From Drone Armies to Systems States: The Political Consequences of Adaptive Coercive Warfare

Updated June 2026: Practitioner testimony from Ukraine’s defence sector refines the paper’s Ukraine finding. Adaptation speed is a vector across five sequential clocks — Ukraine is fast on sensing and engineering but uneven on validation, procurement, and force-wide diffusion. A fifth dimension of adaptive system quality — institutional absorption capacity — is also identified. See the June 2026 research update.

Executive Summary

The central systems-engineering insight of this paper is that the relevant system boundary in modern drone warfare extends well beyond the airframe. What the Department of Defense terms digitally connecting people, processes, data, and capabilities [1][2] is, in the warfighting domain, the mechanism by which battlefield observation becomes design requirement, design requirement becomes fielded capability, and operational feedback becomes doctrine. Durable coercive advantage is therefore an emergent property of the ecosystem as a whole, not an attribute of any single platform. Recent analyses of Ukraine demonstrate that autonomous systems, information operations, electronic warfare, contested logistics, and air defence now operate as one tightly coupled warfighting ecosystem rather than as separate capability silos. [3][4]

This evidence does not support a simple belief that drones automatically dissolve state power. At the tactical edge, they do make observation and attack cheaper, denser, and more widely distributed. At the strategic level, however, effective drone warfare pushes actors toward tighter integration of procurement, standards, software, spectrum management, training, and industrial replenishment — precisely the disciplines that systems engineering addresses at the lifecycle level. Recent evidence is striking: Ukraine plans to purchase roughly 4.5 million FPV drones in 2025 [estimate pending verification] after procuring more than 1.5 million in 2024 [estimate pending verification], with 96 per cent sourced domestically [estimate pending verification; single journalistic source]; Russia is pursuing a full-cycle sovereign UAS ecosystem with 48 research and production centres [estimate pending verification] and a goal of 70 per cent domestic market capture [estimate pending verification]; [5][6] and Western and Ukrainian drone ecosystems remain exposed to component and engine bottlenecks, including continuing dependence on Chinese-origin inputs in critical supply chains. [7][8][9]

The most useful political question is therefore not “who has drones,” but “who can sustain a living digital thread of coercion.” In SE terms, that means asking who can maintain end-to-end traceability from battlefield observation to fielded design change, and who can govern the interfaces between sensing, command, industry, and doctrine with sufficient coherence to learn faster than an adversary. The main implication is that future military advantage will depend less on owning a particular platform than on governing an adaptive ecosystem. States that can institutionalise learning, secure key inputs, standardise data flows, and build affordable layered defences will strengthen state capacity. States and non-state actors that cannot will either remain tactically dangerous but strategically brittle, or will become dependent on outside patrons, suppliers, and sanction-evasion networks. [10][11][12][13]

Introduction

The strongest version of the “drone revolution” thesis is too linear. It assumes that because drones are cheap, available, and lethal, they must necessarily decentralise violence and erode the state’s monopoly over organised force. The recent record is more complicated. In Ukraine, mass drone warfare has been inseparable from state procurement, standards, and increasingly institutionalised data integration. In the Red Sea, the Houthi movement has shown that low-cost drones and missiles can impose large economic costs without possessing full-spectrum state capacity. In Mexico, cartel use of cheap commercial drones has lowered the cost of surveillance and intimidation without producing military parity with the state. In Sudan, drone strikes on power and transport infrastructure have accelerated fragmentation rather than deliver any stable coercive order. [5][22][23][24][25]

What links these environments is not the airframe alone. It is the system in which the airframe sits: sensors, command-and-control services, battle-management software, electronic warfare, training pipelines, component supply, repair capacity, legal authorities, and political oversight. This is precisely the kind of system that systems engineering is designed to analyse: a bounded set of interacting components whose emergent behaviour cannot be predicted from any single part, operating within and shaped by an external environment. The core analytic question is therefore not whether drones matter — they plainly do — but whether an actor can translate observation into requirements, requirements into production and fielding, and battlefield experience back into design and doctrine faster than an adversary can adapt. [1][17][18][19][4]

This matters politically because the location of coercive power is shifting. Precision strike is moving downward toward smaller units and cheaper tools, but the infrastructure required to sustain precision at scale is moving upward into software, data governance, industrial policy, export controls, spectrum access, alliance management, and civil-military innovation networks. A state can possess drones without possessing what this paper calls infrastructural sovereignty: the capacity to control or reliably secure the critical enabling layers behind the drone. Current supply-chain evidence from Ukraine and Russia, and regulatory analysis on AI and UAS export controls, suggest that these enabling layers are now strategic terrain in their own right. [7][8][9][12]

This paper contributes to the systems-engineering literature on system-of-systems (SoS) governance and digital thread implementation by applying those concepts to the political consequences of adaptive coercive warfare. The conceptual framework, comparative cases, and recommendations are written for systems engineers, defence engineers, and policy practitioners engaged in defence-industrial and lifecycle governance. The substantive claim is consistent across those readerships: the future political consequences of drone warfare will be determined less by diffusion of hardware than by variation in adaptive system quality.

Literature Review

Recent strategic literature increasingly treats modern war as a coupled system rather than as a set of independent capability areas. Work from the Center for Strategic and International Studies (CSIS) on Ukraine identifies five mutually reinforcing domains that now define effective warfighting: autonomous systems, information operations, electronic warfare, contested logistics, and evolving air defence. The Royal United Services Institute (RUSI) similarly argues that electronic warfare is no longer a niche enabler but an all-arms concern, and that Ukraine’s battlefield adaptation has fused drones, manoeuvre, sensing, and spectrum contestation into one operational ecology. Research from the RAND Corporation reaches a parallel conclusion, emphasising that Ukraine’s wartime innovation only becomes strategically meaningful when embedded in policy, procurement, and institutional coordination. [3][4][11][10]

At the same time, academic debate has not converged on a single meaning of “drone revolution.” Hutto and Rogers argue that drones are sometimes ordinary and sometimes revolutionary, depending on conflict type, political objectives, and context, and that the two sides of the debate rest upon a false dichotomy between revolutionary and evolutionary accounts. [16] That is an important corrective to both technological determinism and false continuity. It implies that any serious conceptual analysis of drones should begin not with the platform, but with the environment and the adaptive system in which the platform is embedded.

