Abstract
The international economic order of recent decades has relied on specialization, cross-border production, and access to global technology markets. These arrangements expanded national capability while creating dependencies whose consequences become visible during supply disruption, export restrictions, or commercial withdrawal. AI and autonomous systems can change the resources required to maintain selected capabilities, including the labor, expertise, coordination, and organizational scale involved.
This paper develops a qualitative framework for examining how those changes affect sovereignty. The dependency premium describes the additional economic and strategic exposure associated with reliance on a capability once interruption and replacement are considered. The autonomy dividend describes an improvement in the relative practical or economic feasibility of maintaining a capability under domestic, regional, or trusted control attributable to AI and autonomous systems. Foreign suppliers also automate, and gains depend on complementary infrastructure, institutional competence, and implementation.
A Sovereign Capability Stack maps the physical, digital, and institutional dependencies supporting an essential function. It reveals a sovereignty paradox: measures that improve control at one layer can create new dependencies elsewhere. The resulting policy choices can support selective interdependence, combining international exchange with stronger continuity and credible alternatives around consequential functions. Shared capability can improve access where national replication is impractical, although its value depends on arrangements that remain usable under stress. Productive gains can coexist with unequal control over the infrastructure and conditions through which they are obtained.
The paper calls the possible international configuration arising from these interactions the New Sovereignist Order. Its emergence requires evidence that changes in productive feasibility influence state and firm decisions and become embedded in durable international practices. The analysis separates this order-level hypothesis from the practical usefulness of dependency assessment. It preserves the distinction between technological capacity and accountable exercise of that capacity, and identifies foundations for subsequent empirical assessment without proposing a country ranking or numerical sovereignty score.
Executive Summary
Governments are reconsidering technological dependencies while artificial intelligence and autonomous systems change the resources required to sustain some productive and administrative functions. Supply disruption, concentrated markets, export restrictions, and infrastructure risks make continuity and replacement more consequential. AI can alter the available alternatives by extending scarce expertise, automating tasks, or allowing smaller teams to maintain a capability. Realizing those gains requires investment in supporting infrastructure, organizational processes, and technical competence.
The interaction provides the basis for the New Sovereignist Order proposed in this paper. Governments may continue to obtain substantial benefits from international specialization while becoming more selective about dependencies whose interruption would constrain essential functions. The resulting pattern is described as selective interdependence. Its significance depends on whether these choices affect actual procurement, investment, access arrangements, and international relationships over time.
Two analytical concepts organize the comparison. The dependency premium describes the additional economic and strategic exposure associated with reliance on a capability once interruption and replacement are considered. Criticality, concentration, substitutability, recovery time, and exposure to restriction help identify that exposure. The autonomy dividend describes an improvement in the relative practical or economic feasibility of maintaining a capability under domestic, regional, or trusted control attributable to AI and autonomous systems. Both concepts are used qualitatively, without calibrated scores or monetary valuations.
The external alternative remains part of the comparison. Foreign suppliers can obtain greater benefits from automation through their infrastructure, financing, engineering, or scale. Falling domestic costs provide limited evidence of an autonomy dividend if the external supplier improves faster. Evaluation therefore needs a specified function, an acceptable performance standard, and the full resources required for deployment, operation, maintenance, transition, and replacement. The relevant gain may come from supporting existing staff or extending their capacity without eliminating their roles.
Where exposure is consequential and technology improves a controllable alternative, policymakers have stronger grounds to investigate that alternative. The appropriate response can still be diversification, inventories, shared capacity, or negotiated access. Governments may accept substantial costs to preserve an essential function even without an autonomy dividend. Where exposure is limited, automation may support commercially attractive localization while leaving public intervention to be justified separately. The qualitative comparison matrix organizes these possibilities without prescribing a universal decision rule.
The Sovereign Capability Stack traces an essential function through energy, strategic materials, semiconductors, compute, cloud and communications, data, models and software, cybersecurity and identity, physical systems, and human and institutional capacity. It identifies dependencies that may be obscured by national ownership or local deployment. Running an open model locally, for example, can reduce reliance on an external model service while increasing direct requirements for computing infrastructure, electricity, security, and engineering.
These relationships produce the sovereignty paradox. An investment can increase visible domestic capability while leaving a critical constraint elsewhere in the system. Conversely, diversified external provision with portable data and tested fallback arrangements can preserve greater continuity than a concentrated national alternative. Assessment should examine location, jurisdiction, operation, technical control, and continuity together with the availability and practicality of substitution. Optionality refers to alternatives that institutions can actually activate when an important relationship changes.
Countries can combine national capability, shared arrangements, and external access across different functions. These are overlapping choices rather than fixed categories or a country ranking. Coalition sovereignty becomes useful where shared infrastructure and institutions provide credible access during scarcity or disagreement. Strategic specialization can support selected capabilities without requiring national replication of the full stack. Intensive use of imported AI can coexist with substantial external dependence; adoption and the ability to sustain or replace a system require separate assessment.
The distributional implication concerns access to productive capability. AI can expand the availability of services and expertise while the infrastructure, knowledge, and conditions of access remain uneven. Governments may reserve strategically important capacity for domestic users or selected partners. Such choices depend on technological performance and political decisions. The paper treats abundance as a possible consequence of substantial sectoral gains, while recognizing that their distribution within countries also depends on institutions and ownership.
National decisions interact internationally. Restrictions can change expectations, encourage investment in alternatives, and prompt further adjustment by suppliers and other governments. Cooperation can also become more attractive where shared systems provide dependable access at lower cost than separate national provision. The order-level claim requires evidence that these responses persist and become embedded in rules, institutions, and commercial practices beyond established security exceptions.
Effective use of sovereign capability also depends on internal authority. Institutions need to define the permissions granted to autonomous systems, monitor consequential actions, preserve fallback arrangements, and retain the ability to replace systems. Accountability requires identifiable responsibility and effective review and remedy. These normative safeguards should be examined explicitly alongside technological capacity; national control alone provides no assurance that power is exercised accountably.
The policy objective is to preserve usable choices around consequential functions at a defensible cost. The framework can support future evidence-based assessment of capability, dependency exposure, and reliable access. Any numerical score would require additional decisions about indicators, comparability, missing evidence, and uncertainty. Its usefulness would remain separate from proof of an emerging international order. That broader explanation would gain support where AI improves the feasibility of alternatives, changes decisions, and contributes to durable international adjustment.
1. The Rise of Sovereignization
The international economic system that developed after the Second World War and expanded after the Cold War made cross-border integration a central source of national capability. Trade, investment, scientific exchange, and labor mobility allowed countries to obtain resources and expertise that would have been more expensive or difficult to reproduce domestically. The creation of the World Trade Organization in 1995 institutionalized part of this expansion [1]. Falling communication and transportation costs made it economically practical for firms to distribute different stages of production across multiple countries, contributing to the growth of increasingly complex global value chains [2].
The benefits of that system remain substantial. Specialization provides access to larger markets, advanced technology, capital, and internationally distributed expertise. The WTO continues to emphasize the contribution of international integration to prosperity and resilience, while examining pressures on the trading system [2, 3]. Governments nevertheless face dependencies whose failure would interrupt essential functions. OECD analysis identifies supply concentration, critical infrastructure, energy, cybersecurity, and access to technology as interconnected economic-security concerns [4].
AI and autonomous systems can change the resources required to maintain some of those functions. They may allow smaller workforces to sustain production, extend scarce expertise, or operate infrastructure with less continuous staffing. These changes affect the economic feasibility of alternatives to existing dependencies. This paper examines whether their interaction with geopolitical competition and limited trust could contribute to a different organization of the international economy.
1.1 Intellectual origins and the proposed contribution
The connection between interdependence and political power has a long intellectual history. Albert Hirschman examined how the distribution of trade and the difficulty of replacing commercial relationships could create influence over a trading partner. He also recognized the adjustment costs associated with interruption. His account supplies an early foundation for treating commercial efficiency and political dependence as features of the same relationship [5].
Robert Keohane and Joseph Nye developed the distinction between sensitivity and vulnerability interdependence. Sensitivity concerns the costs imposed within an existing policy framework. Vulnerability concerns the costs that remain after an actor has attempted to adjust its policies and relationships. Access to affordable alternatives consequently affects the power produced by interdependence [6]. The dependency premium developed here draws on this insight. It organizes interruption, concentration, substitution, and recovery concerns for contemporary technological systems, without claiming that the strategic value of alternatives is a new discovery.
Stephen Krasner distinguished international legal and Westphalian sovereignty from domestic sovereignty and interdependence sovereignty. His typology separates questions of recognition and external authority from the organization of public authority and control over cross-border movements [7]. The present paper concentrates on the material and institutional conditions under which states sustain particular functions. Its terms effective sovereignty and operational sovereignty are functional descriptions for that purpose, rather than a replacement for established legal or political typologies.
Farrell and Newman show how asymmetric networks can generate coercive leverage when states possess jurisdiction over central nodes [8]. Their account helps explain why the ownership, location, and legal exposure of technological infrastructure matter internationally. The framework developed here adds attention to how productive technologies change the cost of creating alternatives, including circumstances in which automation strengthens the very nodes on which other actors depend.
Contemporary technological-sovereignty research already addresses the relationship between capability and agency. Edler, Blind, Kroll, and Schubert treat technological sovereignty as the ability of states to act through the development of, or reliable access to, critical technologies. Their approach recognizes international cooperation and the functional context in which dependence becomes consequential [9]. Selectivity, access, and specialization therefore enter this paper as established considerations. The Sovereign Capability Stack organizes them around AI and autonomous systems, including their physical inputs and lifecycle requirements.
The adjacent literature on digital sovereignty also shows that sovereignty language can carry different political objectives. Pohle and Thiel trace the concept across competing claims concerning state authority, individual autonomy, democratic control, economic competitiveness, and dependence on foreign digital infrastructure [10]. This variety is relevant here because greater national technological capability should not be assumed to produce the same institutional or normative outcome across political systems. The present framework uses sovereignty primarily to describe practical control over consequential capabilities; questions about how that control is constrained and exercised are addressed separately in Section 10.
Bradford's analysis of competing American, Chinese, and European approaches to digital regulation provides a related account of how states project different models of governance through large technology markets and through their relationships with digital firms [11]. Her framework concerns regulatory power rather than the productive-capability mechanism developed here, but it reinforces the importance of jurisdiction, institutional choice, and the relationship between states and technology firms.
The proposed contribution is a linked explanation of technological change and international organization. AI may alter the relative feasibility of maintaining alternative capability; governments attach different values to external exposure; suppliers and other states respond to their decisions; and repeated adjustments can become embedded in procurement, trade, infrastructure, and alliances. The autonomy dividend captures the first mechanism, while the dependency premium describes the exposure against which alternatives are assessed. Their interaction may reorganize the terms of interdependence across countries.
The account has a defined scope. Autonomous technology can strengthen established production centers as well as improve alternatives, and public investment contributes to sovereign capability only where the resulting arrangements are usable. An order-level transition requires these effects to influence persistent practices beyond established security sectors. Sovereign-AI initiatives provide cases for investigation; their implementation and interaction determine whether they support the broader explanation.
1.2 The changing economic calculation
International provision is attractive when it supplies the required function at lower overall cost or higher performance than a more controllable arrangement. Access to several foreign suppliers may also provide greater resilience than reliance on a small domestic industrial base. Broad fragmentation can sacrifice both efficiency and access to alternatives [2, 12].
The comparison changes when a dependency becomes harder to replace or when the consequences of interruption increase. A platform supporting routine work can become consequential after essential public services are built around it. Industry consolidation can leave fewer alternatives. Export restrictions or commercial withdrawal can alter access even when the underlying service has become less expensive.
AI affects the other side of the comparison when it changes the labor, expertise or scale needed to operate an alternative. Software can extend the work of a small analytical team; industrial automation can reduce the importance of wage differences; and autonomous monitoring can change staffing requirements for remote infrastructure. The effect must be assessed against external suppliers that adopt the same technologies and may benefit more from them.
The autonomy dividend is an improvement in the relative practical or economic feasibility of maintaining domestic, regional, or otherwise trusted capability attributable to AI and autonomous systems. It can be absent or negative. The dependency premium is the additional economic and strategic exposure associated with reliance on a capability once the consequences of interruption and the difficulty of replacement are considered. Section 5 uses both concepts to structure a qualitative comparison of policy options. The concepts carry no calibrated scores or monetary valuations.
1.3 The central argument
The New Sovereignist Order describes a possible international configuration in which states and regional blocs assign increasing value to control over capabilities affecting economic continuity and political freedom of action. Control may be obtained through domestic production, diversified international access, shared infrastructure, or credible substitution arrangements.
Trust is used here to describe confidence that access to a specified capability will remain available during political or economic stress. Trusted arrangements are those in which legal commitments, institutional relationships, supply diversity, and prior performance provide credible grounds for that expectation. Their reliability is assessed against the disruption concerned and can change over time. Political alignment alone provides insufficient assurance.
The resulting economy could remain extensively globalized. Governments would distinguish among dependencies according to their consequences and the cost of managing them. The European Commission's 2026 technological-sovereignty communication illustrates a policy agenda connecting semiconductors, cloud and AI infrastructure, open source, and energy [13]. The analysis here extends beyond digital infrastructure to physical production, labor mobility, and the distribution of the capabilities through which AI-generated gains become available.
Productivity growth may be substantial while access remains unequal. Countries able to combine the required infrastructure and institutions may gain more control over the benefits of automation than countries dependent on externally supplied systems. Where confidence in continued access is limited, preferential access for domestic users or partners becomes one possible response. This distributional claim remains conditional on both technological performance and political choices.
