Introduction

Artificial intelligence has become a central arena of U.S.-China strategic competition because it combines economic scale, dual-use military applications, dependence on advanced infrastructure, and the capacity to shape standards and governance. The rivalry is therefore frequently described as a race for technological leadership. That description is accurate but incomplete. A competition over algorithms or model performance becomes geopolitically consequential only when it changes access to strategic inputs, redistributes economic advantages, alters military adoption, restructures alliances, or modifies the rules through which states and firms participate in global technological systems. The central analytical question is thus not simply which country is ahead in AI, but how competition over AI changes the mechanisms of international power.

The broader U.S.-China rivalry predates the recent acceleration of generative AI. Goldstein (2020) shows that the deterioration of bilateral relations reflects both structural factors, including the distribution of power, and national characteristics specific to the United States and China. Technology intensifies this rivalry because it affects both the material basis of power and the institutions through which power is exercised. Chen (2019) already treated artificial intelligence, big data, and cloud computing as interconnected domains in which the two countries possessed different advantages rather than a single hierarchy. Subsequent developments have reinforced that asymmetry. The 2025 AI Index reported a strong U.S. lead in notable frontier models and private investment, while China continued to lead in AI publication and patent volume and narrowed performance gaps on several benchmarks (Stanford Institute for Human-Centered Artificial Intelligence, 2025). The evidence therefore points to competitive specialization and uneven interdependence rather than simple technological replacement.

This matters for the concept of international order. An international order is not only a ranking of states. It also consists of institutions, rules, standards, networks, production systems, expectations, and patterns of legitimate authority. If AI competition affects semiconductor supply, cloud infrastructure, data governance, technical standards, military decision systems, investment controls, and alliance coordination, it may reshape the order even without producing a clear hegemonic transition. Conversely, technological superiority in a narrow domain does not by itself amount to systemic transformation. A rigorous analysis must distinguish changes in capability from changes in the rules and infrastructures that organize international interaction.

The article asks to what extent U.S.-China rivalry in artificial intelligence should be understood as a technological power struggle and to what extent it signals a transformation of the political, economic, and security order. Rather than treating the answer as predetermined, the analysis advances a qualified proposition: AI rivalry is technological in its immediate instruments but becomes systemic in its effects when competition over strategic assets produces durable changes in dependencies, alliance structures, governance arrangements, and access to critical infrastructures.

The article proceeds through six analytical stages, followed by a limitations section and conclusion. It first clarifies the analytical framework and evidentiary status of the review. It then examines national-security and military implications; economic and infrastructural competition; governance, standards, and alliance formation; and the wider implications for developing countries and future trajectories of rivalry. The discussion then evaluates three possible readings: a conventional technological race within the existing order, a bifurcation into competing technological blocs, and a more complex pattern of selective decoupling combined with continued interdependence and limited cooperation. This structure preserves the distinction between technological competition and systemic transformation without assuming an inevitable Chinese or American victory.

1. Analytical framework and method

This section lays out the article’s analytical and methodological foundations. It first defines the conceptual framework through which AI rivalry is examined as a competition over technological capabilities, strategic assets, infrastructures, dependencies, and rule-making power. It then specifies how these dimensions are operationalized in the analysis and explains the interpretive review method used to assess whether U.S.-China competition remains primarily technological or produces broader effects on the international order. The second part of the section focuses more specifically on the transition from capability competition to strategic-infrastructure rivalry, where control over compute, semiconductors, cloud systems, energy, and organizational capacity becomes central to geopolitical power.

1.1. Analytical framework and interpretive review method

The analysis treats artificial intelligence as a family of enabling technologies rather than a single, homogeneous capability. The National Security Commission on Artificial Intelligence (2021) similarly described AI as a constellation of technologies, while Horowitz (2018) argued that AI is closer to electricity or the internal-combustion engine than to a discrete weapon. This distinction is essential. Leadership in AI depends not only on producing an advanced model but also on the capacity to integrate algorithms with computing infrastructure, data, organizational routines, skilled labor, sensors, platforms, logistics, and decision processes. The strategic consequences of AI therefore depend on adoption capacity.

Ding and Dafoe’s (2021) concept of strategic assets provides a second analytical foundation. They distinguish cumulative, infrastructure, and dependency logics that can make a technology strategically significant. AI exhibits all three. Capabilities may accumulate through learning, investment, data, and talent; frontier systems depend on infrastructures such as advanced chips, data centers, electricity, networks, and cloud services; and states may become dependent on external suppliers for critical components or platforms. This perspective helps explain why AI rivalry spills over into semiconductors, telecommunications, export controls, investment screening, research collaboration, and standards. The relevant object of competition is an ecosystem, not an isolated algorithm.