A second body of literature focuses on legitimacy, accountability, and non-state diffusion. Brookings Institution research suggests that perceived legitimacy varies not simply with the existence of drones but with how they are used and constrained — specifically, that multilateral constraint and civilian casualty avoidance are the principal determinants of public legitimacy judgements. [15] UNIDIR finds that non-state armed groups in Africa most often use uncrewed systems for intelligence, surveillance, reconnaissance, and propaganda, and warns that shared knowledge and expertise are likely to expand that pattern over time. [14] RUSI’s analysis of AI-supported targeting in Gaza argues that the key issue is not whether AI is used, but whether it is used to improve discrimination or instead to accelerate target generation without equal improvement in verification and oversight. [29]

A third body of literature comes from systems engineering. The Department of Defense defines digital engineering as an integrated digital approach using authoritative data and models across the lifecycle. [1][2] A NASA handbook explains how system modelling can be integrated with standard systems-engineering processes. [17] INCOSE defines an engineered system as something designed or adapted to interact with an anticipated operational environment to achieve a purpose under constraints. [18] Department of Defense guidance on digital threads describes them as communication frameworks that enable connected data flows and integrated views across a system’s lifecycle. [1]

What remains underdeveloped in the literature is the synthesis of these three conversations. Strategic studies show that autonomy, logistics, EW, and air defence are interdependent. Political studies show that drones alter sovereignty, legitimacy, and patterns of non-state violence. Systems engineering offers a rigorous vocabulary for traceability, lifecycle learning, and boundary definition. But relatively few works join them to ask the central question of this paper: how adaptive coercive warfare changes the political architecture of the state itself. This paper treats that synthesis as its principal contribution. [3][4][10][12]

Conceptual Framework

The foundational move in this paper is to redefine the system of interest. In traditional acquisition or force-planning debates, the system is often treated as a platform or a platform family. That boundary is now too narrow. The relevant system is the broader coercive ecosystem: reconnaissance and strike drones, EW, command software, training pipelines, procurement authorities, industrial suppliers, repair loops, legal constraints, and the external financial and sanctions environment in which they operate. Systems-engineering sources support this move by emphasising that system properties emerge from interactions among components and between a system and its environment, not from the characteristics of any one part taken in isolation. [18][17][4][19]

This broader boundary makes it possible to adapt the digital-thread concept into a warfighting frame. In industrial development, a digital thread links design, manufacturing, testing, sustainment, and feedback. In adaptive coercive warfare, the analogous thread links battlefield observation, requirement-setting, software changes, procurement choices, industrial ramp-up, fielding, battle damage assessment, and doctrinal revision. The conceptual value lies in traceability and speed: whether an actor can keep the full loop coherent enough to learn faster than the adversary. [1][19]

This paper therefore uses adaptive coercive warfare to mean coercion organised around continuous iteration between sensing, command, software, industrial replenishment, and doctrine. It uses systems state as an analytical term for a state, coalition, or quasi-state actor that can define the operating environment, derive and revise requirements from that environment, align those requirements with industrial feasibility, and keep the observation-to-adaptation loop running fast enough to preserve coercive advantage.

A note on the systems-state concept. The term is analytical rather than doctrinal. It is introduced here to capture a specific capability pattern that current sources describe but do not yet name consistently: the capacity to close the adaptive loop at sufficient speed and coherence. A conventionally powerful state may possess large defence budgets, mature industrial capacity, and advanced platforms, yet still lack the digital thread, data governance, and cross-domain interface standards required to sustain adaptive coercive warfare at tempo. Conversely, a less wealthy actor that has invested in tight observation-to-production linkages — as Ukraine has done under wartime pressure — may exhibit higher adaptive system quality than a better-resourced but more siloed adversary. The systems-state concept names this distinction; it does not restate conventional measures of military power. [1][10][6]

A closely related concept is infrastructural sovereignty. In this paper, infrastructural sovereignty means the degree to which an actor controls, secures, or can reliably access the enabling layers of coercion: frequencies, data architecture, software stacks, integration standards, high-risk components, repair capacity, and export permissions. A state may have drones without possessing infrastructural sovereignty. Ukraine’s long-range programmes remain constrained by mini-turbojet bottlenecks; Russia’s Geran output has grown despite sanctions through external component channels; and RUSI’s supply-chain work argues that Western drone production is still heavily exposed to Chinese-origin materials and components. [8][9][7]

The figure below maps the system boundary used throughout the paper. It extends the “weapon system” outward into the data, industrial, and governance layers that determine whether drone warfare becomes merely tactical or strategically sustainable. [1][10][4][7]

flowchart TB
    subgraph Edge["Operational edge"]
      ISR["ISR drones and sensors"]
      Strike["FPV, loitering, and interceptor drones"]
      EW["Electronic warfare and spectrum contestation"]
      Ops["Operators and small units"]
    end

    subgraph C2["Decision and integration layer"]
      COP["Common operating picture"]
      Tasking["Tasking and battle management"]
      Doctrine["Tactics, doctrine, lessons learned"]
    end

    subgraph Industry["Industrial and training layer"]
      RnD["R&D, startups, volunteer engineering"]
      Assembly["Assembly and platform integration"]
      Components["Engines, chips, sensors, software, links"]
      Train["Training pipelines"]
      Repair["Repair, replacement, sustainment"]
    end

    subgraph Governance["Governance and external layer"]
      Proc["Procurement and budgeting"]
      Reg["Spectrum, standards, export control, law"]
      Allies["Allies, investors, sanctions, foreign suppliers"]
      Infra["Critical infrastructure governance"]
    end