1.4 Scope and evidentiary approach
This is a conceptual policy paper supported by existing research, institutional reports, and selected policy examples. It offers a framework for investigation rather than an estimated model, original country ranking, or demonstration that a new order has already consolidated. Statistics describe specific technologies and periods; announcements illustrate policy intentions unless implementation is documented.
The analysis concentrates on economic and technological sovereignty. Military and monetary systems intersect with it through procurement, payments, and sanctions, but their distinctive institutions receive limited treatment. Conclusions about civilian technology should not be transferred automatically to military command or monetary authority. Including those domains more fully would require additional analysis of deterrence, public finance, and legal powers.
Physical resources, capital, engineering, and institutional capacity continue to constrain autonomous systems. Smaller states often obtain more capability through international relationships than they could sustain independently. Diversification can improve resilience, while generalized localization can impose costs without consistent security benefits [12]. These constraints are examined throughout the paper rather than treated as temporary obstacles that AI necessarily removes.
1.5 When sovereignization becomes an international order
International orders are sustained through rules, institutions, and expectations that organize relationships among states. Bull’s account of international society explains how common rules and institutions can operate in the absence of world government [14]. For the present analysis, national initiatives become relevant to order when their repeated interaction changes the practices governing relationships across countries.
The terminology has described established arrangements as well as projects for change. Accounts of the postwar liberal international order emphasize economic openness, multilateral institutions, and cooperation supported by the United States and its partners [15]. The New International Economic Order declaration adopted by the UN General Assembly on 1 May 1974 proposed changes in economic relations, including control over natural resources and improved access to technology for developing countries [16]. In September 1990, George H. W. Bush used “new world order” prospectively to describe opportunities for international cooperation as Cold War confrontation receded [17]. These usages serve different purposes. Evidence of implementation and durability is needed to distinguish an established configuration from an announced political project.
The New Sovereignist Order is proposed prospectively as an analytical account of changing conditions of acceptable dependence. Economic-security and technological-sovereignty policy already examines which dependencies can be managed through diversification, reliable external access, or additional domestic capability [4, 13]. The claim examined here concerns whether AI and autonomous systems change the feasibility of those responses sufficiently to help reorganize international relationships.
An order-level transition becomes visible when these choices persist and interact. Dependency analysis enters routine economic policy; technology partnerships acquire durable access rules; and procurement and investment incorporate continuity, jurisdiction, interoperability, and control. Restrictions can stimulate substitution and countermeasures, while autonomous systems change the feasibility of selected alternatives. The resulting practices must affect actual infrastructure, investment, and behavior to support an order-level claim.
Such a system can remain highly interconnected. Trade, capital, technology, knowledge, and people can continue crossing borders while states become more selective about the dependencies embedded within those exchanges. The New Sovereignist Order therefore describes a potential reorganization of interdependence around sovereignty, optionality, and differentiated trust rather than a retreat into national isolation.
Section 8 examines whether these developments are becoming sufficiently persistent and widespread to support an order-level claim.
1.6 A note on the term “sovereignist”
Sovereignist already carries meanings in political debate and scholarship. Work on new sovereigntism, for example, examines resistance to international legal authority and protection of domestic legal orders [18]. Here the term describes an orientation toward preserving practical control over capabilities that affect a state's capacity to act. It does not designate a political party family or presume support for protectionism, immigration restriction, or withdrawal from international institutions.
Domestic political movements can reinforce demands for national control. Carnegie’s analysis of populist middle powers also finds divergent economic agendas and obstacles to durable cooperation among such movements [19]. In this framework, security, resilience, and economic-policy concerns can produce similar preferences for control across political orientations.
A sovereignist technology strategy can rely extensively on cooperation. What matters within this framework is whether institutions retain usable choices when a supplier, technology, or international relationship changes.
2. The Economic Foundations of Sovereignty
The exercise of sovereign authority requires revenue, administration, infrastructure, information, and people capable of implementing decisions. Many additional functions on which public authority depends are produced outside government, including electricity, transportation, finance, food, and technical services. Their reliability affects what the state can accomplish.
Political economy has long examined these foundations. Besley and Persson describe legal and fiscal capacity as accumulated institutional investments influencing what governments can regulate, finance, and implement [20]. Recent World Bank work examines institutional capacity at the organizational level, through the ability of ministries and agencies to carry out their mandates [21]. AI and autonomous systems affect this relationship when they change the resources required to perform a function or the dependencies through which it is delivered.
2.1 Formal sovereignty and effective sovereignty
Formal sovereignty establishes the recognized authority of the state. Effective sovereignty, as used here, concerns its practical ability to exercise lawful authority and sustain essential functions under ordinary conditions and disruption. The two can diverge. A government can possess jurisdiction over an essential service while lacking the means to maintain it after the loss of a supplier.
Fiscal, legal, and administrative capacity remain central. Technological dependence adds constraints when implementation requires access to computing infrastructure, proprietary systems, industrial equipment, or external expertise. A national AI strategy has limited practical force if the organizations responsible for it lack sufficient electricity, data, engineering, or operating capability.
Michael Mann's concept of infrastructural power directs attention to the state's ability to implement decisions through society [22]. Contemporary applications, including financial governance, examine how public power depends on the infrastructure through which it operates [23]. The present analysis extends that concern to systems assembled across private firms, international suppliers, and autonomous technologies. Stress can reveal constraints that ordinary operation conceals.
2.2 Sovereignty as productive capacity
National capability can be supplied by public institutions, private firms, or international arrangements. Ownership is consequently an incomplete guide to effective control. A privately operated data center can support public capability, while a state-owned facility can remain unable to function without an unavailable external component.
The relevant assessment concerns the system supporting the function. Location, jurisdiction, technical knowledge, maintenance, supply diversity, and replacement time all matter. For example, a factory may be domestically owned and operated but depend on imported machinery and remote diagnostics. Its contribution to sovereignty depends partly on whether production can continue when those services fail.
This approach also avoids attributing governmental control to every commercially successful national firm. A firm may contribute important capacity without being subject to government direction over its commercial decisions. Access arrangements and legal powers determine what public institutions can rely on in a crisis.
2.3 Globalization and the externalization of capability
International exchange separates access to capability from the need to reproduce every input nationally. Countries can use equipment, finance, professional expertise, and manufactured goods produced through systems located elsewhere. Global value chains extend this arrangement by distributing different stages of production among specialized locations.
Externalization can strengthen resilience when it provides alternatives to domestic constraints. A country purchasing from several independent suppliers may be less vulnerable than one dependent on a single national producer. OECD supply-chain analysis finds that broad localization can impose substantial costs without consistently reducing volatility [12]. The relevant issue is the reliability and substitutability of access, rather than whether trade crosses a border.
AI may make selected alternatives more feasible, but international provision remains the appropriate comparison. The paper therefore examines changes in the cost of restructuring dependence rather than assuming that the historical externalization of capability will generally reverse.
2.4 Dependence, concentration, and substitutability
External dependence becomes consequential when interruption causes substantial harm and alternatives cannot be activated quickly. The OECD's economic-security framework emphasizes criticality, concentration, and difficulty of diversification or substitution [4]. These characteristics distinguish a vulnerability from an ordinary commercial relationship.
Consider two countries importing the same share of an industrial input. One can switch among compatible suppliers within months. The other must redesign equipment and retrain personnel before replacing its sole supplier. Similar import shares conceal different constraints. In digital services, dependence can deepen when applications, data, and employee skills become specific to one platform; cloud-switching rules address some of these barriers [24].
Replacement time is especially important for essential functions. An alternative available in three years provides little continuity for a service that must resume next week. Concentration can also exist domestically: jurisdiction over a national monopoly does not create an alternative technology or workforce. Operational options must be assessed under the particular disruption that policy is intended to address.
2.5 Operational sovereignty
Operational sovereignty is the capacity to preserve an essential function when an important external dependency becomes unavailable, restricted, or unreliable. It may be secured through domestic capability, diversified suppliers, reserves, regional arrangements, or interoperable systems. The appropriate combination depends on the function and on the duration and consequences of interruption.
Governments have little reason to maintain costly substitutes for every imported product. Additional redundancy becomes easier to justify for functions such as electricity, emergency communications, payments, healthcare, and critical computing. This prioritization is consistent with economic-security analysis linking supply chains and infrastructure to national capacity [4]. The alternatives carry their own risks. Domestic facilities can fail; reserves can become obsolete; institutions can lack the skills to activate a backup; and protected firms can receive support unrelated to a genuine vulnerability. Public investment should therefore identify the service to be preserved and compare the costs of credible options.
AI and autonomous systems can change those costs where labor, expertise, or continuous staffing has been a binding constraint. Foreign producers face the same technological opportunities. The next section examines how the resulting change in productive requirements can affect the location and organization of capability.
3. How AI Changes Productive Requirements
AI affects sovereignty when it changes the inputs required to sustain a capability. Software can extend analytical and administrative work, while robotics connects computational capability to physical action. The effect develops unevenly as reliability, deployment costs, supporting infrastructure, and organizational adaptation determine what institutions can use.
Automation can improve a domestic alternative or reinforce an established foreign supplier. It can extend local access to expertise while increasing reliance on remote computing services. Assessment therefore follows the function being maintained and its full supporting requirements.
3.1 AI as a productive input
AI can increase the productivity of people who continue performing a function or carry out tasks previously requiring direct human execution. Agentic systems can connect tasks through information retrieval, tool use, and sequences of actions within defined permissions. These arrangements can expand capacity through augmentation or partial automation while retaining a human role in the underlying function.
A ministry may allow analysts to review more evidence without eliminating the need for domain expertise. A software team may maintain a larger system with similar staffing. Such changes can lower the minimum organizational scale at which a function becomes viable. They can be valuable where a small institution cannot sustain a specialist team for every intermittent need.
Realizing these gains can require complementary investment in data, organizational processes, integration, evaluation, security, and institutional competence, together with time for experimentation and adaptation. Human verification and exception handling also carry costs. These requirements belong in any assessment of the autonomy dividend.
3.2 AI-supported expertise
Software can support document analysis, coding, translation, compliance review, scheduling, and monitoring with less continuous human execution. Its sovereignty relevance is greatest where scarce expertise has constrained an institution’s ability to sustain a function.
AI-supported staff may extend technical or analytical capacity across more organizations and locations. Local-language resources can improve access to information. These mechanisms can reduce some advantages associated with geographic concentrations of expertise, but the infrastructure supplying the capability may remain concentrated elsewhere. Access through a foreign model or cloud platform can improve performance without creating an independently maintainable national capability.
The distinction between consuming intelligence and sustaining its production is central to the Sovereign Capability Stack. It also requires comparison with external organizations using AI to improve the productivity of their own, potentially larger specialist communities.
3.3 Industrial automation
Industrial automation provides existing evidence of changes in labor requirements and production geography. The International Federation of Robotics recorded 542,000 industrial robot installations in 2024 and approximately 4.66 million robots operating worldwide by year end. Asia accounted for 74 percent of installations and China for 54 percent [25]. Adoption is therefore strongest in many locations that already possess considerable industrial capability. Krenz, Prettner, and Strulik find that robot adoption can contribute to reshoring where automation narrows labor-cost disadvantages. Their model also suggests that returning production need not restore the low-skilled employment associated with the earlier industrial model [26]. Artuc, Bastos, and Rijkers find that robotization can increase international trade through changes in comparative advantage [27]. Together, these studies support examining relative effects rather than treating automation as uniformly localizing. Section 6 applies that comparison to trade and production.
3.4 Automation in physical environments
Robotics becomes more consequential to sovereignty as machines operate beyond fixed production environments. Specialized systems in logistics, inspection, agriculture, and services combine perception, movement, and software with varying degrees of adaptation. Technical feasibility and commercial adoption remain distinct; a successful demonstration may not establish reliable operation across an entire sector.
Humanoid machines receive attention because they could operate in spaces designed for people, but their commercial base remains limited. Reuters reported IFR estimates of approximately 7,000 humanoid sales in 2025 for industrial and professional service applications, many to research institutions or companies developing AI rather than directly replacing routine labor [28]. IFR separately recorded 542,000 industrial installations in 2024 [25]. Its service-robot survey recorded nearly 200,000 professional-use sales among 294 suppliers for that year, a sample rather than a whole-industry estimate [29]. These categories and years should not be treated as a single directly comparable market census.
The sovereignty mechanism does not require general-purpose humanoids. Purpose-built machines can address tasks in warehouses, factories, farms, or infrastructure where the environment and process can be designed together. Their significance appears where economically reliable operation reduces a constraint that previously required workers or specialists to be present locally.
3.5 From analysis to action
Combining analysis with physical execution can change the organization of production. A system may identify a defect and direct machinery to remove a component, or plan inventory movement and dispatch equipment to carry it out. These are illustrative operating arrangements; the degree of useful autonomy depends on task reliability, safety, and integration.
Human work remains necessary in engineering, maintenance, exception handling, supervision, and organizational responsibility. Its quantity and location can change. An installation requiring occasional technical visits rather than continuous staffing may become feasible in a different location. A smaller team may manage a larger fleet. Whether either arrangement reduces overall cost depends on capital, maintenance, communications, and the alternatives available to the operator.
The constraint that automation relaxes matters more than a generic measure of technological sophistication. Reducing routine labor can be valuable in a high-wage setting; extending expertise can matter in a small administration; autonomous monitoring can affect geographically dispersed infrastructure. Later sections examine these mechanisms without assuming that gains in one task translate into economy-wide autonomy.
3.6 Relative effects and binding constraints
The external alternative changes alongside domestic capability. An established supplier may capture greater gains from automation because it has better infrastructure, scale, financing, or engineering. Technological progress contributes to an autonomy dividend only where the relative feasibility of a more controllable arrangement improves.
