.88% CAGR. Command structures must adapt.
Private capital at scale: Anduril at $30.5 billion (Series G, June 2025), reportedly approaching $60 billion. Shield AI is at $5.6 billion, reportedly approaching $12 billion. Saronic at $4 billion. Helsing at ~€5.5 billion. Combined VC/PE in dual-use defense startups exceeded $9 billion in 2024. This capital demands software-scale returns from hardware-constrained markets. The valuation correction timeline matters: at 30x revenue for Anduril and 200x+ P/E for Palantir, any disruption in the procurement pipeline (a continuing resolution, a political cycle shift, a regulatory shock) compresses multiples violently. This is a 2027–2029 risk, not a 2035 risk.
A Readiness Framework for Autonomous Warfare Positioning
For leaders navigating this transformation, these five filters evaluate competitive positioning:
System Integration Capability. Can your organization integrate platforms, AI decision engines, and network infrastructure into a coherent operational system? The Army’s decision to consolidate Anduril’s 120 contracts into a single enterprise vehicle signals a shift toward system-level procurement.
Data Advantage. No training data exists for peer conflict with active adversarial adaptation. Organizations that invest in simulation environments and hardware-in-the-loop testing gain a compounding advantage.
Economic Scalability. The bifurcation between expendable mass and exquisite autonomy requires clarity on which cost structure you are competing in.
Regulatory Positioning. The EU AI Act largely excludes military applications, but interpretation is evolving. 156 states supported a UN resolution on the governance of autonomous weapons (UN ODA, November 2025).
Time-to-Field. Organizations that compress the cycle from AI simulation to supervised deployment, as Shield AI’s three-phase Hivemind pipeline demonstrates, hold a decisive operational advantage.
Signals That Have Not Yet Reached Mainstream Industry Media
The sharpest operators detect signals before they become consensus. Four patterns are emerging that have not yet reached mainstream industry media.
The “Dark Swarm” doctrine is replacing cloud-assisted operations. Military doctrine is shifting toward heavy edge-compute platforms designed to operate completely offline in heavily jammed environments. Shield AI’s Hivemind operated without GPS or communications in Ukraine, defeating both Ukrainian and Russian EW jamming across 130+ sorties (Fortune, December 2025). Russia’s shift to fiber-optic-guided Lancet drones, immune to RF jamming, confirms that the future battlefield assumes communications denial as the default condition, not the exception. The commercial implication: edge AI compute hardware becomes the critical procurement category, not cloud infrastructure.
“Algorithmic sovereignty” is replacing hardware sovereignty. Governments are demanding indigenous AI models and rejecting black-box neural networks from partners, driven by concerns over allied kill-switch dependencies. This pattern is consistent across procurement conversations in Europe, the Middle East, and Asia-Pacific, though it has not yet been formalized in treaty language. If this accelerates, it fragments the interoperability assumption underpinning NATO’s autonomous warfare doctrine and creates protected national markets for European, Gulf, and Asian AI companies that can deliver sovereign autonomy stacks.
Talent migration from Big Tech to defense AI is creating a clearance bottleneck that fundamentally alters unit economics. Cleared AI engineers command 2–3x commercial market rates. The security clearance pipeline takes 6–18 months. Defense tech companies burn cash on cleared labor at rates that compress margins and extend paths to profitability. This is not an operational detail. It is a structural cost driver that VC valuation models underweight. The approximately 2,000 AI-relevant PhDs the United States produces annually cannot fill the demand; the defense sector cannot compete with commercial compensation for uncleared talent; and the clearance process excludes a significant portion of the talent pool. This constraint is generational, not cyclical.
Space-based autonomous platforms represent the fastest-growing addressable segment. Autonomous space systems are projected to grow at 13.56% CAGR through 2035, tying orbital ISR to autonomous kinetic effects. Shield AI’s December 2025 partnership with Sedaro extends Hivemind autonomy to satellite operations, targeting multi-agent teaming for proximity operations and defensive counter-space in communications-limited environments.
Three Scenarios for the Next Decade and Where Capital Should Position
The autonomous warfare ecosystem could land in fundamentally different places over the next decade. Each demands different capital positioning.
Network Dominance: The Orchestration Problem Gets Solved
Interoperability standards mature. AI-enabled command and control architectures achieve cross-service and cross-allied integration. Software orchestrators (Anduril Lattice, Palantir AIP, Shield AI Hivemind) dominate the ecosystem. Traditional primes become platform suppliers within larger digital architectures. Value is concentrated in the software and data layers. Defense procurement transitions fully to enterprise software models.
Drivers: Successful CJADC2 implementation. NATO DIANA investments are reaching operational maturity. Quantum-assured navigation solving GPS-denial. Political will to share algorithms across allies.
Consequence: Software engineering becomes a critical determinant of military superiority, alongside industrial capacity and logistics.
Capital positioning: Orchestration-layer companies command premium multiples. Platform pure-plays compress toward commodity margins. The investment analogy is the cloud transition, in which infrastructure providers captured more value than the enterprises running on their platforms.