    Components --> Assembly --> Strike
    RnD --> Assembly
    Train --> Ops
    Repair --> Strike
    ISR --> COP --> Tasking --> Strike
    EW --> Tasking
    Doctrine --> Tasking
    Strike --> Feedback["Battle damage, telemetry, battlefield feedback"]
    Feedback --> COP
    Feedback --> Doctrine
    Feedback --> RnD
    Proc --> Assembly
    Reg --> Components
    Reg --> EW
    Allies --> Components
    Allies --> Proc
    Infra --> COP

The second indispensable concept is the sense-making loop. Traditional doctrinal and acquisition cycles are too slow for environments in which countermeasures, software patches, frequencies, supply routes, and payload designs change within weeks or even days. RAND’s Ukraine work argues that sustaining advantage requires stronger formal learning channels without extinguishing Ukraine’s informal innovation culture. Current battlefield reporting likewise shows iterative adaptation in interceptors, software, and counter-drone methods across theatres. [11][10][20][34]

flowchart LR
    Observe["Observe battlefield data, losses, interceptions, EW effects"]
    Validate["Validate reports with telemetry and operator feedback"]
    Model["Update threat and environment models"]
    Require["Revise requirements for payload, range, software, training"]
    Produce["Adjust production, procurement, and software releases"]
    Field["Field updates and retrain operators"]
    Employ["Employ capabilities in the contested environment"]
    Learn["Institutionalise lessons in doctrine and standards"]

    Observe --> Validate --> Model --> Require --> Produce --> Field --> Employ --> Observe
    Validate --> Learn
    Learn --> Require

Adaptive system quality is a system-level property of a coercive ecosystem, defined as its capacity to maintain end-to-end loop integrity — from sensing through command, production, and fielding — under adversarial perturbation. It is assessed across four dimensions: loop closure rate (the speed and completeness of observe-orient-decide-act cycles); requirements traceability (bidirectional linkage from operational observation to fielded design change, consistent with digital-thread principles); interface coherence (the degree to which subsystem interfaces share common data models and standards, enabling seamless handoff across the ecosystem boundary); and supply-chain resilience (the redundancy and diversity of critical enabling inputs relative to single-point-of-failure risk). High adaptive system quality does not require perfection in any single dimension; it requires sufficient coherence across all four that the loop continues to close faster than an adversary can disrupt it.

Operationalising adaptation speed. As a primary measurable proxy, adaptation speed can be assessed as the elapsed time between a documented battlefield event — a platform loss, an EW interception, or a new countermeasure deployment — and a verifiable change in production specification, operator training syllabus, or fielded software release. A secondary proxy is the frequency of formal doctrine revision cycles. Both can be assessed qualitatively across cases using publicly available procurement announcements, training updates, and policy documents. Comparative estimates for each case are provided in the case studies; full quantitative data are not yet available in open sources and that gap should be treated as a priority research task before any quantitative claims about adaptation speed are made.

Methodology

The most appropriate research design for this paper is structured, focused comparison. In the classic formulation, the method is structured because the researcher asks each case the same general questions, and focused because the researcher examines only those aspects of each case that bear directly on the research objective. That design is especially well suited to drone warfare because the surface manifestations of drone use vary so widely across theatres. Without a common question set, comparison quickly becomes anecdotal and hype-driven. [33][16]

Accordingly, every case is assessed through the same analytic sequence: operational environment, requirement structure, industrial feasibility, adaptation speed, and political effects. This sequence is drawn from the systems framework above and is intended to discipline causal inference.

Dimension	Core question	Why it matters
Operational environment	What terrain, target set, infrastructure exposure, and electromagnetic conditions define the battlespace?	Drones produce different effects in trench warfare, maritime chokepoints, cities, borders, and weak-state environments.
Requirements	What mix of ISR, strike, EW, C2, logistics, and operator skill does that environment require?	Political effects often follow from the support architecture, not the drone alone.
Industrial feasibility	What can be produced, repaired, replaced, and funded at scale, and with what dependencies?	Tactical innovation without replenishment rarely creates durable leverage.
Sense-making loop	How quickly do lessons move from field observation into software, procurement, training, and doctrine?	Adaptation speed is the bridge between battlefield performance and state capacity.
Political effects	How do these linkages alter sovereignty, legitimacy, alliance patterns, internal authority, and coercive monopoly?	This is the paper’s primary dependent variable.

The source strategy follows the same logic. Where official or primary materials exist, they are prioritised. Those sources are supplemented with contemporary research from RAND, CSIS, and RUSI, and with Reuters reporting when official data are incomplete or when fast-moving industrial and organisational changes are not yet captured in formal documents. RAND’s Ukraine project uses mixed methods including literature review, stakeholder interviews, workshops, and field visits; this paper cannot replicate that full design, but it follows the same general principle of source triangulation and explicit caution where facts remain contested, particularly in cases involving AI-enabled targeting claims. [11][10][29]

Note on source coverage. The source base is predominantly Anglophone and Western. The following non-Western analytical sources are cited in the relevant case studies: for the Russia and Ukraine case, Observer Research Foundation (New Delhi) [37]; for the Red Sea and Houthis, Al-Jazeera Centre for Studies (Doha) [38]; for Sudan, Institute for Security Studies (Pretoria) [39]; for Turkey, SETA Foundation (Ankara) [40]; for Iran, Middle East Council on Global Affairs (Doha) [41]; and for Israel and Gaza, SETA Foundation (Ankara) [42]. Where no English-language publication from the originally proposed organisation was identified, a substitute non-Western source has been used; details are noted in the relevant footnotes.

Case selection is purposive rather than statistical. The six principal cases were chosen because they vary across the political conditions most relevant to the paper’s thesis: state capacity, industrial depth, infrastructural sovereignty, reliance on external suppliers or patrons, density of non-state participation, and legal-regulatory visibility. [16][22][23][24][25][26][28]

All figures derived from journalistic or single-source reporting are flagged as [estimate pending verification] throughout the text. Before final publication, all directional claims should be replaced with independently verified figures where available, and explicitly labelled as estimates if authoritative data remain unavailable.