This supports directional, conditional hypotheses. Positive dividends are more plausible where automatable labor or staffing costs are a major domestic disadvantage and complementary inputs are already available. Dividends may be weak or negative where scarce advanced hardware, energy, capital, or industrial knowledge remain binding and incumbents can expand faster. These are propositions for sectoral comparison, not established classifications of industries or countries.
A production system can also gain control at one layer while losing it elsewhere. Domestic automation may reduce labor requirements while making continuity more dependent on imported processors, minerals, or manufacturer services. The full dependency structure must be considered before interpreting a productivity gain as greater sovereignty.
3.7 Why physical AI matters to national capability
Physical AI extends automation into functions that supply food, manufacture equipment, move goods, and maintain infrastructure. These activities have historically depended on combinations of local labor, industrial ecosystems, and imported production. Changes in their staffing requirements can affect both offshoring and migration, even where total economic activity and employment continue to expand.
Autonomous production remains dependent on physical and institutional inputs. Energy, materials, semiconductors, software, networks, maintenance, and technical expertise determine whether a deployed system can continue functioning. The following section maps these dependencies through the Sovereign Capability Stack.
4. The Sovereign Capability Stack
Technological capability is assembled across resources and organizations that may fall under different forms of control. A national model depends on computing infrastructure; compute depends on hardware, electricity, cooling, communications, and software; autonomous machines add components, maintenance, and operating services. Human talent and institutions determine whether those resources can be integrated and sustained.
The Sovereign Capability Stack maps these dependencies, including the energy and material inputs beneath digital systems. It is a diagnostic framework for identifying critical constraints and alternatives. It does not prescribe domestic ownership of every layer. A function may be more resilient through diversified international access than through a smaller national production system.
The European Commission's 2026 technological-sovereignty communication connects semiconductors, cloud and AI infrastructure, open source, and energy [13]. The framework here extends that analysis to strategic materials, national knowledge resources, trust infrastructure, physical machines, and the institutional capacity required throughout the lifecycle.
4.1 Energy and electrical infrastructure
Computing infrastructure and autonomous production depend on reliable electricity. The International Energy Agency estimates that data centers consumed approximately 485 TWh globally in 2025 and projects around 950 TWh by 2030 under its central scenario. It also identifies grid connections, transformers, generation capacity, and other equipment as constraints on deployment [30]. These figures describe data-center electricity use rather than the energy requirements of the entire autonomous economy.
Purchased processors and completed buildings provide limited capacity where electricity or grid access is unavailable. Automated factories, telecommunications, and robot fleets create additional requirements for continuity, charging, and recovery. Assessment at this layer should examine available generation, grid capacity, fuel exposure, storage, and the time needed to expand or restore supply.
Regional electricity markets and cross-border infrastructure can provide useful redundancy when access is diversified and reliable. AI can also improve forecasting, maintenance, and the operation of energy networks [30]. The relationship runs in both directions: energy enables autonomous capability, while well-governed automation can improve the reliability and productivity of energy systems.
4.2 Critical minerals and strategic materials
The physical foundations of AI include materials used in semiconductors, electrical equipment, batteries, communications networks, and industrial machines. The requirements differ by technology and design. A battery chemistry, a permanent-magnet motor, and an accelerator do not depend on an identical basket of minerals. Sovereignty analysis should trace the materials needed for the specific function rather than treat critical minerals as a single interchangeable input.
The IEA's Global Critical Minerals Outlook 2025 documents growing concentration in key refining markets between 2020 and 2024 and examines the consequences of supply disruption and expanding export restrictions [31]. Access to ore alone provides limited assurance where processing, purification, or component production is concentrated elsewhere. A material representing a small share of purchase cost can still become a binding constraint if no qualified substitute is available.
China's July 2023 announcement of controls on gallium- and germanium-related items illustrates how materials access can enter economic-security policy. The Ministry of Commerce described licensing controls justified by national-security and national-interest considerations, rather than an unconditional prohibition on all exports [32]. This historical example identifies an instrument of control; it does not establish the current licensing position for any particular product or destination.
For autonomous infrastructure, exposure can extend across the operating life of the asset. An installation may continue functioning with existing components during a supply interruption but face constraints when equipment fails or capacity must expand. Inventories, repairability, alternative components, and supplier qualification can therefore be as relevant as the location of primary production.
Automation may improve the productivity of mining, processing, or recycling, but its sovereignty effect remains comparative. It can strengthen a diversified alternative or reinforce the efficiency of an established processing center. Changes in chemistry, material efficiency, recycling, and product design can reduce particular dependencies, although their feasibility must be assessed over the time available for substitution.
The appropriate response will vary by material and application. Domestic extraction may be unnecessary where supply is diverse and substitutes are practical. Conversely, a processing bottleneck can justify investment or coordinated access even when the raw resource is widely distributed. Environmental requirements, infrastructure, permitting, skills, and financing form part of the cost of establishing an alternative. This layer connects physical autonomy to resource geography and prevents reduced labor dependence from being mistaken for material self-sufficiency.
4.3 Semiconductors and advanced hardware
The semiconductor value chain combines design software, intellectual property, manufacturing equipment, materials, fabrication, memory, packaging, assembly, and testing. OECD analysis identifies substantial geographic concentration and cross-border dependencies across several of these stages [33]. Domestic access to one stage consequently leaves exposure to other specialized inputs.
Countries may secure capacity through selected domestic production, diversified suppliers, inventories, long-term arrangements, or participation in multinational production networks. The relevant question is whether those arrangements sustain the required function during interruption and over successive hardware generations.
Automation can change the cost of producing some hardware, but established suppliers also benefit from it. Semiconductor manufacturing therefore provides an important limit to the assumption that autonomous technology makes replication straightforward. Capital, equipment, process knowledge, and supplier ecosystems can remain binding constraints even where direct labor requirements fall.
4.4 Compute capacity
Compute supports model development and adaptation, inference, scientific research, simulation, and industrial applications. The required capacity varies substantially by task. Training a frontier model and operating a smaller application create different requirements for scale, security, and continuity.
Installed processors provide only a partial measure. Institutions also need to know who controls allocation, which software governs access, what happens during scarcity, and how workloads can move to other infrastructure. Replacement hardware, cooling, networking, maintenance, and electricity affect whether nominal capacity remains usable.
External compute can support operational sovereignty when access is diversified and workloads are portable. A small domestic facility without a backup may provide less continuity than several reliable external options. The relevant assessment compares the computational function that each arrangement can sustain under specified disruption scenarios.
4.5 Cloud, communications, and data-center infrastructure
This layer connects machines, users, applications, and data through facilities, fiber, mobile networks, internet exchange points, submarine cables, satellites, and cloud-management software. The ITU reports that submarine telecommunications cables carry more than 99 percent of international data traffic and identifies their resilience as a critical-infrastructure concern [34]. The statistic concerns international traffic rather than all internet use.
A country can gain capability through global connectivity while becoming exposed to concentration in routes, landing points, operators, or external networks. Cloud infrastructure creates related dependencies where local facilities rely on remote administration, authentication, proprietary management systems, or external technical teams. The Commission's cloud-sovereignty framework examines these operational and jurisdictional characteristics [35].
Continuity planning should establish which functions remain available during degraded connectivity or loss of remote services. Physical location may improve latency, security, or legal oversight, but operation depends on the full network and management architecture. Section 9 examines the resulting gap between local presence and practical control.
4.6 Data and national knowledge resources
AI's usefulness depends partly on the information available in the environment where it is deployed. Legislation, administrative records, industrial information, scientific data, languages, geospatial resources, and public statistics can determine whether a system provides reliable local capability.
Data governance involves lawful access, quality, metadata, interoperability, preservation, and protection of sensitive information. Fragmented or unusable data can limit national AI capacity even where models and compute are available. The World Bank's AI-foundations assessment emphasizes context and relevant data alongside connectivity, compute, and competency [36].
Local-language datasets and evaluation resources can improve the usefulness of systems developed elsewhere. This offers a way to build capability without reproducing frontier-model development [36]. Storage location remains relevant to particular risks, but the productive value of national knowledge also depends on whether institutions can lawfully retrieve, interpret, and use it across systems.
4.7 Models, software, and digital platforms
Institutions can develop models, adapt open models, use proprietary APIs, or purchase applications whose underlying model is largely invisible. Each arrangement distributes dependencies differently. API access can reduce internal operating requirements while leaving availability, pricing, updates, and aspects of technical control with the supplier. Locally operated models provide greater control over some of these functions but require compute, engineering, security, and maintenance.
Open models illustrate how sovereignty can shift between layers of the capability stack. Running an open model locally can reduce dependence on an external model provider and provide greater control over deployment, modification, and continuity. It can simultaneously increase the institution's direct dependence on computing infrastructure, accelerators, electricity, cooling, technical expertise, and the software required to operate the model. If those resources remain externally controlled or difficult to substitute, greater sovereignty at the model layer may simply relocate dependency to compute, hardware, or energy.
The surrounding software can be equally consequential. Databases, operating systems, development tools, orchestration, identity services, and cloud platforms affect integration and replacement. Open technologies can preserve options, and the European Commission explicitly links open source with dependency reduction [13, 37]. Their practical value depends on whether organizations possess the skills, infrastructure, documentation, and resources required to operate and maintain them.
Model access and model independence should therefore be assessed separately. The ability to download model weights or change an API endpoint does not establish that an institution can preserve the underlying function after a transition. Effective sovereignty depends on the dependencies required to operate the alternative across the wider capability stack.
4.8 Cybersecurity, identity, and trust infrastructure
Autonomous systems require reliable ways to establish which actor is operating and what it is permitted to do. Software agents may access databases, use tools, or initiate transactions; machines can move through facilities and operate equipment. Delegating such actions requires identity, authentication, authorization, revocation, and evidence of the authority exercised.
NIST's AI Agent Standards Initiative identifies secure interoperability, identity, and authorization as areas requiring coordinated work [38]. In physical systems, comparable questions arise when machines cross organizational boundaries or change operators. These are proposed requirements for trustworthy operation, rather than evidence that every market already uses a common machine-identity standard.
Security and trust infrastructure can itself become a dependency. Essential functions may rely on external certificate services, authentication platforms, or security tools. Institutions need to know whether authority can still be verified, restricted, and withdrawn if those services fail or become inaccessible.
4.9 Robotics and autonomous physical systems
Robotics converts computational capability into physical action in factories, warehouses, farms, transportation, inspection, and services. Its contribution to sovereignty depends on the function performed and the resources needed to sustain it. A machine's physical presence provides limited information about the software and service relationships on which it relies.
Potential dependencies include firmware, activation, fleet management, connectivity, specialized components, diagnostics, maps, credentials, safety assurance, and updates. Their extent varies among products. A deployment should therefore be evaluated over its expected operating life, including maintenance, transfer, and supplier exit.
Morgan Stanley’s 2025 Humanoid 100 offers a selected investment-research mapping of this industrial ecosystem. It distinguishes a computational “brain” of models, software, and semiconductors; a physical “body” of sensors, batteries, actuators, and other components, including rare earths and magnets; and firms integrating complete humanoid systems [39]. This mapping illustrates the breadth of upstream inputs. It does not measure the vulnerability of a particular deployment or establish independence at any layer.
This lifecycle perspective becomes consequential when fleets support essential functions. Procurement can identify which services must remain available locally, which data and records must be transferable, and what alternative arrangements exist if manufacturer support ends. Section 9 develops the manufacturer-dependence problem; Section 11 considers the corresponding procurement and standards requirements.
4.10 Human talent and institutional capacity
Engineering, operations, maintenance, security, procurement, regulation, and management support every technical layer. Equipment can often be purchased faster than institutions can develop the knowledge required to operate and govern it. This can leave apparently substantial national assets dependent on a small external technical community.
AI may extend scarce expertise while increasing the importance of people able to integrate, evaluate, repair, and supervise systems. International talent and research cooperation can contribute directly to that capacity; OECD work examines the scarcity and mobility of specialized AI skills [40]. Domestic education and organizational learning determine whether access develops into durable competence.
Human and institutional capacity is therefore a cross-cutting condition. It governs whether alternatives can actually be activated and whether responsibility remains clear after technology becomes embedded in an essential function.
4.11 Assessing interdependence across the stack
A stack assessment should begin with the function to preserve, identify consequential dependencies at each layer, and examine whether proposed alternatives share the same upstream constraint. Two suppliers provide limited diversification if both require the same unavailable component, management platform, or processing facility.
The framework does not require equal strength everywhere. It identifies where a weakness can interrupt the function and which intervention addresses that weakness at acceptable cost. Improving one layer can also increase requirements elsewhere: local compute adds energy demand, and locally operated software requires maintenance capability.
This provides the link to the comparison of sovereign-capability options in Section 5. A technically feasible alternative becomes strategically useful only if its supporting dependencies remain manageable. The following section examines current exposure alongside the performance, resources, and dependencies of more controllable arrangements.
5. The Changing Cost of Sovereign Capability
Greater control over a technological capability requires resources. Domestic production, backup infrastructure, alternative suppliers, and transition arrangements can cost more than relying on an established external provider. Governments must consider what additional continuity or freedom of action these investments would provide and whether the same objectives could be achieved through less costly arrangements.
AI and autonomous systems can change the alternatives available. Automation may make a previously labor-intensive activity feasible with a smaller workforce. AI-supported expertise can help an institution maintain functions that previously required larger specialist teams. These opportunities also extend to foreign suppliers, whose existing advantages may become stronger as they automate.
This section develops a qualitative framework for comparing such choices. Two analytical concepts, the dependency premium and the autonomy dividend, organize the relationship between the exposure created by dependence and the changing feasibility of alternatives. Their measurement requires sector-specific evidence about costs, performance, dependencies, and implementation. The framework provides a basis for that investigation without assigning a monetary value to sovereignty.