Fragmented Autonomy: Siloed Systems in a Contested Landscape (most probable)
Autonomous systems are widely deployed, but interoperability remains structurally constrained by institutional incentives, classification barriers, and vendor lock-in. CJADC2 delivers information sharing but not autonomous coordination. The commons tragedy in standards persists. Electronic warfare and adversarial AI regularly degrade autonomous performance, forcing reversion to human-controlled operations.
One critical caveat: a wartime forcing function, whether a Taiwan scenario or an escalation in the Middle East beyond current operations, could compress interoperability timelines from fifteen years to eighteen months. History demonstrates that institutional incentives collapse under existential urgency. The article’s interoperability critique holds in peacetime conditions. Under wartime conditions, the calculus changes fundamentally.
Drivers: Common tragedy in standards. Proprietary differentiation. Classification barriers. Spectrum congestion at scale.
Consequence: Autonomous systems provide tactical advantage but not the transformative superiority the investment thesis assumes. Industry grows at the lower end of forecast ranges (10% CAGR, not 14%).
Capital positioning: Picks-and-shovels plays (compute, spectrum, sensors) and counter-UAS deliver more reliable returns than orchestration bets. The TransDigm Group Inc./HEICO analogy applies: capturing high-margin aftermarket spend on fielded systems, rather than betting on platform disruption.
Constraint Collapse: Physics and Governance Break the Scaling Narrative
Spectrum congestion prevents large-scale autonomous coordination. AI systems trained on non-adversarial data degrade significantly when faced with peer opposition. A major autonomous targeting error triggers regulatory shock and binding international restrictions. Counter-UAS cost exchange ratios reach favorable levels. Drone hardware valuations collapse.
Drivers: Electromagnetic spectrum physics. Adversarial AI adaptation. A high-profile civilian casualty event. Binding governance from the UN CCW and the General Assembly mandates.
Consequence: The “autonomous warfare revolution” narrative deflates. Organizations that overinvest in offensive drone platforms face stranded assets.
Capital positioning: Defensive technologies, quantum-PNT, and edge-AI plays retain value. Counter-autonomy and resilience investments become the highest-returning segment. This is the scenario in which traditional primes, with their certification authorities, government relationships, and sustainment revenue, reassert dominance.
The Blindspot: Why the Platform Superiority Thesis Is Incomplete
The industry assumption: The nation or company that builds the most sophisticated autonomous platform wins.
Why is this dangerously incomplete?
Three realities undermine this framing. First, there is limited real-world testing against peer-level adversarial AI systems. Performance metrics from SIPRI targeting studies and NATO ISR trials derive from controlled environments without adversarial adaptation. These are upper bounds, not operational guarantees (RAND; DARPA testing gap assessments).
Second, the cost advantage inverts at scale when OPEX is properly accounted for. The lifecycle illustration above, where hardware accounts for only 40% of total cost, is not yet reflected in market projections that anchor on unit procurement prices.
Third, the network's interoperability thesis depends on a tragedy of the commons. Defense contractors compete on proprietary differentiation. Military services compete for budget by demonstrating unique capabilities. No mechanism exists to align individual program incentives with joint force benefits.
The bottom line: The winners will not be the companies with the most advanced prototypes. They will be the organizations that solve the orchestration problem while respecting the physical constraints (spectrum, compute, power, governance) that most narratives ignore.
Executive Decisions: Orchestration, Compute, and Counter-Autonomy
Prioritize Orchestration Infrastructure Over Platform Procurement. The Anduril ($20B ceiling, March 2026) and Palantir ($10B ceiling, July 2025) enterprise contracts are contract vehicles with maximum potential values, not obligated spend. But they represent a procurement architecture that favors software-defined integration. Evaluate whether your organization is investing in the integration layer or is still buying individual systems that may become stranded assets. Note the fiscal risk: these vehicles depend on continued appropriations. A continuing resolution or defense topline flattening in FY2028–2029 could freeze new obligations and compress neoprime revenue projections.
Treat Compute, Data, and Spectrum as Capital Investments, Not IT Procurement. McKinsey identifies a $160–230 billion “computing gap” in the U.S. alone (McKinsey, Future Defense Tech, February 2026). The supply chain behind this gap (defense-grade AI accelerators manufactured through TSMC, Intel Foundry Services, and CHIPS Act domestic fabs) represents both a chokepoint and an investment opportunity. Leaders should treat compute infrastructure, data sovereignty, and spectrum management as core capital allocation, not IT procurement.
Invest in Counter-Autonomy with the Same Urgency as Offensive Capability. The counter-UAS market, growing from $2 billion to $20 billion by 2035, is not secondary. It is the defensive complement to every offensive capability deployed. Organizations investing only in offensive autonomy are building one-sided portfolios.
As Marcus Aurelius observed, the obstacle is the way. In autonomous warfare, the obstacle is not technology. It is the capacity to integrate, orchestrate, and govern systems of extraordinary complexity while respecting the physical and ethical constraints that define responsible power.
When the next peer conflict arrives, and the autonomous networks are tested not against simulations but against adversarial intelligence operating at machine speed, will military superiority belong to the side with the most advanced individual platforms, or to the side that built the architecture to orchestrate thousands of systems as one?
Andy Demir provides board-level analysis on aerospace, defense, space, and advanced industrial markets.
For board-level conversations on strategy, partnerships, and expansion, connect via LinkedIn.
Related Reading