Drone Technology and Industrial Contours

The most important technological fact is that the drone is rarely the full system. Ukraine’s current procurement plans illustrate both the scale and the institutional depth of the change. Kyiv purchased more than 1.5 million FPV drones in 2024 [estimate pending verification] and plans to buy roughly 4.5 million in 2025 [estimate pending verification]; simultaneously, Ukrainian authorities are pushing manufacturers to integrate unmanned systems into shared situational-awareness and fire-correction environments in order to build a common operating picture. CSIS also reports that some AI-enabled unmanned systems can now be learned in as little as 30 minutes to one day [estimate pending verification], which lowers operator bottlenecks but increases the importance of standards, interfaces, and data fusion. [5][19][15]

This systems logic also defines contemporary air defence. Reuters reports that Ukraine is pursuing a layered defence against Shahed-type attacks that combines interceptor drones, EW, gun teams, and other assets, [20] while CSIS’s analysis of Iran’s 2026 Gulf campaign argues that large waves of low-cost drones can impose operational and economic strain precisely because they force defenders to spend expensive interceptors against cheap attackers. [27] Reuters reporting from the Gulf reinforces the point: several Gulf states are exploring low-cost Ukrainian-designed interceptor drones as a cheaper alternative to relying primarily on missile defence. [31] In effect, offensive and defensive drone ecosystems are now part of the same strategic problem.

Industrial bottlenecks are therefore not secondary variables. They are part of combat effectiveness. Ukraine’s deep-strike drone fleet is constrained by a mini-turbojet shortage centred on a small group of European suppliers. Russia’s Geran production, by contrast, has reportedly expanded through covert Chinese engine shipments disguised as refrigeration units [estimate pending verification]. [8][9] RUSI’s supply-chain work generalises the problem: Western and allied multirotor ecosystems remain heavily exposed to Chinese-origin semiconductors, magnets, sensors, and related upstream inputs. [7] From a systems-engineering perspective, these bottlenecks are interface failures: points in the supply chain where the boundary between an actor’s controlled system and its uncontrolled environment is most exposed to adversarial disruption.

Export-oriented and sanctions-adapted ecosystems show another version of the same dynamic. Baykar told Reuters in 2024 that it would invest $300 million [estimate pending verification] over five years to develop jet engines and internalise more component production. [26] Meanwhile, Ukrainian drone diplomacy has moved from battlefield adaptation toward transnational influence: Ukraine has sent air-defence teams to three Gulf states [estimate pending verification], and its broader drone expertise is becoming a source of diplomatic leverage, potential exports, and industrial positioning even though regulatory and production constraints remain substantial. [32][33] These cases show that drone ecosystems are now instruments of political economy and foreign policy as much as of combat.

The supply-chain map below summarises where coercive dependence tends to hide. From a systems-engineering standpoint, it is a map of where interface governance is most consequential. [7][8][9][26]

flowchart LR
    Raw["Raw materials and industrial inputs"] --> Components["Chips, optics, motors, batteries, engines, radios"]
    Components --> Software["Firmware, autonomy modules, battle-management software"]
    Software --> Integration["Platform integration and assembly"]
    Integration --> Fielding["Fielding, operator training, maintenance"]
    Fielding --> Use["ISR, strike, interception, EW support"]
    Use --> Feedback["Telemetry, losses, battle damage, operator feedback"]
    Feedback --> Software
    Feedback --> Integration
    Feedback --> Procurement["Procurement priorities and contracts"]

    Ext1["External chokepoints: component origin, sanctions, shipping, finance"] --> Components
    Reg2["Regulatory chokepoints: export control, standards, data access, spectrum rules"] --> Software
    Reg2 --> Procurement
    Alliances["Alliance and partner inputs: aid, co-production, licensing, foreign training"] --> Components
    Alliances --> Procurement

Operational Environment Typology

The evidence strongly suggests that no single theory of drones can explain modern warfare. Instead, the political effects of drones vary with conflict type, geography, target vulnerability, actor mix, intensity of sensing, and the speed of adaptation. The typology below is designed to keep the comparative analysis environment-first rather than platform-first. [16][3]

Environment type	Representative pattern	Dominant dependency	Core political question
Industrial continental attritional war	Ukraine–Russia mass production, EW saturation, and common-operating-picture integration [3][35][5][19]	National industry plus allied supply and data standards	Can the state synchronise output, defence, and doctrinal learning faster than a peer adversary?
Maritime chokepoint coercion	Red Sea disruption, convoy protection, and coalition naval burden-sharing [22][13]	Geographic concentration and defender cost disadvantage	How far can a low-cost actor impose global trade costs without full state capacity?
Urban high-surveillance conflict	Dense intelligence, startup-state integration, and accelerated targeting in Israel/Gaza [36][29]	Software, target-verification standards, and legitimacy	Does accelerated machine-assisted targeting generate coercive efficiency or political backlash?
Criminal-border coercion	Commercial drones for scouting, smuggling, intimidation, and localised attack in Mexico [23]	Commercial availability and uneven territorial governance	When does cheap aerial access erode local monopoly of violence without producing true military parity?
Weak-state infrastructure war	Drone attacks on electricity, transport links, and urban centres in Sudan [24][25]	External patrons, fragile grids, and territorial fragmentation	How do drones accelerate political collapse by weaponising the systems through which citizens experience governance?
Export-patronage and sanctions-adapted competition	Turkish export ecosystems and Iranian transfer networks [26][27][28]	Industrial policy, transfer channels, and sanctions evasion	How do drone ecosystems become tools of statecraft, proxy leverage, and regime resilience?