5.1 Sovereignty has a cost
International specialization gives countries access to capabilities supported by accumulated expertise, established infrastructure, large markets, and specialized suppliers. Importing equipment avoids the expense of reproducing its entire research and manufacturing ecosystem. Commercial computing services allow users to share infrastructure costs that would otherwise have to be borne internally.
These advantages remain relevant when governments become concerned about dependence. OECD supply-chain analysis finds that generalized localization can reduce trade and output without consistently improving resilience. Diversified international networks can provide alternatives that a smaller domestic system lacks [12].
The starting point for assessing sovereign capability should therefore include the value of existing external provision. Additional control may justify paying more, particularly where interruption would affect an essential function. The comparison must also account for the resources committed to the alternative, its operating requirements, and the vulnerabilities it introduces. A domestic facility that depends on unavailable components may provide less continuity than its location suggests.
5.2 The dependency premium
The dependency premium refers to the additional economic and strategic exposure associated with reliance on a capability once the consequences of interruption and the difficulty of replacement are considered. The term captures considerations that may be poorly reflected in the ordinary purchase price of a product or service.
Five characteristics help organize the assessment. Criticality concerns the harm caused by losing the capability. Concentration concerns whether supply or control rests with a small number of genuinely independent actors. Substitutability concerns the availability of another technology, supplier, or process that can preserve the function. Recovery time concerns how long that transition would take. Exposure to restriction concerns access that can be altered through political, legal, commercial, or technical decisions. These considerations are consistent with the dependency concerns examined in OECD economic-security analysis [4].
Their significance depends on the particular function and arrangement. A cloud application that can move between providers within an acceptable period presents a different exposure from an essential government system requiring extensive redesign before migration. For autonomous machinery, access to parts, maintenance information, software, and technical personnel can determine whether continued operation is possible.
The assessment should examine how these characteristics interact. Several nominal suppliers may depend on the same infrastructure or jurisdiction. A replacement may exist commercially but remain unusable during the disruption being considered. Where reliable probabilities are unavailable, explicit scenarios can still identify which functions would fail, what alternatives could be activated, and how long recovery would take.
5.3 The autonomy dividend
The autonomy dividend refers to an improvement in the relative practical or economic feasibility of maintaining a capability under domestic, regional, or otherwise trusted control attributable to AI and autonomous systems. It concerns the institution’s practical options, assessed separately from the degree of discretion delegated to a machine.
The comparison is with the external alternative, which can also improve. A positive dividend arises where technological change narrows a disadvantage or relaxes a constraint that previously made the more controllable arrangement difficult to sustain. Lower domestic operating costs alone are insufficient evidence if an external supplier achieves greater improvements.
Several developments can create such an opportunity. Robotics can reduce reliance on difficult-to-recruit workers. AI can extend the output of scarce specialists. Autonomous monitoring can reduce requirements for continuous staffing at dispersed installations. More capable software can make selected functions manageable within smaller organizations.
These changes can expand the choices available to a government or institution without making complete domestic production desirable. A shared regional operation may become feasible at a scale that participating countries can support. An institution may acquire the ability to operate a model locally while continuing to use external services for other workloads.
The dividend concerns the improvement in that alternative. Whether adopting it strengthens sovereignty depends on the full supporting system. As Section 4 explains, gaining control at the model layer can increase direct requirements for compute, electricity, hardware, and technical operations.
5.4 Relative feasibility matters
A useful comparison must specify the function to be preserved, the required performance, and the period over which continuity matters. Two arrangements that differ substantially in reliability, safety, output, or service quality cannot be treated as equivalent merely because one is cheaper.
The assessment should consider the resources required throughout the operating life of the alternative. Acquisition is only part of the commitment. Integration, energy, maintenance, security, evaluation, training, and eventual replacement can determine whether a capability remains usable. Public support also requires careful treatment: a subsidy may reduce the price paid by the purchaser while leaving the underlying resource requirements unchanged.
Consider, as an illustration, a logistics activity that can be served by a domestic distribution facility or an external provider. Robotic handling may allow the domestic facility to maintain throughput with fewer workers, easing a recruitment constraint. The comparison still needs to include equipment, software, technicians, electricity, maintenance, and the cost of delivering an equivalent service. It must also account for automation by the external provider.
If the domestic option becomes more feasible but relies on indispensable remote manufacturer services, the assessment must examine that new dependency alongside the reduction in labor requirements. The example identifies questions for investigation; it supplies no measured result.
Attributing an improvement to autonomy also requires separating technological effects from changes in wages, energy prices, tariffs, exchange rates, or public support. This is where sector-specific research would be needed to establish the magnitude and direction of an autonomy dividend.
5.5 Comparing policy options
The dependency premium and autonomy dividend help distinguish situations that warrant different responses. Table 1 organizes these comparisons qualitatively. Its categories refer to a specified capability and alternative arrangement, rather than fixed characteristics of an entire country or industry.
| Dependency premium | Weak, absent, or negative autonomy dividend | Material positive autonomy dividend |
|---|---|---|
| High | Compare diversification, inventories, shared capacity, and negotiated access. Expensive domestic provision may still be justified where the consequences of interruption are severe. | Examine whether improved alternatives provide meaningful additional control or continuity. Compare domestic, regional, and other arrangements against their full operating requirements. |
| Low | International sourcing may remain appropriate. Monitor concentration and the practicality of switching. | Commercial localization may become attractive. Any public support requires a separate justification. |
A consequential dependency combined with a material improvement in the alternative gives policymakers stronger grounds to investigate additional controllable capacity. It leaves open which arrangement should be chosen. Diversification or a shared regional facility may still provide better continuity than national replication.
A positive autonomy dividend is also unnecessary for every sovereignty investment. Governments may accept substantial costs to preserve a function whose interruption would be intolerable. Conversely, technology may make local production commercially attractive even where the original dependency carries little strategic exposure.
The practical comparison concerns what each option would preserve under the relevant disruption. A second supplier is useful only if it remains available when the first fails. Domestic hosting provides limited protection against a failure affecting essential foreign software. The dependency map developed through the Sovereign Capability Stack should therefore inform the comparison.
5.6 Optionality as a policy objective
Optionality means having credible alternatives when an important technological or commercial relationship changes. It can allow an institution to maintain essential services, negotiate from a stronger position, or adjust its technology without having to rebuild the entire capability.
This approach has an affinity with real-options analysis, which examines investment under uncertainty and the value of preserving choices over time [41]. The application here remains qualitative. A staged investment, maintainable backup, or tested migration arrangement may preserve useful choices before a government commits to a more expensive domestic alternative.
Those options require continuing attention. Documentation becomes outdated, technical staff leave, and substitute systems can cease to meet operational requirements. A portability clause offers limited practical value if no one has tested whether the application can be transferred. Periodic exercises can reveal whether alternatives remain usable and whether the investment continues to address the original exposure.
The policy objective is an improvement in usable freedom of action, assessed against the resources required to obtain and maintain it.
5.7 Limits and trade-offs
Sovereignty objectives do not remove the risks of public investment. Governments can overestimate the importance of a dependency, underestimate implementation costs, or support an incumbent whose performance remains weak. Industrial policy can create fiscal burdens and become vulnerable to political capture, particularly when strategic claims substitute for clear objectives and evaluation [42].
Domestic capability can also introduce concentration. Replacing several international suppliers with one national provider may improve jurisdictional control while reducing technical choice. A protected technology can become costly to replace after institutions organize around it. These possibilities require the proposed alternative to be examined as carefully as the existing dependency.
Assessment should also distinguish different policy objectives. Continuity, employment, industrial development, innovation, and national security may support the same investment, but their benefits and costs can diverge. An automated facility may strengthen productive capacity while creating relatively few jobs. Those outcomes should be evaluated separately.
A credible intervention should identify the function at risk, the disruption it addresses, the alternatives considered, and the evidence needed to justify continued support. Where information remains incomplete, limited deployment or staged investment can provide a basis for learning before larger commitments are made.
The framework developed here supports those judgments. It explains why changing technological capability can make additional control more feasible while preserving the need for evidence about costs, performance, and remaining dependencies. The following section examines how these considerations affect production, trade, and the geography of international economic activity.
6. The Reconfiguration of Globalization
The New Sovereignist Order would develop inside an international economy that remains deeply integrated. The available evidence points more clearly to reconfiguration than to a broad reversal of globalization. AI and autonomous systems can alter where production is viable and which dependencies governments consider acceptable without requiring a generalized collapse in cross-border exchange.
The resulting pattern is likely to combine continued specialization with stronger efforts to preserve alternatives around capabilities that are difficult to replace. Production can move domestically, regionally, or toward trusted partners while other activities become more globally distributed through digital services and automated coordination.
Global value chains still account for a large share of international trade. The WTO reports that GVC-related trade represented 46.3 percent of global trade in 2024, below its 2022 peak [43]. UNCTAD’s December 2025 update projected that trade in goods and services would exceed $35 trillion that year [44].
6.1 Automation and the geography of production
Production locations reflect wages, energy, capital, infrastructure, logistics, supplier ecosystems, technical skills, regulation, and proximity to customers. Automation changes the weight of these factors. Facilities using fewer production workers become less sensitive to wage differences, while AI-supported engineering, inspection, maintenance, inventory, and logistics can increase the output of smaller teams.
These changes can improve the competitiveness of higher-cost locations where labor formed an important part of the disadvantage. Established industrial clusters can also automate and preserve or deepen their advantages. The location effect therefore follows relative productivity rather than the presence of automation itself.
6.2 Reshoring and the limits of national production
Reshoring relocates production from a foreign economy to the domestic one. It can support sovereignization where the relocation improves continuity or provides credible alternatives. The research discussed in Section 3 indicates that automation can contribute to this process while returning production without restoring the same number or composition of jobs [26].
Domestic production can remain dependent on imported machinery, software, components, or materials. Replacing several external suppliers with one national producer can also increase concentration. These risks need to be evaluated alongside the cost of maintaining domestic capacity.
Reshoring is therefore one instrument for reducing a dependency premium. Its value depends on whether the relocated capability actually improves continuity and substitution under the disruption the policy is intended to address.
OECD modelling illustrates the costs of treating national localization as a general resilience strategy. Its supply-chain analysis estimates that broad relocalization could reduce global trade by more than 18 percent and global real GDP by more than 5 percent, without consistently reducing volatility [12]. The result reinforces the need to compare reshoring with diversification, inventories, interoperability, and regional alternatives rather than assume that domestic production is inherently more resilient.
6.3 Nearshoring, regionalization, and trusted production networks
Production does not have to return within national borders for states to reduce exposure. Nearshoring can shorten supply routes; regionalization can distribute production among countries connected through common rules and infrastructure; trusted networks can place strategically important activities among partners expected to preserve access during political stress.
These arrangements retain more of the economic value of specialization than comprehensive localization. They also connect directly with coalition sovereignty: shared computing infrastructure, semiconductor capacity, energy networks, or industrial production can provide greater resilience than a smaller national alternative when access rules are credible and several participants remain capable of substitution.
Current trade indicators already show elements of this reconfiguration. UNCTAD reported that nearshoring and friendshoring indicators strengthened again in the third quarter of 2025 after weakening in the first half of the year [44]. The WTO describes global value chains as being rewired through digitalization, diversification, industrial policy, and changing geopolitical conditions [43].
6.4 Autonomous production and the changing industrial model
Physical AI changes the internal organization of production as well as its location. Increasingly automated facilities combine robotic handling, machine vision, predictive maintenance, automated inspection, software-based planning, and autonomous movement of materials. The resulting model places greater weight on capital, reliable electricity, software, engineering, cybersecurity, and maintenance.
The transition will remain uneven. Structured environments and standardized objects are easier to automate than work requiring frequent adaptation to changing physical conditions. Capital-intensive sectors can support larger automation investments than industries with thin margins. Countries whose manufacturing advantage rests heavily on low-cost labor may therefore face pressure in some sectors while retaining advantages in others.
6.5 Supply chains as strategic infrastructure
Modern production networks increasingly resemble infrastructure because disruption in one specialized input can affect many downstream activities. Semiconductors support data centers, vehicles, telecommunications, machinery, and medical equipment. Cloud platforms support finance, government operations, communications, and industrial systems. Similar cascading effects arise from critical minerals, software dependencies, and specialized maintenance services.
International exposure itself is not evidence of weakness. Vulnerability arises where an essential node is concentrated and alternatives are limited. Governments therefore need enough visibility to identify dependencies whose failure would have consequences beyond an individual firm.
OECD data illustrate the continued depth of external exposure in strategic manufacturing: roughly one quarter of inputs used by strategic manufacturing industries across OECD countries are sourced from abroad, while a similar share of output depends on foreign final demand [12].
6.6 What remains global
Strong incentives for international exchange remain. Semiconductor production is distributed across specialized firms and jurisdictions. Critical minerals follow geography. Scientific research benefits from international collaboration. Advanced equipment incorporates knowledge and components from multiple countries, while technology companies recruit globally and operate across large markets.
AI can strengthen these connections. Automated translation lowers linguistic barriers, software services can be distributed globally at low marginal cost, and AI can improve coordination across international production networks. Local productive capability and deeper international integration can therefore expand at the same time.
6.7 Selective interdependence
These developments support selective interdependence: continued international exchange combined with different standards of resilience for different dependencies. Global sourcing remains appropriate where suppliers are diversified, replacement is practical, and interruption would have limited consequences. More concentrated dependencies can justify inventories, portability requirements, regional capacity, trusted production networks, or domestic alternatives.
Geographic proximity and political alignment can reduce some risks without guaranteeing continuity. The policy task is to match the degree of control to the consequences of failure.
Misclassified strategic industries, weak domestic incumbents, and fragmented markets can undermine resilience [12, 42]. The next section examines how changes in productive capability may affect international labor mobility.