The value of this typology is practical as well as analytic. It explains why the same hardware category can produce very different political outcomes across theatres. It also shows why misclassification is dangerous: a state that treats maritime coercion like trench attrition, or organised crime like intermediate-state war, is likely to build the wrong procurement strategy, regulate the wrong chokepoints, and infer the wrong political lessons from tactical events. [16][10]

Comparative Case Studies

The six cases below use one common template: context, adaptive system, dependencies, political effects, and adaptation speed estimate. The purpose is not to retell each conflict. It is to compare how different coercive architectures redistribute authority among the state, firms, external patrons, and armed networks, and to apply the adaptation speed proxy across cases. [33][16]

Ukraine

Ukraine is the clearest case of adaptive coercive warfare strengthening rather than simply bypassing state capacity. The state has not displaced volunteer innovation or private manufacturing; instead, it has progressively absorbed them into wartime procurement, data integration, and informal-to-formal learning channels. Mass procurement of domestically sourced FPV drones, accelerated training pipelines, growing integration into shared situational-awareness systems, and RAND’s emphasis on clearer institutional roles and longer-term contracting all point in the same direction: tactical decentralisation can coexist with stronger state capacity when the state standardises and scales bottom-up innovation instead of suppressing it. Yet the case also reveals the limits of sovereignty under wartime dependence, including continuing exposure to external air-defence support, engine bottlenecks, and component vulnerabilities. Politically, Ukraine demonstrates how battlefield adaptation can become both a state-building process and a source of external diplomatic leverage. [5][19][15][10][11][8][32][33]

Adaptation speed: fast (weeks). Ukrainian engineers have shortened design-to-production cycles from months to weeks, pushing new airframe configurations, EW-resistant firmware, and counter-measure responses into service faster than any Western acquisition process. [43] This is the most clearly documented instance of a functioning sense-making loop operating at operational tempo among the six cases in this study.

For a non-Western perspective on Russia’s parallel industrial adaptation, see Observer Research Foundation (New Delhi), which documents how Russia has sought to build sovereign drone production capacity in response to battlefield attrition but at significantly longer cycle times than Ukraine. [37]

Red Sea and the Houthis

The Red Sea case shows how quasi-state actors can impose strategic costs without achieving full sovereign capability. Houthi operations have disrupted shipping, forced multinational naval governance, and shifted security burdens onto external coalitions because the geography of the Bab al-Mandab chokepoint concentrates vulnerability. Reuters reports that traffic recovered to 36–37 ships per day [estimate pending verification] by June 2025, up 60 per cent [estimate pending verification] from the 2024 trough but still far below the pre-crisis range of 72–75 ships [estimate pending verification]. [22] The European Union confirms that EUNAVFOR ASPIDES was extended through February 2027 to help protect freedom of navigation. [13] The political consequence is not state replacement but coalition externalisation: a relatively low-cost actor can force expensive, prolonged, multinational defensive governance where trade routes and legal constraints magnify the effect of relatively simple systems.

Adaptation speed: moderate (months), patron-dependent. The Samad-4 variant entered service in September 2024; by February 2025, Iranian-supplied jet engine technology and FPV system integration were reportedly under development, suggesting design-to-field cycles of roughly three to six months. [44] This is faster than traditional state procurement but significantly slower than the Ukrainian model, and it is reliant on Iranian supply chains and commercial components rather than indigenous engineering capacity.

The Al-Jazeera Centre for Studies argues that the Houthis have succeeded in leveraging geographic chokepoint concentration to shift the coalition burden-sharing calculus regardless of their limited industrial base — a pattern that confirms the paper’s typology: geography can substitute for adaptive system quality in environments where defender costs are structurally asymmetric. [38]

Mexico and the Cartels

Mexico illustrates the criminal-governance version of drone diffusion. Reuters reports that Mexican criminal groups have used cheap commercial drones for more than a decade to conduct surveillance and transport contraband, while official U.S. border testimony has highlighted continuing incidents involving illicit UAS activity between ports of entry. [23] These patterns do not amount to military symmetry with the state. What they do produce is a lower-cost form of territorial monitoring, intimidation, and coercive experimentation in environments where state presence is uneven. Politically, the result is not a “drone army” but a sharper erosion of local monopoly of violence, greater pressure for cross-border counter-UAS cooperation, and a heightened risk that emergency security responses become increasingly militarised without solving the governance deficits that make cartel drone use effective in the first place.

Adaptation speed: commercially driven, months for platform adoption. Recent evidence complicates a simple “not applicable” assessment. CJNG conducted Mexico’s first confirmed FPV drone attack in April 2025; by September 2025 the Sinaloa cartel had also adopted FPV systems, suggesting a three-to-six-month cross-cartel diffusion cycle for new platform types. [45] Seizures of custom-modified FPV payloads with circuitry resembling Ukrainian DIY loitering munitions, and credible reporting of Latin American operatives joining Ukrainian volunteer units specifically to acquire FPV skills, indicate that knowledge transfer — not indigenous R&D — is the primary adaptation mechanism. The institutional adaptation proxy is not applicable: cartels have no requirements-traceability architecture or formal doctrine revision cycles. The commercial and social-network adaptation pathway is however active and accelerating, and should not be dismissed as strategically irrelevant.

Turkey and Iran as Second-Tier Drone Powers

In Turkey and Iran, drones function as instruments of middle-power statecraft. Turkey’s model is export-centred and increasingly vertically integrated: Baykar has sold to dozens of countries, internalised more of its production chain, and is investing heavily in domestic engines. SETA (Ankara) frames Turkey’s emerging autonomous systems — including the GÖKSUR and ULAQ platforms — as an enabling technology for future warfighting across all domains, reflecting a deliberate state strategy to leverage drone production as both a military and a diplomatic instrument. [40] Iran’s model is sanctions-adapted and transfer-oriented: CSIS’s 2026 Gulf analysis portrays drones as the primary instrument of sustained low-cost pressure, [27] while Reuters reports that Iranian officials are explicitly willing to share “defensive capabilities” with foreign partners. [28] The Middle East Council on Global Affairs characterises Iran’s drone programme as a deliberate instrument of strategic disruption calibrated to impose asymmetric costs on better-resourced adversaries, degrading U.S. aerial hegemony through distributed, attrition-based pressure. [41]

Adaptation speed — Turkey: industrial (years). Baykar’s successive platform generations — TB2, Akinci, Kizilelma — operate on multi-year R&D cycles driven by export strategy and combat experience across multiple theatres, not by weeks-scale battlefield feedback loops. Adaptation speed — Iran: slow to moderate (months to a year). Geran/Shahed variant evolution provides the clearest available proxy: a heavier 90 kg warhead variant was first observed in May 2024 and widely deployed by May 2025, indicating a design-to-deployment cycle of approximately twelve months. Component-level adaptation — including anti-jamming antenna integration and Starlink communications modules — occurs on a shorter three-to-six-month timescale following battlefield observation. These cycles are notably slower than Ukraine’s weeks-scale model but faster than traditional state procurement, consistent with an actor that relies on incremental modification of proven platforms rather than rapid iterative redesign.