7. Labor, Skills and International Mobility
Human capability is an input to sovereign capability. AI matters to international labor mobility where it changes the tasks, skills, or staffing needed to sustain a function. Family, humanitarian, educational, and other forms of migration follow additional mechanisms. The hypothesis examined here concerns selected labor requirements rather than a general decline in cross-border movement.
OECD countries recorded 6.2 million new permanent immigrants in 2024, 15 percent more than in 2019. Family-related admission accounted for 44 percent and work-related admission for 17 percent [45]. These categories do not determine subsequent labor-market participation. The figures caution against treating employer demand as an explanation of migration as a whole.
7.1 Tasks, staffing and productive capacity
The ILO estimates that one in four workers is employed in an occupation with some exposure to generative AI, with 3.3 percent of global employment in its highest exposure category. It expects job transformation to be more common than complete occupational replacement [46]. Acemoglu and Restrepo distinguish displacement from existing tasks from the creation of new tasks in which labor retains or gains a comparative advantage [47].
For sovereign capability, the relevant question is whether AI changes the human resources required to maintain a specified function at an acceptable standard. A warehouse may need fewer workers to move goods but more technicians to maintain its fleet. AI can reduce hospital administration without resolving a shortage of clinicians. Productivity gains can also expand output and create complementary work. These effects make the composition and location of labor demand more informative than the share of individual tasks automated.
7.2 Evidence from robotics and migrant labor
Studies of industrial robotics show several relationships between automation and migrant labor. Research on US local labor markets between 1990 and 2014 found larger estimated employment effects among immigrants than native-born workers, partly associated with their concentration in exposed tasks. One additional robot per thousand workers was associated with reductions of 0.38 percentage points in the native employment-to-population ratio and 0.67 percentage points among immigrants [48]. Danish employer-employee evidence for 1995–2019 found that municipalities receiving larger increases in non-Western immigrant labor subsequently experienced lower firm-level robot adoption, with stronger relationships where workers were more substitutable for robotic tasks [49].
A 2026 US study found that low-skilled immigrants responded to robot exposure partly by moving between local labor markets. It found no significant effect on international immigration itself [50]. Taken together, these findings support task-specific substitution and adjustment mechanisms. They provide limited grounds for extrapolating an economy-wide migration response to contemporary AI. Such a claim would require evidence connecting actual deployment to recruitment, vacancies, and mobility while accounting for other influences. The studies concern different countries, periods, and margins of adjustment. Their findings should be used to identify mechanisms for investigation, with current software and physical-AI deployment assessed on its own evidence.
7.3 Physical work and demographic constraints
Agriculture, construction, warehousing, transport, cleaning, hospitality, and care contain tasks requiring workers to be present where production or service occurs. Robotics changes that requirement where machines can perform enough of the work reliably and economically. Structured tasks are generally easier to automate than changing sites, variable objects, or work involving sustained interpersonal contact.
The comparison follows Section 5: equivalent function and acceptable performance, lifecycle and transition costs, maintenance, energy, and remaining human work. Available technical support can be as important as equipment prices. Institutions may accept additional costs where continuity warrants them, but a demonstration supplies limited evidence of a maintainable alternative.
Demographic demand can offset substitution. OECD data for 2020–21 identify approximately 830,000 foreign-born doctors and 1.75 million foreign-born nurses across countries with available data, around 22 percent and 17 percent of the respective workforces [45]. Among countries with comparable data for 2000–01 and 2020–21, foreign-born doctors increased by 86 percent and foreign-born nurses by 136 percent. The OECD connects growing reliance partly to aging populations and demand for care [45]. AI-supported administration, logistics, or monitoring can extend their capacity while leaving demand for clinical judgment, responsibility, and human relationships. A country can consequently reduce some production-worker requirements while continuing to recruit heavily for care and health services.
7.4 International talent as a condition of capability
Building and governing autonomous systems requires researchers, engineers, technicians, infrastructure operators, and other specialists. An OECD review reports that specialized AI skills remain a small part of the workforce and identifies demand for complementary capabilities in data interpretation, management, and problem-solving [40].
A country seeking to reduce reliance on foreign technology may initially need more foreign expertise to build the alternative. Semiconductor facilities, data centers, robotics fleets, and oversight bodies require different technical communities. Equipment can often be acquired faster than that knowledge develops. Recruitment and research cooperation can support capability while education and organizational learning build local competence.
The relationship should be assessed through the availability, retention, and substitutability of required skills. A foreign-born engineer can strengthen local operating capacity; reliance on one unavailable external maintenance team can weaken it. Nationality alone is an inadequate proxy for whether institutions can sustain a function. The relevant distinction concerns usable expertise and the arrangements through which it remains accessible. Domestic education and organizational learning can help institutions maintain that capability over time, while shared research and technical specialization offer options where national scale is insufficient.
7.5 Implications and limits
Baldwin’s account of “globotics” examines automation alongside telemigration, through which service work crosses borders without the worker migrating [51]. AI-supported expertise and imported autonomous equipment provide related possibilities for maintaining production with different combinations of local staff, remote services, and mobile specialists.
These arrangements may change particular recruitment needs while leaving aggregate migration substantial. They can also affect origin countries through employment opportunities abroad and the movement of specialists needed to maintain domestic capability. The direction depends on deployment, wages, demographic demand, institutional choices, and complementary investment.
The paper therefore retains labor mobility as a bounded implication of changing productive requirements. A broader claim about divergence between the mobility of people and technology belongs to subsequent empirical research. For sovereignty assessment, the immediate question is whether institutions can obtain and retain the skills required to operate, govern, and replace consequential systems. Section 8 considers how differences in that capacity interact with unequal infrastructure and access.
8. The Uneven Distribution of Sovereign Capability
AI and autonomous systems enter an international economy with large differences in capital, infrastructure, scientific capacity, industrial experience, and institutions. Those differences affect which states can develop technologies, which can obtain reliable access, and which retain alternatives when access changes.
UNCTAD's 2025 report identifies substantial concentration in AI research and intellectual property: the United States and China together accounted for approximately one third of AI publications and 60 percent of AI patents. It also reported that the United States hosted roughly one third of the TOP500 supercomputers and more than half of their aggregate computing performance [52]. These measures describe particular datasets and should not be read as a census of all AI capability.
As of June 2025, high-income countries accounted for 77 percent of global co-location data-center capacity, while low-income countries accounted for less than 0.1 percent. The World Bank also reported that high-income countries hosted 86 percent of TOP500 systems and 97 percent of their capacity [36]. Stanford's 2026 AI Index records $285.9 billion in US private AI investment in 2025, compared with $12.4 billion in China, while noting that private-investment comparisons understate Chinese state-directed financing [53]. Infrastructure, investment, and scientific output measure different capabilities; none alone determines sovereignty.
The following sections examine differences in capability and ways of organizing access. National provision, shared arrangements, and external sourcing can coexist within the same country and function. Strategic specialization can operate across all three. These approaches provide questions for assessment rather than mutually exclusive country categories or a ranking.
8.1 Breadth of national capability
Countries differ in the breadth of capability and credible alternatives they can sustain across the stack. Large markets, fiscal resources, and established engineering communities can support more extensive options. Critical gaps in materials, equipment, energy, or software can nevertheless interrupt particular functions even where national capacity is substantial.
Assessment should identify those gaps and examine the ability to adapt as technologies and access arrangements change. Breadth describes the range of functions that can be sustained; it supplies no threshold for classifying a country as technologically self-sufficient.
8.2 Coalition sovereignty
Coalition sovereignty arises where states share access to capabilities that would be difficult or uneconomic to reproduce separately. Infrastructure, research, procurement, standards, and supply arrangements can preserve specialization while improving continuity.
Its value depends on whether access survives scarcity or disagreement. Financing obligations, allocation rules, interoperability, legal commitments, and dispute resolution determine whether participants can use a shared capability when needed. An agreement contributes little if access disappears when another member’s demand rises. Coalition sovereignty therefore requires workable institutional arrangements as well as common interests.
8.3 Strategic specialization
Concentrating resources in selected domains can be more feasible than pursuing extensive national provision across the stack. A country can build on an existing advantage or address a dependency whose interruption would be particularly damaging, while securing complementary inputs externally.
Saudi Arabia’s May 2025 launch of HUMAIN illustrates an announced investment strategy spanning data centers, cloud infrastructure, models, and applications, including Arabic-language capability [54]. The announcement establishes an intended approach rather than operational independence. Its contribution to sovereign capability depends on what becomes usable locally, the terms of access to complementary technologies, and the institutions’ ability to sustain the resulting systems. Strategic specialization can support national provision, shared capability, or negotiated external access.
8.4 AI use and external dependence
External AI services and imported equipment can provide useful capability without the expense of reproducing their development ecosystem. The World Bank identifies opportunities for smaller and locally deployable applications in economies lacking frontier-scale infrastructure [36]. Such access can be a defensible use of scarce resources.
Intensive use can coexist with substantial external dependence. Exposure becomes consequential where an essential function cannot be maintained, transferred, or replaced within the required period. Several procurement decisions can accumulate reliance on a small number of providers. Assessment should therefore examine both the capability obtained and the options retained. Domestic expertise, portable data, and diversified access can improve continuity without national manufacture of every underlying technology.
8.5 Developing economies and asymmetric dependence
For countries with limited resources, reliable electricity, connectivity, cybersecurity, public data, and technical skills may provide more usable capability than a nationally branded frontier model. The World Bank’s connectivity, compute, context, and competency framework identifies these foundations [36].
India’s digital-public-infrastructure approach offers an architectural illustration. Under its 2023 G20 presidency, ministers endorsed a voluntary framework emphasizing security, interoperability, open standards, governance, and inclusive access [55]. This demonstrates an approach to shared digital systems rather than AI independence. Its sovereignty implications depend on actual suppliers, institutions, data protections, and exit options.
Imported technology also requires enough local expertise to assess, adapt, and maintain it. Regional research, procurement, and training can distribute these costs where national scale is insufficient.
8.6 Regional infrastructure and unequal access within coalitions
Shared arrangements can redistribute dependence within a coalition. The location of infrastructure and the rules governing allocation affect which members have usable access. Smaller participants may otherwise exchange dependence on a global supplier for reliance on a regional gatekeeper.
Assessment should examine emergency access, allocation, affordability, dispute procedures, and participants’ operating skills. Formal membership alone establishes little about these conditions. Shared investment strengthens sovereign capability where it creates options participants can use even when their immediate interests diverge.
8.7 Unequal access to productive gains
AI can expand access to services and expertise while control over compute, energy, hardware, financing, and technical knowledge remains uneven. Broad use of applications can coexist with concentrated control over their supply. The distributional question concerns who can sustain access and influence its conditions when scarcity or disagreement arises.
Abundance within states does not guarantee abundance between states. Here abundance refers to substantial sectoral gains in available goods, services, or expertise. Governments may reserve strategic capacity for domestic users or selected partners, but this remains conditional on capability and policy choices. Material constraints continue to matter, and ownership, public services, taxation, and labor institutions influence the distribution of gains within each country. Unequal access is therefore a possible implication to examine, without presuming that general abundance has arrived.
8.8 Strategic interaction and the distribution of power
Sovereignization changes the environment to which other states respond. A restriction intended to protect strategic capability can raise the cost of dependence for foreign users. Those users may seek substitutes, reorganize procurement, or invest in competing infrastructure. Existing suppliers may then perceive a loss of market access or technological advantage and support further protective measures. The dependency premium is partly produced through these interactions, rather than fixed outside them.
The security-dilemma literature examines how measures intended to improve one state's position can reduce others' perceived security and produce reciprocal responses [56]. Technological-sovereignty research also emphasizes that national interventions can reshape the international structures within which states act [9]. Applying that insight to technology requires attention to the instrument and its purpose. Security restrictions can address genuine risks. Their effects nevertheless extend to investment incentives and expectations about future access, including among third countries not directly targeted.
The United States' October 2022 controls on advanced computing and semiconductor-manufacturing items illustrate the use of access restrictions in a technology supply chain [57]. China's July 2023 announcement of controls on gallium- and germanium-related items illustrates control over a different upstream position [32]. These documented measures show that leverage can operate at different layers. Their sequence alone does not establish that every measure was retaliation or that they caused any particular relocation decision.
Sanctions provide a related mechanism where access to finance, technology, infrastructure, or commercial relationships can be restricted through legal authority. Their relevance to the present framework lies in the way they can convert existing dependencies into constraints on another actor’s freedom of action. A country or firm that depends on externally controlled payment channels, software, cloud services, advanced hardware, logistics, or maintenance may discover that formal ownership of an asset provides limited operational autonomy once access to a critical input is restricted. A fuller analysis of sanctions regimes would require separate treatment of their legal, financial, and geopolitical dimensions.
A reciprocal mechanism would require additional evidence. Restrictions would have to change purchasers' expectations, lead to investment in alternatives, and provoke further adjustment by suppliers or governments. AI would matter where it changed the cost or credibility of the response. A country with an attainable automated alternative could react differently from one unable to maintain the same function without the restricted input.
Firms participate in this process. Providers may develop local deployments, partner with domestic operators, change licensing terms, or support standards to retain customers concerned about dependence. They can also favor proprietary architectures that make substitution harder. Governments purchasing these services are responding to commercial strategies as well as interstate rivalry. Section 9 examines how corporate control becomes connected to state jurisdiction.
Escalation is only one possible outcome. Diversified supply can reduce the leverage of a restriction. Transparent licensing, credible access agreements, and interoperable alternatives may reduce incentives for abrupt withdrawal. Cooperative arrangements can become more attractive when national replication is expensive. The empirical question is whether responses amplify insecurity, stabilize access, or produce different effects across sectors.