Israel and Gaza

The Israel–Gaza case demonstrates the other edge of systems integration: the possibility that higher-speed targeting and close startup-state fusion may increase operational tempo while simultaneously intensifying legitimacy and accountability risk. Reuters reports that Israel’s wartime “green path” fast-tracked selected startups, generated orders to 101 startups and small firms [estimate pending verification], moved more than 25 from development to production [estimate pending verification], and saw roughly half of anti-drone technology [estimate pending verification] used during the conflict come from startups. [36] RUSI’s analysis warns that the principal concern is not AI as such but the apparent use of AI to expand and accelerate the target cycle, with the underlying facts of public reporting remaining contested unless the systems themselves are independently scrutinised. [29]

SETA (Ankara) frames the conflict as a testing ground for AI warfare in which commercial innovation velocity is deliberately harnessed to compress targeting cycles — raising the same accountability questions that RUSI identifies, but from a Turkish analytical perspective that emphasises the precedent-setting nature of the conflict for other state actors developing autonomous weapons. [42] Politically, this is the paper’s clearest example of a state becoming more coercively integrated while simultaneously becoming more vulnerable to legitimacy loss, legal contestation, and strategic backlash.

Adaptation speed: fast (weeks). Israel’s startup-state integration model produces the same weeks-scale design-to-field cycles as Ukraine’s volunteer-engineering model, though through commercial contracting rather than wartime bottom-up innovation. Both cases confirm that fast adaptation speed is achievable through different institutional mechanisms.

Sudan as the Weak-State Example

Sudan is the strongest case for how drones accelerate fragmentation where infrastructural sovereignty is already weak. Reuters documented blackouts across large army-controlled areas after drone attacks on the Merowe Dam complex and other power infrastructure. [24] ACLED reports that drone strikes in 2025 were 47 per cent higher [estimate pending verification] than in the first eleven months of 2024 and that the RSF’s improved drone arsenal enabled longer-range attacks against SAF-controlled areas, including Port Sudan. [25][30] Politically, the significance is clear: drone warfare in weak states often weaponises the everyday infrastructures through which governance is experienced — electricity, aid access, transit, and hospital operations. The result is not efficient coercive order but a deepening of territorial partition, humanitarian collapse, and dependence on rival external patrons.

Adaptation speed: externally driven, no internal loop. The RSF’s introduction of Chinese-made FH-95 drones in late 2024, first confirmed by satellite imagery at Nyala Airport, reflects UAE supply decisions rather than internal design cycles. No evidence of indigenous drone modification or battlefield-feedback-driven redesign has been found in open sources. ISS (Pretoria) documents the broader pattern: African armed groups are increasingly acquiring military-grade drones through external patron channels, with adaptation determined by patron capabilities rather than operational feedback — confirming that the sense-making loop framework does not apply where institutional absorptive capacity is absent. [39]

Comparative Synthesis

The cases point toward one general conclusion: drones redistribute coercive power differently depending on the quality of the surrounding system. Where state institutions can absorb bottom-up innovation and coordinate industrial adaptation, drones can strengthen public authority. Where geography concentrates vulnerability, drones can let non-state actors externalise security costs onto coalitions. Where governance is thin, they deepen fragmentation. And where state-tech integration is already high, they can sharpen both effectiveness and accountability pressures at once. [10][11][22][24][29]

The adaptation speed estimates now allow a first cross-case comparison. Fast adaptation (weeks) is associated with the clearest cases of sustained coercive advantage — Ukraine and Israel. Moderate adaptation (months) is associated with patron-dependent actors who can still impose strategic costs through geographic leverage — the Houthis. Industrial-cycle adaptation (years) characterises export-focused middle powers whose advantage lies in platform generation rather than battlefield iteration. Where adaptation is externally driven or not applicable, the sense-making loop is either absent or substituted by geographic, demographic, or commercial factors.

Case	State capacity	Infrastructural sovereignty	Adaptation speed	Dominant political effect
Ukraine [5][10][11][8]	High under wartime mobilisation	Medium and improving, but externally constrained	Fast (weeks)	State strengthening through institutionalised adaptation, offset by strategic bottlenecks
Red Sea / Houthis [22][13]	Quasi-state and sponsor-dependent	Low domestically, high through geography and networked support	Moderate (months, patron-dependent)	Outsized disruption of trade and multinational burden-shifting
Mexico / cartels [23][45]	Uneven and territorialised	Partial and locally contested	Commercially driven (months, network-transfer)	Erosion of local monopoly of violence without full military parity
Turkey [26][40]	Strong	High and rising	Industrial (years, export-driven)	Export leverage and industrial prestige through vertically integrating drone production
Iran [27][28][41]	Strong coercive core under sanctions	Medium under sanctions pressure	Slow to moderate (months to a year)	Proxy leverage, transfer-based influence, and low-cost pressure campaigns
Israel / Gaza [36][29][42]	Very high for the state	High for the state, very low for the territory under attack	Fast (weeks, startup-integrated)	Greater coercive integration paired with intensified legitimacy and accountability risk
Sudan [24][25][30][39]	Fragmented	Low and externally shaped	Externally driven, no internal loop	Infrastructure attack, partition, and deeper humanitarian collapse

Conclusion and Recommendations

The paper’s core conclusion is that the decisive political variable in drone warfare is not drone possession, but adaptive system quality. Actors that can maintain a live digital thread between sensing, software, industrial replenishment, training, and governance are more likely to convert tactical innovation into durable coercive advantage. Actors that cannot may still inflict harm, but they are more likely to remain dependent, brittle, or fragmentary. In that sense, the future contest is less about “drone armies” than about which actors become systems states. [1][4][10][11]

For systems engineers and policy practitioners working on defence-industrial and lifecycle governance, the implications fall into six areas.