Over time, these interactions can reorganize the distribution of power. States able to sustain or replace critical functions have options unavailable to states that cannot. Where alternatives remain concentrated, the hierarchy may deepen despite widespread adoption of AI. Where technologies lower entry barriers or coalitions provide reliable access, smaller states may gain freedom of action. Neither outcome follows from automation alone.
8.9 When sovereignization becomes an international order
Using the threshold established in Section 1.5, sovereignization becomes relevant to international order when expectations and practices surrounding critical dependence become durable enough to shape the behavior of governments, firms, and institutions. The relevant evidence concerns implementation, persistence, and interaction across countries.
The proposed organizing norm concerns acceptable dependence. A structural transition would be visible if governments routinely regarded continuity and substitution capacity as legitimate considerations in economic policy, and if suppliers, trading partners, and institutions adapted around those expectations. Disputes would then concern the boundaries and costs of sovereignty requirements within a recurring institutional practice, rather than isolated emergency measures.
Relevant evidence includes the spread of dependency analysis beyond traditional security sectors; persistent requirements in procurement, infrastructure, trade, and investment; access arrangements that survive political change; and differences in substitution capacity that affect bargaining behavior. The AI-specific claim additionally requires observable improvements in the relative feasibility of alternative capability attributable to AI or autonomous systems. Unequal access to productivity gains provides a further distributional implication to investigate.
The analysis should compare these developments with earlier economic-security practice. Rising protectionism or a temporary crisis response cannot alone establish the proposed mechanism. Studies should specify their sectors, observation period, baseline, and rival explanations before drawing conclusions about a system-level transition.
Evidence would weaken the order-level claim if sovereignty requirements remained confined to a few established security domains, receded as immediate tensions eased, or failed to affect actual procurement and investment. Sectoral tests should identify candidate activities in advance using the conditions set out in Section 3.6. If automation consistently widened the disadvantage of controllable alternatives in those activities, that evidence would weaken the proposed production mechanism. Persistent fragmentation might still occur in that setting, but it would need to be explained primarily through geopolitical rivalry or concentration rather than demonstrated improvements in the feasibility of those alternatives.
The New Sovereignist Order is therefore a hypothesis about a particular combination of technological change, strategic interaction, and institutional adaptation. The following section examines whether investments made under that logic preserve usable control or move dependence into less visible parts of the system.
9. The Sovereignty Paradox
The pursuit of technological sovereignty can create new dependencies inside the systems intended to reduce them. A government can keep sensitive data inside national territory while processing them on imported accelerators. A domestic cloud provider can operate local data centers while relying on proprietary software or remote technical support. A national model can be trained locally while depending on external hardware, development tools, or research. Industrial robots can operate in domestic factories while receiving updates, diagnostics, or fleet-management services from platforms controlled abroad.
This paper calls this the sovereignty paradox: measures that increase apparent or immediate autonomy can relocate dependence into technical layers that are harder to observe, govern, or replace. Physical location, domestic ownership, and national branding can strengthen control, but none establishes effective sovereignty independently.
9.1 Sovereign AI and layered dependence
The growing interest in sovereign AI reflects legitimate objectives around sensitive data, jurisdiction, national infrastructure, local languages, and reduced exposure to concentrated providers. These objectives can be pursued through domestic data centers, sovereign cloud configurations, local models, public compute, or partnerships that provide stronger operational control.
Each arrangement addresses some dependencies while leaving others intact. A domestic data center improves control over physical infrastructure without producing semiconductor independence. A locally operated model can reduce reliance on an external API while increasing requirements for compute, engineering, security, and maintenance. The European Commission's Cloud Sovereignty Framework reflects this layered view by assessing legal control, data and AI, operations, supply chains, technology, security, and strategic considerations rather than treating physical location as sufficient [35].
The semiconductor value chain provides the clearest example. OECD analysis identifies substantial concentration across design tools, manufacturing equipment, materials, fabrication, packaging, and related stages [33]. Domestic compute infrastructure can therefore increase national capacity while remaining exposed to internationally concentrated upstream inputs. Similar patterns appear in robotics, where a locally assembled machine can depend on imported sensors, motors, batteries, processors, or control software.
9.2 Upstream dependence and accumulated lock-in
Sovereignization frequently shifts dependence upstream. Domestic compute increases demand for processors, networking equipment, cooling systems, electricity, software, and maintenance expertise. Automation can reduce dependence on labor while increasing dependence on robotics, industrial software, sensors, chips, energy, and technical support. Some of these new dependencies may be more concentrated than the ones they replace.
Dependence also accumulates after procurement. Applications are built around platform interfaces, data are stored in provider-specific formats, employees acquire specialized skills, and security or compliance processes become tied to a particular architecture. AI can deepen this problem as agents, retrieval systems, evaluations, and institutional workflows become organized around a model or provider. The EU Data Act's switching provisions illustrate the broader policy response to this form of lock-in in data-processing services [24].
A technology provider also has commercial agency. It decides where to invest, how to allocate scarce capacity, which services to support, and how to structure access, subject to its contracts and legal obligations. These decisions can affect customers independently of any government instruction. Dependence on a corporation should therefore be distinguished from dependence on a state, even where the two are connected through jurisdiction.
The US Department of Justice's explanation of the CLOUD Act provides a bounded example. It describes how a provider subject to US jurisdiction can be required through valid legal process to disclose covered data within its possession, custody, or control regardless of storage location, subject to applicable legal requirements and safeguards [58]. This concerns data access, not a general power to discontinue foreign AI services. Restrictions on supplying technology arise through separate instruments, including export controls. The sovereignty assessment must identify the particular legal authority, corporate structure, technical access, and conflicting obligations involved rather than infer a government's power from a supplier's nationality alone.
9.3 Physical autonomy and manufacturer dependence
Autonomous physical systems make the paradox especially visible because ownership of the machine can diverge from control over its operation. Robots can depend on manufacturer services for activation, identity, configuration, telemetry, software updates, diagnostics, maintenance records, maps, credentials, or proprietary replacement components.
These dependencies become consequential when machines perform essential economic functions. Manufacturers can discontinue products, change licensing terms, be acquired, fail, or leave a market while the physical equipment remains useful. Procurement of consequential autonomous systems therefore needs to address maintenance information, local repair, software continuity, operational-data portability, update procedures, identity and credentials, spare parts, and transfer between owners or operators.
Localization should therefore be assessed according to the dependency it actually changes. Keeping sensitive data within national territory can reduce legal or security exposure, but operational control also depends on who administers the software, controls encryption keys, operates authentication, provides updates, and can migrate the workload. Physical presence is one form of control within a wider operating architecture.
9.4 Five dimensions of technological sovereignty
The dependencies above can be assessed through five dimensions. Location concerns where infrastructure, data, and operations are situated. Jurisdiction concerns which laws, courts, contracts, and governmental authorities apply. Operation concerns whether the system can continue without continuous participation by the original provider. Technical control concerns whether authorized actors can inspect, maintain, secure, integrate, modify, and replace critical components. Continuity concerns whether the underlying function can survive the loss of an important supplier, service, network, or component.
These dimensions interact and need not be maximized for every application. A consumer productivity tool and a system operating critical infrastructure justify different requirements. The useful question is which dependency can interrupt the function and what level of control is proportionate to that consequence.
The five dimensions can also reveal asymmetry inside a single system. A national platform may provide strong control over location and jurisdiction while remaining difficult to maintain without the original provider. Another system may rely on foreign infrastructure while retaining strong portability, documentation, multiple suppliers, and tested fallback arrangements. The second configuration can provide greater practical continuity despite being less visibly national.
9.5 Continuity, domestic concentration, and preserved optionality
Continuity provides a practical test of sovereignty because disruption reveals which dependencies are consequential. Institutions can examine what happens if international connectivity fails, a cloud provider withdraws, replacement processors are unavailable, a robotics manufacturer ends support, or a critical authentication service becomes inaccessible. The objective is to know what will fail, how quickly, which functions can continue, and which alternatives can be activated.
Sovereignization can also create concentration inside the domestic economy. A government that supports one national provider and then builds essential public functions around it may replace foreign dependence with domestic lock-in. Competition, interoperability, portability, and exit arrangements therefore remain relevant even where the supplier is locally owned.
The sovereignty paradox ultimately returns to optionality. A technological system provides stronger practical sovereignty when providers can be substituted, data can move, equipment can be maintained, critical software can continue operating, and institutions possess the knowledge and authority needed to activate alternatives. Additional infrastructure strengthens sovereignty when it expands future choices rather than narrowing them.
Continuity exercises should test whether nominal alternatives can be activated within the required period. They may reveal that an additional provider, local operating mode, stock of components, or documented recovery procedure provides sufficient resilience without full domestic replication. The results should inform procurement and operating plans rather than remain a compliance record.
This logic also applies to the state itself. As public institutions become more dependent on AI and autonomous systems to perform consequential functions, control over external suppliers is no longer sufficient. They must retain authority over the systems through which sovereign capability is exercised. The next section examines that internal boundary.
10. Human Control and Institutional Accountability
AI can expand state capacity while making public institutions dependent on systems they struggle to direct, interrupt, or replace. Control over suppliers addresses only part of that exposure. Institutions also need authority over the purposes, permissions, and operating arrangements through which autonomous systems perform consequential functions. This chapter concerns those institutional requirements; technical assurance of a model’s safety or alignment requires separate evaluation.
10.1 Authority and institutional control
An institution should specify what a system may do, the resources it may use, and the conditions requiring escalation. A machine can execute intermediate actions without individual approval while its objectives and permissions remain subject to responsible officials. Describing that discretion does not require treating the machine as a bearer of public authority.
The boundaries depend on consequence, reversibility, and operating conditions. Summarizing documents differs from deciding access to benefits or intervening in infrastructure. Decisions about acceptable burdens, competing rights, and public objectives require an accountable institutional basis. Permissions should be reviewed when model versions, operating conditions, or capabilities change.
The Council of Europe Framework Convention supplies principles concerning human dignity, autonomy, transparency, accountability, and reliability within its scope [59]. Their legal application depends on jurisdiction and treaty obligations. The sovereignty concern is whether institutions retain the knowledge and authority needed to govern the system after it becomes embedded in public work.
Technical performance cannot determine whose access to a public service should be restricted or which risks a population should bear. AI may inform those choices, but a responsible institution must establish their objectives and legal basis. Constitutional and administrative arrangements vary across countries; the assessment should identify the actual authority involved rather than infer it from a system’s capabilities.
10.2 Oversight, records and continuity
Responsibilities can become dispersed among developers, cloud providers, integrators, operators, and public agencies. The deploying institution should identify who authorizes use, monitors performance, handles incidents, and can suspend or replace the system. NIST’s AI Risk Management Framework supports documented responsibilities, continuing monitoring, oversight, and decommissioning arrangements [60]. Supplier commitments support these functions without displacing public responsibility.
Records should make consequential actions and changes reconstructable, including the version deployed, its permissions, relevant inputs, interventions, and incident responses. A nominal approval step offers limited protection where the reviewer lacks time, expertise, information, or authority. The AI Act’s oversight provisions and NIST’s framework address relevant capabilities and responsibilities [60, 61].
For rapid processes, control can operate through bounded permissions, monitoring, escalation, and periodic review. Intervention must also account for continuity: stopping an automated process can interrupt an essential service. Exercises should establish what an override affects, which fallback functions remain available, and who can authorize recovery. Retained expertise and documented procedures help determine whether an alternative can actually be activated. Physical systems add maintenance providers, manufacturers, and component suppliers to the responsibility chain. Oversight needs to cover those relationships throughout the operating life of the deployment, including changes in support, software, and ownership.
10.3 Accountability, contestability and remedy
Affected people need access to an institution able to examine evidence, correct errors, and provide remedy. The Council of Europe Convention addresses documentation, complaints, and procedural safeguards within its scope [59]. An explanation should be relevant to the decision and the person’s circumstances, rather than confined to general descriptions of model architecture.
Greater national technological control can coexist with weak public oversight. Digital-sovereignty claims serve different political purposes [10]. The capacity to sustain and direct a system must therefore be assessed separately from whether its use is legitimate and accountable. A government may have strong operational capability while offering inadequate opportunities to challenge its exercise. Identifiable responsibility, effective review, and remedy remain distinct requirements rather than assumed consequences of domestic ownership. Errors can arise from records, implementation choices, model limitations, or interactions among components. A challenge process needs someone able to examine the relevant evidence and correct the outcome, with records sufficient to establish how the decision was reached.
10.4 Practical assessment
Table 2 consolidates the institutional questions developed here. The requirements should be proportionate to the function and the consequences of failure. They identify evidence to investigate rather than supply a certification or a numerical scoring rule.
| Test | Principal question |
|---|---|
| Control | Can the responsible institution define the system’s authority, monitor its operation, and intervene when required? |
| Continuity | Can the underlying function continue if the system, provider, or critical dependency becomes unavailable? |
| Substitutability | Can the technology be replaced within an acceptable period without losing the capability it supports? |
| Accountability | Can identifiable people and institutions be held responsible for the system’s use, with effective review and remedy available to those affected? |
Control requires evidence that authorization, monitoring, escalation, and intervention work in practice. Continuity concerns fallback arrangements and acceptable degraded operation. Substitutability concerns the time and resources needed for replacement, including portable data, documentation, and transition support. Accountability concerns identifiable responsibility and effective review and remedy.
A future sovereignty assessment should report these characteristics transparently. Operational capacity and normative accountability should remain distinguishable, so that a score does not obscure either strong capability with weak oversight or strong safeguards with fragile infrastructure. Public institutions need both reliable means of acting and constraints governing how those means are used. The following chapter considers how policy can preserve those capabilities.
11. Governing the New Sovereignist Order
Governments managing consequential technological dependencies face choices among forms of provision and access. Those choices involve cost, resilience, innovation, security, and participation in global markets. They require institutions capable of understanding supply chains and operational dependencies that extend beyond administrative boundaries.