Govern the system boundary, not just the platform. The coercive ecosystem — not the drone — is the correct unit of strategic investment and assessment. Capability reviews, acquisition strategies, and threat assessments should define system boundaries that include sensing, command, software, training, supply chain, and feedback loops, not platforms in isolation. [1][2]

Implement the digital thread. The single most consequential technical investment is a data architecture and interface-standards regime that links battlefield observation to design requirement, production, and fielded capability, and carries operational feedback back through the same chain. Without a working digital thread, sense-making loop speed is limited by the brittleness of informal channels. [1][2]

Treat supply-chain topology as a strategic security assessment. Single-point-of-failure risks in critical component supply — engines, chips, sensors, software libraries — are interface failures with strategic consequences. Supply-chain topology analysis should be conducted with the rigour applied to safety-critical systems, identifying and governing chokepoints before adversarial disruption exposes them. [7][9][12]

Build formal requirements-traceability architecture. States that can formally trace requirements from operational observation through procurement to fielded change — and back — will out-adapt adversaries that rely on informal feedback. Bidirectional traceability, familiar from safety-critical SE practice, is a direct political-military advantage in contested drone warfare. [1][17][18]

Replace episodic acquisition with adaptive lifecycle governance. Traditional procurement cycles are too slow for contested drone environments. Lifecycle governance frameworks should be redesigned to accommodate rapid design iteration, short production runs, and continuous fielding — drawing on the digital-engineering governance models that the Department of Defense and allied procurement bodies are developing. [1][2]

Frame counter-UAS as a system-of-systems integration challenge. Layered counter-drone defence is not a platform procurement question. It is an integration and interface-management problem: ensuring that EW, interceptor drones, point defences, and command software share a common operating picture and can be tasked coherently under contested conditions. [4][20][31]

The official EU response in the Red Sea, DoD digital-engineering guidance, RAND’s Ukraine recommendations, and RUSI’s supply-chain analysis all support this direction of travel. [13][1][10][7]

Next Steps for Final Publication

The following actions remain before this paper is ready for final publication.

Quantitative annex. Annex A lists all directional claims currently marked [estimate pending verification] with their sources and the independent data required. These should be resolved before submission; claims that cannot be independently verified should be explicitly labelled as estimates in the final text.
Footnote [16] author confirmation. Author attribution for the EJIS article [16] is based on institutional affiliation metadata and should be confirmed via direct journal access before submission.
Peer review. At least one reviewer with a systems-engineering background should review the conceptual framework and recommendations before final release.

Annex A: Quantitative Claims Pending Verification

The following claims are flagged [estimate pending verification] in the body text. Each entry notes the claim, its current source, and the independent data needed before publication.

Claim	Current source	Verification required
Ukraine purchased >1.5 million FPV drones in 2024	Reuters [5]	Cross-check against Ukraine MoD procurement data or KSE Institute report
Ukraine plans to purchase ~4.5 million FPV drones in 2025	Reuters [5]	Verify against 2025 procurement outturn once published
96% of Ukraine’s FPV drones sourced domestically	Reuters [5]	Single journalistic source; cross-check against CSIS [19] or RAND [10] supply-chain data
Russia operating 48 drone R&D and production centres	Reuters via CSIS [6]	Verify against CAST [37] or independent industrial analysis
Russia targeting 70% domestic market capture for UAS	Reuters via CSIS [6]	Verify against official Russian procurement statements
AI-enabled UAS trainable in 30 minutes to one day	CSIS [19]	Verify system type and training context; generalisation risk is high
Russia’s Geran production expanded via Chinese engine shipments disguised as refrigeration units	Reuters [9]	Verify against US government sanctions designations or OSINT supply-chain analysis
Baykar investing $300 million over five years in domestic jet engine development	Reuters [26]	Verify against Baykar investor communications or Turkish defence industry reports
Ukraine sent air-defence teams to three Gulf states	Reuters [33]	Verify against official Ukrainian MFA or Gulf government statements
Red Sea traffic at 36–37 ships/day by June 2025 (up 60% from 2024 trough)	Reuters [22]	Verify against EUNAVFOR ASPIDES, Lloyd’s, or IMO shipping data
Red Sea pre-crisis baseline of 72–75 ships/day	Reuters [22]	Verify against Lloyd’s List or IMO annual shipping statistics
Sudan drone strikes 47% higher in 2025 vs first 11 months of 2024	ACLED [25]	ACLED methodology note should be cited; verify coding criteria for drone events
Israel’s “green path” generated orders to 101 startups	Reuters [36]	Verify against Israeli MoD or IDF official statements
More than 25 Israeli startups moved from development to production	Reuters [36]	Single journalistic source; verify against Israeli defence ministry data
Roughly half of Israel’s anti-drone technology came from startups	Reuters [36]	Single journalistic source; no independent verification found
Iran Geran/Shahed 90 kg warhead variant observed May 2024, widely deployed May 2025	Open-source munitions tracking	Verify against OSMP, Oryx, or similar open-source munitions databases
CJNG first confirmed FPV drone attack April 2025	DroneXL [45]	Verify against US Customs and Border Protection or DEA reporting

Footnotes

[1] Office of the Under Secretary of Defense for Research and Engineering. Digital Engineering Practice. Department of Defense.

[2] Office of the Under Secretary of Defense for Research and Engineering. Digital Engineering Strategy. Department of Defense, 2018.

[3] Slusher, Matthew. Lessons from the Ukraine Conflict: Modern Warfare in the Age of Autonomy, Information, and Resilience. CSIS, May 2025.

[4] Watling, Jack, and Noah Sylvia. Competitive Electronic Warfare in Modern Land Operations. RUSI, January 2025.

[5] Reuters. “Ukraine to Sharply Raise Purchases of Home Produced FPV Drones in 2025.” March 2025.