The governance task is to identify which functions deserve additional protection, determine the required performance and level of control, compare credible arrangements and their full operating requirements, and periodically reassess whether the intervention remains justified. Sovereign capability therefore includes the capacity of government to govern its own technological dependencies.
11.1 National sovereign-capability strategies
A national strategy should begin with functions rather than a growing list of strategic technologies. Governments can identify functions whose prolonged interruption would materially affect economic continuity, public safety, essential services, or the exercise of lawful authority, and then trace the dependencies supporting those functions through the Sovereign Capability Stack.
The analysis should establish where concentration exists, how long replacement would take, which upstream inputs are vulnerable, and what level of service must survive during disruption. Different capabilities can then receive different responses: domestic capacity where interruption is intolerable and alternatives are unreliable; regional or trusted access where fixed costs are too high for national replication; diversification where several suppliers exist; inventories where disruption is temporary; and interoperability where switching costs create the principal vulnerability.
These classifications should be reviewed periodically. Technologies commoditize, industries consolidate, new architectures emerge, and autonomous systems can remove one constraint while creating another. Sovereign-capability planning therefore resembles continuing risk management more than permanent designation of protected sectors.
Current government capacity remains uneven. The OECD's 2026 Digital Government Outlook reports that 21 of 36 surveyed countries provide central support for procuring AI goods and services, while risk assessment, audit, and systematic impact measurement remain less consistent [62]. The gap matters because sovereign technology policy can fail through ordinary procurement decisions even when national strategy is sophisticated.
11.2 Procurement as an instrument of sovereignty
Public procurement can shape technological dependence long after the initial purchase. Agencies reorganize workflows, data, employee skills, security processes, and applications around chosen platforms, allowing switching costs to grow gradually. AI increases this risk because models and agents become integrated into institutional processes rather than used as isolated tools.
Sovereignty considerations should therefore enter procurement before dependence accumulates. Contracts for consequential systems can address data portability, technical documentation, transition assistance, software updates, security obligations, audit access, interfaces, incident reporting, continuity arrangements, and the conditions under which important technical features may change. Supply-chain visibility can expose remote administration, subcontractors, software dependencies, and hardware concentration that are not apparent from the location or nationality of the primary provider.
The European Commission's Cloud Sovereignty Framework illustrates the method by translating sovereignty into measurable legal, operational, supply-chain, technological, data, and security characteristics [35]. The broader principle is that sovereignty becomes governable when procurement authorities can evaluate observable capabilities rather than rely on labels such as national or sovereign.
11.3 Interoperability, portability, and maintainable openness
The ability to leave a technological relationship is one of the clearest indicators of optionality. Interoperable interfaces, common data formats, identity protocols, workload portability, documented APIs, and transferable credentials can reduce switching costs in cloud, AI, and autonomous systems. NIST's AI Agent Standards Initiative reflects the growing importance of secure interoperability as agents begin acting across organizational boundaries [38].
Open-source software and open standards can strengthen the same objective by reducing the technical control exercised by one vendor. Their value depends on maintainability. Source code or model weights create only potential optionality unless institutions possess the engineers, infrastructure, security processes, documentation, and financing required to operate them. Governments can support maintainable openness through procurement, shared maintenance, security investment, public-sector technical teams, and collaboration with other institutions using the same components.
Portability should be tested rather than assumed. Two AI models may expose technically compatible interfaces while producing meaningfully different behavior, and moving a critical workload can require new evaluations, security reviews, and workflow changes. Periodic migration tests provide stronger evidence of optionality than contractual statements alone.
11.4 Regional capability and trusted technology relationships
Some layers of the Sovereign Capability Stack have fixed costs too large for efficient national duplication. Advanced semiconductor production, critical-materials processing, frontier compute, research infrastructure, energy networks, and some forms of industrial capacity benefit substantially from scale. Regional or allied capability can therefore provide more practical sovereignty than national replication.
Trusted relationships can include shared infrastructure, long-term supply agreements, joint investment, research cooperation, common standards, reciprocal access, and coordinated procurement. Their value depends on durability. Access rules, allocation during scarcity, investment obligations, dispute resolution, interoperability, and exit arrangements determine whether cooperation remains reliable during the conditions for which it was designed.
The European Union's 2026 Open Source Strategy makes a related point by linking open technologies with reduced lock-in while also emphasizing maintenance, stewardship, security, and skills [37]. Open technology is most useful to sovereignty when several institutions can share the cost of sustaining an alternative rather than each attempting to build an isolated national substitute.
11.5 Sovereign AI and autonomous physical infrastructure
The growing use of the term “sovereign AI” creates a risk that it becomes too broad to carry analytical value. The five dimensions developed in Section 9—location, jurisdiction, operation, technical control, and continuity—provide a more useful assessment. Different uses can require different combinations rather than a single maximum standard.
The same approach should extend to physical AI. A robot or autonomous vehicle combines a physical asset with software, data, identity, communications, maintenance, and often continuing manufacturer services. Consequential deployments can require locally available operating modes, maintenance information, operational-data portability, transferable credentials, defined update procedures, spare-parts arrangements, and manufacturer-exit provisions. Common standards for identity, authority, records, and communication can reduce the risk that each manufacturer creates a closed control system around its machines.
A graduated approach can avoid imposing maximum sovereignty on every system. The European Commission's proposed Cloud and AI Development Act uses differentiated sovereignty assurance levels, allowing stronger requirements where ownership, third-country exposure, or continuity matter more [63]. A similar risk-based approach can be applied to autonomous physical systems according to the function they perform.
11.6 Institutional constraints and international rules
Greater technological sovereignty can expand the information and operational power available to government. National control over data and AI infrastructure can reduce dependence on foreign providers while also increasing the state's ability to combine information, automate decisions, or conduct persistent monitoring. The human-control requirements developed in Section 10 therefore belong inside sovereignty strategies rather than in a separate ethics layer.
Independent oversight, courts, auditors, regulators, impact assessment, contestability, and remedy remain relevant as the technological means through which authority is exercised change. The Council of Europe Framework Convention provides one international reference point by placing AI lifecycle activity within obligations concerning human rights, democracy, and the rule of law [59].
Selective interdependence also creates continuing demand for international governance. Semiconductor supply chains, energy networks, data infrastructure, research communities, and autonomous systems cross borders. Standards influence interoperability; trade and investment rules affect access; international agreements can provide continuity; and identity, cybersecurity, liability, safety, and evidence become more important as autonomous systems operate across jurisdictions. International governance may therefore become more important as states place greater value on sovereignty because the quality of international rules affects whether dependence remains controllable.
11.7 Governing for optionality
The governance objective connecting these policies is preservation of credible options. Industrial policy can create an alternative supplier. Procurement can prevent unnecessary lock-in. Interoperability can reduce switching costs. Open technologies can preserve technical choice. Regional infrastructure can provide scale. Trusted partnerships can diversify access. Human-control requirements can preserve institutional authority over autonomous systems. A government considering intervention can identify the dependency it is addressing, establish the function and performance to preserve, compare the lifecycle and transition requirements of credible alternatives, and specify what new option the intervention creates. The resulting capability can then be tested periodically rather than assumed to exist because infrastructure has been purchased or legislation adopted.
These policies strengthen the ability to decide where control matters, understand the dependencies supporting critical functions, and preserve alternatives when access becomes unreliable. International exchange can remain extensive while institutions place greater weight on the conditions under which it sustains national freedom of action.
12. Conclusion: Sovereignty as Capacity, Not Isolation
International exchange remains a major source of national capability. AI and autonomous systems can change the feasibility of selected alternatives while geopolitical and commercial conditions change the consequences of dependence. The New Sovereignist Order is the proposed international configuration arising where those developments influence decisions and become embedded in durable relationships.
12.1 The changing cost of sovereign capability
The dependency premium and autonomy dividend organize a qualitative comparison between exposure and alternatives. A consequential dependency combined with an improved controllable alternative warrants investigation. The choice depends on required performance, lifecycle resources, remaining dependencies, and competing objectives. Governments may also accept substantial costs to preserve essential functions where automation provides no advantage.
The comparative requirement is central. Gains by established external suppliers, implementation costs, and constraints in supporting infrastructure can absorb or outweigh domestic improvements. An autonomy dividend requires evidence about the more controllable arrangement relative to the relevant external alternative.
12.2 Sovereignty Through Optionality
National provision, diversified suppliers, interoperable systems, inventories, and shared infrastructure can each preserve options. Their contribution depends on whether institutions can sustain the function during the disruption concerned. Coalition sovereignty requires access arrangements that remain usable during scarcity or disagreement.
The architecture of dependence can matter more than its volume. Extensive international sourcing can preserve freedom of action where alternatives are available, while substantial domestic production can remain vulnerable to one indispensable input or provider. These are conditions for assessment rather than fixed country categories.
12.3 Unequal access to productive capability
Productivity gains can expand alongside unequal control over the infrastructure producing them. Broad use of AI applications provides limited information about the terms on which institutions can continue using, adapting, or replacing them. Countries can combine domestic investment, specialization, coalition access, and imported services in different ways.
Domestic or preferential access to strategically important capability remains a possible policy choice. Its effects depend on technological feasibility, institutions, and political decisions. The paper’s distributional claim can be investigated through access conditions and usable capability without requiring a prediction of general abundance.
12.4 Layered dependence and human authority
The Sovereign Capability Stack makes upstream requirements visible. A local model can require imported hardware; an automated factory can depend on manufacturer services; shared infrastructure can depend on allocation rules that fail during stress. The sovereignty paradox directs assessment toward the dependencies an investment creates as well as those it reduces. The same inquiry applies inside public institutions. Authority to set objectives, constrain actions, preserve services, and replace systems must remain usable. Accountability adds the requirement for effective review and remedy. Technological capacity and the normative conditions of its exercise should be examined distinctly, with neither inferred from national ownership alone.
12.5 A testable research and measurement agenda
The paper advances two distinguishable claims: AI and autonomous systems can improve the relative feasibility of selected controllable capabilities, and repeated responses can contribute to a durable reorganization of international interdependence. Evidence supporting one claim would not automatically establish the other.
The production mechanism would be weakened if, in sectors selected in advance using the conditions set out in Section 3.6, automation consistently strengthened external suppliers without improving the relative feasibility of controllable alternatives. Comparisons should hold the required function and performance broadly constant, examine operating and transition costs, and separate technological effects from subsidies, demand shifts, and restrictions. A negative result could invalidate a sectoral hypothesis rather than merely refine it.
The order-level claim would be weakened if sovereignty requirements remained confined to established security exceptions, access arrangements proved temporary, or dependency analysis failed to affect procurement and investment. Persistent fragmentation without a discernible role for changed productive feasibility would require another explanation. More supportive evidence would trace improvements attributable to AI through decisions, responses by suppliers or states, and durable rules and access arrangements. Studies should specify their scope, observation period, baseline, and competing explanations in advance.
The stack and dependency characteristics provide foundations for a later AI sovereignty assessment. Such work would need indicators for usable capability, control, continuity, substitution, and reliable external or coalition access, with explicit treatment of evidence quality and missing data. Overlapping measures should be checked to avoid double counting. The autonomy dividend requires comparative evidence over time and should be kept separate from an initial snapshot of capability. Accountability should remain visible as a distinct assessment rather than being silently combined with material capacity.
A further research question concerns whether deployment changes the composition of international labor demand while increasing the value of technical talent. Current task and robotics evidence supplies a starting point, but a broader mobility claim requires separate testing. Neither that research question nor a future score is needed to establish the usefulness of dependency mapping.
Practical usefulness and explanatory success remain separate. The framework may help institutions compare options even if the proposed international transition fails to occur. This paper supplies no empirical basis for universal scoring weights, numerical thresholds, or an implementation deadline.
12.6 Sovereignty as capacity
Sovereign capability depends on what institutions can produce, continue accessing, maintain, and replace when important relationships change. The framework evaluates those choices across physical resources, technological systems, and institutional arrangements. Where AI changes their feasibility and states and firms adapt around the consequences, it can contribute to a reorganization of interdependence. Whether that process consolidates into a New Sovereignist Order remains an empirical question.
References
[1] World Trade Organization. The Multilateral System: 50 Years of Achievement. Geneva: WTO, 1998. https://www.wto.org/english/thewto_e/minist_e/min98_e/mc98_e/introd_e.htm
[2] World Trade Organization. World Trade Report 2023: Re-globalization for a Secure, Inclusive and Sustainable Future. Geneva: WTO, 2023. https://www.wto.org/english/res_e/publications_e/wtr23_e.htm
[3] World Trade Organization. World Trade Report 2026. Geneva: WTO, September 2026. https://www.wto.org/english/res_e/publications_e/world-trade-report-2026_e.htm
[4] OECD. Economic Security in a Changing World. Paris: OECD Publishing, 2025. https://doi.org/10.1787/4eac89c7-en
[5] Hirschman, Albert O. National Power and the Structure of Foreign Trade. Berkeley and Los Angeles: University of California Press, 1945. Catalogue record for the 1945 edition. https://openlibrary.org/books/OL17766062M/National_power_and_the_structure_of_foreign_trade
[6] Keohane, Robert O., and Joseph S. Nye Jr. Power and Interdependence. 4th ed. Boston: Longman, 2012. Originally published 1977. ISBN 9780205082919.