[6] Bondar, Kateryna. How Russia Is Building a Sovereign Drone Ecosystem for AI-Driven Autonomy. CSIS, April 2026.

[7] Tollast, Robert. Drones: Decoupling Supply Chains from China. RUSI, November 2025.

[8] Reuters. “Ukraine’s Attack Drone Fleet Faces a Mini Jet Engine Supply Crunch.” April 2026.

[9] Reuters. “Chinese Engines, Shipped as ‘Cooling Units,’ Power Russian Drones Used in Ukraine.” July 2025.

[10] Paillé, Pauline, Mattias Eken, Thomas Kenchington, and Beatrice Aubert. From Policy to Victory: Recommendations to Ukraine for Harnessing Defence Technology. RAND, November 2025.

[11] RAND Europe. Wartime Innovation and Adaptation: Supporting Ukraine’s Digital Transformation. RAND, 2025.

[12] Shumate, William, et al. Export Controls on Artificial Intelligence and Uncrewed Aircraft Systems: Interagency Challenges. RAND, February 2026.

[13] Council of the European Union. Red Sea: Council Extends the Mandate of Operation ASPIDES to Safeguard Freedom of Navigation. February 2026.

[14] Figueiredo, Bárbara Morais. The Use of Uncrewed Aerial Systems by Non-State Armed Groups: Exploring Trends in Africa. UNIDIR, January 2024.

[15] Lushenko, Paul, and Sarah Kreps. “What Makes a Drone Strike ‘Legitimate’ in the Eyes of the Public?” Brookings Institution, 5 May 2022.

[16] Hutto, James Wesley, and James Patton Rogers. “The Drone Revolution: Towards a Synthesis in the Drone Debate.” European Journal of International Security, 2025. DOI: 10.1017/eis.2025.10005.

[17] NASA. NASA Systems Modeling Handbook for Systems Engineering. NASA-HDBK-1009A, March 2025.

[18] INCOSE. Systems Engineering Handbook, 5th edition. Wiley, 2023.

[19] Bondar, Kateryna. Ukraine’s Future Vision and Current Capabilities for Waging AI-Enabled Autonomous Warfare. CSIS, March 2025.

[20] Reuters. “Inside Ukraine’s Drive to Defeat the Dreaded Shahed Drone.” April 2026.

[22] Reuters. “Red Sea Marine Traffic Up 60% After Houthis Narrowed Targets, EU Commander Says.” June 2025.

[23] Reuters. “Cartel Drones Become Flashpoint Between US and Mexico.” February 2026.

[24] Reuters. “Power Outages Hit Army-Controlled Sudan After Drone Attacks.” January 2025.

[25] ACLED. “Fighting Moves to Kordofan as Sudan’s East–West Divide Solidifies.” December 2025.

[26] Reuters. “Turkish Drone Maker Baykar to Invest $300 mln to Develop Jet Engine, CEO Says.” October 2024.

[27] Bondar, Kateryna. Unpacking Iran’s Drone Campaign in the Gulf: Early Lessons for Future Drone Warfare. CSIS, March 2026.

[28] Reuters. “Iran Willing to Share Defensive Capabilities With Asian Partners, Deputy Defence Minister Says.” April 2026.

[29] Watling, Jack. The Israel Defence Forces’ Use of AI in Gaza: A Case of Misplaced Purpose. RUSI, July 2024.

[30] ACLED. “Q&A: How Do Escalating Tensions in the Middle East Affect the Sudan Conflict?” April 2026.

[31] Reuters. “Gulf States Eye Cheap Ukrainian Interceptor Drone as Iranian Attacks Drain Missile Stocks.” April 2026.

[32] Reuters. “Drone Diplomacy Wins Ukraine Valuable Allies, but Now It Must Deliver.” April 2026.

[33] Reuters. “Ukraine Sends Drone Experts to Three Countries in Middle East, Zelenskiy Says.” March 2026.

[34] Reuters. “US Turns to Ukrainian Counter-Drone Tech After Iran Attacks, Sources Say.” April 2026.

[35] Watling, Jack. Emergent Approaches to Combined Arms Manoeuvre in Ukraine. RUSI, October 2025.

[36] Reuters. “Israeli Startups Make Global Plans After Key Role in War.” January 2025.

[37] Observer Research Foundation (New Delhi). “The Evolution of Russia’s Drone Warfare in Ukraine.” ORF Expert Speak.

[38] Al-Jazeera Centre for Studies. “The Battle Over Shipping Lanes Tips Toward the Houthis.” Policy Brief, Doha.

[39] Institute for Security Studies (ISS). “Drones: A Propaganda Tool for Africa’s Armed Groups?” ISS Today, Pretoria.

[40] SETA Foundation for Political, Economic and Social Research. “GÖKSUR, ULAQ and the Algorithmic Defense of Blue Homeland.” SETA, Ankara.

[41] Middle East Council on Global Affairs. “Iran’s Missile and Drone Program: Disrupting U.S. Aerial Hegemony.” Doha.

[42] SETA Foundation for Political, Economic and Social Research. “Gaza as a Testing Ground: Israel’s AI Warfare.” SETA, Ankara.

[43] Army Recognition. “Ukraine Emerges as World Leader in Drone Technology Driven by Battle-Proven Innovation.” 2025.

[44] Sana’a Center for Strategic Studies. “New Technologies in Houthi Drones — The Yemen Review, January–March 2025.”

[45] DroneXL. “Cartels Deploy FPV Drones and Anti-UAS Systems in Criminal Arms Race.” September 2025. See also: Small Wars Journal. “Mapping Weaponized Drone Attacks Attributed to Mexican Drug Cartels.” February 2026.

FCPI Index — measuring indispensability of critical infrastructure layers
Platform Sovereignty — how digital platforms become governance objects
Indispensability Threshold — when dependency becomes structural
Starlink vs Starshield — satellite infrastructure as coercive dependency
Cloud Providers as Choke Points — how cloud dependency creates sovereignty risk
What Is Distributed Coercion — coercive structures across non-state actors

Platform Sovereignty FCPI Index Starlink vs Starshield