[7] Krasner, Stephen D. Sovereignty: Organized Hypocrisy. Princeton, NJ: Princeton University Press, 1999. https://doi.org/10.2307/j.ctt7s9d5
[8] Farrell, Henry, and Abraham L. Newman. “Weaponized Interdependence: How Global Economic Networks Shape State Coercion.” International Security 44, no. 1 (2019): 42–79. https://doi.org/10.1162/ISEC_a_00351
[9] Edler, Jakob, Knut Blind, Henning Kroll, and Torben Schubert. “Technology Sovereignty as an Emerging Frame for Innovation Policy: Defining Rationales, Ends and Means.” Research Policy 52, no. 6 (2023): 104765. https://doi.org/10.1016/j.respol.2023.104765
[10] Pohle, Julia, and Thorsten Thiel. “Digital Sovereignty.” Internet Policy Review 9, no. 4 (2020). https://doi.org/10.14763/2020.4.1532
[11] Bradford, Anu. Digital Empires: The Global Battle to Regulate Technology. New York: Oxford University Press, 2023. https://doi.org/10.1093/oso/9780197649268.001.0001
[12] OECD. OECD Supply Chain Resilience Review: Navigating Risks. Paris: OECD Publishing, 2025. https://doi.org/10.1787/94e3a8ea-en
[13] European Commission. Communication on European Tech Sovereignty, Accompanied by an EU Open Source Strategy. COM(2026) 503. Brussels: European Commission, June 3, 2026. Accessed 24 September 2026. https://digital-strategy.ec.europa.eu/en/library/communication-european-tech-sovereignty-accompanied-eu-open-source-strategy
[14] Bull, Hedley. The Anarchical Society: A Study of Order in World Politics. 2nd ed. London: Macmillan, 1995. Originally published 1977. https://doi.org/10.1007/978-1-349-24028-9
[15] Ikenberry, G. John. “The End of Liberal International Order?” International Affairs 94, no. 1 (2018): 7–23. https://doi.org/10.1093/ia/iix241
[16] United Nations General Assembly. Declaration on the Establishment of a New International Economic Order. Resolution 3201 (S-VI), 1 May 1974. United Nations Audiovisual Library of International Law, instrument and procedural-history record. Accessed 25 September 2026. https://legal.un.org/avl/ha/ga_3201/ga_3201.html
[17] Bush, George H. W. “Address Before a Joint Session of the Congress on the Persian Gulf Crisis and the Federal Budget Deficit.” 11 September 1990. George H. W. Bush Presidential Library and Museum. Accessed 25 September 2026. https://www.bush41library.gov/digital-research-room/finding-aid/public-papers/address-joint-session-congress-persian-gulf-crisis
[18] Pollack, Mark A. “The New, New Sovereigntism, or How the European Union Became Disenchanted with International Law and Defiantly Protective of Its Domestic Legal Order.” In Whither the West? International Law in Europe and the United States, edited by Chiara Giorgetti and Guglielmo Verdirame, 73–112. Cambridge: Cambridge University Press, 2021. https://doi.org/10.1017/9781316996836.005
[19] Chivvis, Christopher S., Rohan Mukherjee, and Milan Vaishnav. “How Populist Middle Powers Are—and Are Not—Reshaping Global Politics.” Carnegie Endowment for International Peace, August 6, 2026. Accessed 24 September 2026. https://carnegieendowment.org/research/2026/08/how-populist-middle-powers-areand-are-notreshaping-global-politics
[20] Besley, Timothy, and Torsten Persson. “The Origins of State Capacity: Property Rights, Taxation, and Politics.” American Economic Review 99, no. 4 (2009): 1218–1244. https://doi.org/10.1257/aer.99.4.1218
[21] Kim, Galileu, Tanu Kumar, Rita Ramalho, and Stuart Russell. “Rethinking State Capacity: A Practical Framework for Better Policy Implementation.” World Bank Blogs, 5 March 2026. Accessed 24 September 2026. https://blogs.worldbank.org/en/governance/rethinking-state-capacity--a-practical-framework-for-better-poli
[22] Mann, Michael. “The Autonomous Power of the State: Its Origins, Mechanisms and Results.” European Journal of Sociology 25, no. 2 (1984): 185–213. https://doi.org/10.1017/S0003975600004239
[23] Coombs, Nathan. “Infrastructural Power in Financial Governance: Its Meaning, Applications, and Varieties.” In The Cambridge Global Handbook of Financial Infrastructure, edited by Carola Westermeier, Malcolm Campbell-Verduyn, and Barbara Brandl, 37–48. Cambridge: Cambridge University Press, 2025. https://doi.org/10.1017/9781009428118.005
[24] European Commission. “Data Act Explained.” Policy factsheet. Accessed 24 September 2026. https://digital-strategy.ec.europa.eu/en/factpages/data-act-explained
[25] International Federation of Robotics. “Global Robot Demand in Factories Doubles Over 10 Years.” Press release accompanying World Robotics 2025: Industrial Robots, 25 September 2025. Accessed 25 September 2026. https://ifr.org/ifr-press-releases/news/global-robot-demand-in-factories-doubles-over-10-years
[26] Krenz, Astrid, Klaus Prettner, and Holger Strulik. “Robots, Reshoring, and the Lot of Low-Skilled Workers.” European Economic Review 136 (2021): 103744. https://doi.org/10.1016/j.euroecorev.2021.103744
[27] Artuc, Erhan, Paulo Bastos, and Bob Rijkers. “Robots, Tasks, and Trade.” Journal of International Economics 145 (2023): 103828. https://doi.org/10.1016/j.jinteco.2023.103828
[28] Reuters. “Humanoid Robot Sales Tally Hit 7,000 Globally Last Year.” September 21, 2026. Reporting International Federation of Robotics data on 2025 humanoid-robot sales. Accessed 24 September 2026. https://www.reuters.com/technology/humanoid-robot-sales-tally-hit-7000-globally-last-year-2026-09-21/
[29] International Federation of Robotics. “Service Robots See Global Growth Boom.” World Robotics 2025—Service Robots. Frankfurt: International Federation of Robotics, October 7, 2025. Accessed 24 September 2026. https://ifr.org/ifr-press-releases/news/service-robots-see-global-growth-boom
[30] International Energy Agency. Key Questions on Energy and AI. Paris: IEA, 2026. https://www.iea.org/reports/key-questions-on-energy-and-ai
[31] International Energy Agency. Global Critical Minerals Outlook 2025. Paris: IEA, 2025. https://www.iea.org/reports/global-critical-minerals-outlook-2025
[32] Ministry of Commerce of the People’s Republic of China. “MOFCOM Regular Press Conference (July 6, 2023).” 6 July 2023. Accessed 24 September 2026. https://english.mofcom.gov.cn/News/PressConference/art/2023/art_36fb2d80e4b4453891bb8fc83e2b3c4e.html
[33] OECD. Mapping the Semiconductor Value Chain: Working Towards Identifying Dependencies and Vulnerabilities. OECD Science, Technology and Industry Policy Papers, No. 182. Paris: OECD Publishing, 2025. https://doi.org/10.1787/4154cdbf-en
[34] International Telecommunication Union. Measuring Digital Development: Global Connectivity Report 2025. Geneva: ITU, 2025. https://www.itu.int/itu-d/reports/statistics/global-connectivity-report-2025/
[35] European Commission. Cloud Sovereignty Framework—Implementation Guidance. Brussels: European Commission, 2026. Accessed 24 September 2026. https://commission.europa.eu/document/download/2ad80a48-166f-4c77-a513-80c53ca2a128_en?filename=Cloud+Sovereignty+Framework+-+Implementation+guidance.pdf
[36] World Bank. Digital Progress and Trends Report 2025: Strengthening AI Foundations. Washington, DC: World Bank, 2025. https://documents1.worldbank.org/curated/en/099112525160536089/pdf/P505350-59c98ca8-0803-4f23-b470-17f3dab010ab.pdf
[37] European Commission. The EU Open Source Strategy. Brussels: European Commission, June 2026. Accessed 24 September 2026. https://digital-strategy.ec.europa.eu/en/factpages/eu-open-source-strategy
[38] National Institute of Standards and Technology. “Announcing the AI Agent Standards Initiative for Interoperable and Secure Innovation.” Gaithersburg, MD: NIST, February 17, 2026. Accessed 24 September 2026. https://www.nist.gov/news-events/news/2026/02/announcing-ai-agent-standards-initiative-interoperable-and-secure
[39] Morgan Stanley Research. The Humanoid 100: Mapping the Humanoid Robot Value Chain. Equity research report, 6 February 2025, pp. 2–4, 7.
[40] OECD. “Skills in the AI Age.” OECD Artificial Intelligence Papers, No. 60. Paris: OECD Publishing, 2026. https://doi.org/10.1787/972bd15e-en
[41] Dixit, Avinash K., and Robert S. Pindyck. Investment under Uncertainty. Princeton, NJ: Princeton University Press, 1994. https://doi.org/10.2307/j.ctt7sncv
[42] Baquie, Sandra, Yueling Huang, Florence Jaumotte, Jaden Kim, Rafael Machado Parente, and Samuel Pienknagura. Industrial Policies: Handle with Care. IMF Staff Discussion Note 2025/002. Washington, DC: International Monetary Fund, 2025. https://doi.org/10.5089/9798229000642.006
[43] World Trade Organization, Asian Development Bank, Institute of Developing Economies–Japan External Trade Organization, University of International Business and Economics, and World Economic Forum. Global Value Chain Development Report 2025: Rewiring GVCs in a Changing Global Economy. Geneva: World Trade Organization, 2025. https://www.wto.org/english/res_e/publications_e/gvcreport2025_e.htm
[44] UN Trade and Development. Global Trade Update: Global Trade Poised for a Record-Breaking 2025 as Flows Expected to Surge Past $35 Trillion. Geneva: UNCTAD, December 2025. https://unctad.org/publication/global-trade-update-december-2025-global-trade-poised-record-breaking-2025-flows
[45] OECD. International Migration Outlook 2025. Paris: OECD Publishing, 2025. https://doi.org/10.1787/ae26c893-en
[46] Gmyrek, Paweł, Janine Berg, Karol Kamiński, Filip Konopczyński, Agnieszka Ładna, Balint Nafradi, Konrad Rosłaniec, and Marek Troszyński. Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140. Geneva: International Labour Organization, 2025. https://doi.org/10.54394/HETP0387
[47] Acemoglu, Daron, and Pascual Restrepo. “Automation and New Tasks: How Technology Displaces and Reinstates Labor.” Journal of Economic Perspectives 33, no. 2 (2019): 3–30. https://doi.org/10.1257/jep.33.2.3
[48] Javed, Mohsin. “Robots, Natives and Immigrants in US Local Labor Markets.” Labour Economics 85 (2023): 102456. https://doi.org/10.1016/j.labeco.2023.102456
[49] Mann, Katja, and Dario Pozzoli. “Robots and Immigration.” Journal of Economic Behavior & Organization 227 (2024): 106708. https://doi.org/10.1016/j.jebo.2024.106708
[50] Chopra, Anand, and Ronit Mukherji. “Automation and Local Labour Markets: Impact of Immigrant Mobility.” Labour Economics 100 (2026): 102868. https://doi.org/10.1016/j.labeco.2026.102868
[51] Baldwin, Richard. The Globotics Upheaval: Globalization, Robotics, and the Future of Work. New York: Oxford University Press, 2019. ISBN 9780190901769. https://repository.graduateinstitute.ch/record/296793
[52] UN Trade and Development. Technology and Innovation Report 2025: Inclusive Artificial Intelligence for Development. Geneva: United Nations Conference on Trade and Development, 2025. https://unctad.org/publication/technology-and-innovation-report-2025
[53] Stanford Institute for Human-Centered Artificial Intelligence. The 2026 AI Index Report. Stanford, CA: Stanford University, 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report
[54] Public Investment Fund. “HRH Crown Prince Launches HUMAIN as Global AI Powerhouse.” 12 May 2025. Accessed 24 September 2026. https://www.pif.gov.sa/en/news-and-insights/press-releases/2025/hrh-crown-prince-launches-humain-as-global-ai-powerhouse/
[55] Government of India, Ministry of Electronics and Information Technology. “G20 Digital Economy Ministers’ Meeting (DEMM) Concluded with Adoption of the G20 Digital Economy Outcome Document and Chair’s Summary.” Press Information Bureau, 19 August 2023. Accessed 24 September 2026. https://www.pib.gov.in/PressReleasePage.aspx?PRID=1950443
[56] Jervis, Robert. “Cooperation under the Security Dilemma.” World Politics 30, no. 2 (1978): 167–214. https://doi.org/10.2307/2009958
[57] US Department of Commerce, Bureau of Industry and Security. “Commerce Implements New Export Controls on Advanced Computing and Semiconductor Manufacturing Items to the People’s Republic of China (PRC).” 7 October 2022. Accessed 24 September 2026. https://www.bis.gov/press-release/commerce-implements-new-export-controls-advanced-computing-semiconductor-manufacturing-items-peoples
[58] US Department of Justice. Promoting Public Safety, Privacy, and the Rule of Law Around the World: The Purpose and Impact of the CLOUD Act. White paper, April 2019. https://www.justice.gov/archives/opa/press-release/file/1153446/dl?inline=
[59] Council of Europe. Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. CETS No. 225. Strasbourg: Council of Europe, 2024. https://www.coe.int/en/web/artificial-intelligence/the-framework-convention-on-artificial-intelligence
[60] National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST AI 100-1. Gaithersburg, MD: NIST, 2023. https://doi.org/10.6028/NIST.AI.100-1
[61] European Parliament and Council of the European Union. Regulation (EU) 2024/1689 of 13 June 2024 Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act). Official Journal of the European Union, L 2024/1689, 12 July 2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
[62] OECD. Digital Government Outlook 2026: From Foundations to Transformational Impact. Paris: OECD Publishing, 2026. https://doi.org/10.1787/0496b2bc-en
[63] European Commission. Proposal for the Cloud and AI Development Act (CADA). COM(2026) 502. Brussels: European Commission, June 3, 2026. Accessed 24 September 2026. https://digital-strategy.ec.europa.eu/en/library/proposal-cloud-and-ai-development-act-cada