The article adopts a structured interpretive review rather than a systematic review or an empirical causal design. It synthesizes peer-reviewed scholarship, book chapters, institutional reports, and comparative indicators across five dimensions: security and military adoption; cybersecurity and information risks; investment and technological infrastructure; governance and alliances; and effects on developing countries. Sources are used to compare mechanisms and assess the plausibility of competing interpretations of order change. The review does not claim exhaustive coverage of all U.S. and Chinese AI policies, nor does it infer causal effects from descriptive indicators alone.

This methodological clarification is important because the research question concerns systemic transformation, a claim that cannot be tested by descriptive indicators alone. The analysis therefore relies on theoretically guided synthesis rather than causal hypothesis testing. Evidence is weighted more heavily when multiple source types converge—for example, comparative data on research and investment, documented government strategies, peer-reviewed work on military adoption, and scholarship on global governance. Evidence is treated more cautiously when it rests on a single forecast, media narrative, or unsupported quantitative estimate. The argument therefore distinguishes observed trends from plausible scenarios.

Four indicators determine when technological competition becomes an order-level transformation. First, competition must alter the distribution or accessibility of strategic capabilities. Second, it must reorganize networks of dependence, for example through supply-chain controls or cloud and semiconductor chokepoints. Third, it must generate rule-making competition over standards, safety, data, surveillance, or military use. Fourth, it must affect the strategic choices of third states by changing the costs and benefits of alignment. Order change therefore depends on mechanisms rather than rhetoric alone.

This framework also avoids the assumption that the United States and China represent internally uniform models. The United States combines federal policy, private-sector concentration, military procurement, research universities, and alliances. China combines party-state planning, commercial firms, public research institutions, industrial policy, and international infrastructure initiatives. Their systems differ, but neither reduces to a simple opposition between an unregulated market and a fully state-controlled economy. Both governments increasingly treat AI, compute, semiconductors, talent, and data as matters of national capability, while private firms remain indispensable to innovation and diffusion.

Table 1. Analytical dimensions linking AI rivalry to international-order change

Dimension	Immediate object of competition	Potential order-level effect
Security	Military adoption, cyber capability, decision speed	Changes in deterrence, crisis stability, operational advantage
Infrastructure	Compute, semiconductors, cloud, energy, data	Strategic dependencies, chokepoints, selective decoupling
Economy	Investment, talent, firms, supply chains	Redistribution of innovation capacity and market access
Governance	Standards, regulation, safety, data rules	Competing norms and institutional influence
Third states	Infrastructure choices, financing, partnerships	Alignment pressures, diversification, bargaining agency

1.2. From capability competition to strategic infrastructure

The distinction between capability and infrastructure matters for interpreting recent AI developments. Model performance can change quickly, and benchmark leadership may be temporary. Infrastructure and organizational capacity usually change more slowly. Training advanced models requires access to specialized computing, software ecosystems, reliable energy, highly skilled personnel, and capital. Scaling deployment requires distribution networks, trusted institutions, and domain-specific expertise. For this reason, the strategic significance of AI depends on who controls bottlenecks and who can mobilize complementary assets.

The U.S.-China relationship illustrates this asymmetry. Stanford’s 2025 indicators showed that U.S.-based institutions produced substantially more notable AI models in 2024 and attracted far more private AI investment, while Chinese institutions led in publication totals and patents and were rapidly closing some technical performance gaps. Min et al. (2026) similarly find strong growth in Chinese output but a more complex picture when quality indicators are considered. These patterns make binary claims of absolute leadership analytically weak. Power resides in different layers of the AI stack, and those layers may not move together.

For international order, this implies that control over infrastructure can matter more than a temporary benchmark advantage. Export restrictions, investment controls, domestic industrial policy, and efforts to secure trusted supply chains are attempts to manage dependency. They may slow diffusion, incentivize substitution, and encourage the formation of partially separate technological ecosystems. Yet complete separation is difficult because research communities, markets, capital, and supply chains remain transnational. The resulting structure is better described as selective interdependence under strategic pressure than as complete technological autarky.

2. AI, national security, and strategic stability

National security provides the clearest setting in which AI rivalry can move from technological competition to strategic competition. The first part of this section examines how AI affects military capability, organizational adaptation, decision speed, and strategic advantage without assuming that technological superiority translates automatically into operational dominance. The second turns to cybersecurity, information operations, and crisis management, where AI may alter the speed, ambiguity, and informational conditions of confrontation. Together, these two dimensions show that AI’s security implications lie not only in weapons or platforms, but also in the institutions, infrastructures, and decision processes that convert technological capabilities into strategic effects.

2.1. Military capability, organizational adoption, and strategic advantage

National security is the domain where technological competition most directly becomes a power problem. AI can support intelligence analysis, logistics, targeting, autonomous systems, cyber defense, decision support, and information operations. Horowitz (2018) cautions, however, that technological invention does not automatically produce military advantage. Organizations must adopt, integrate, test, and trust new systems. This means the military balance depends as much on institutional learning as on model capability.

The United States has explicitly connected AI innovation to national-security competition. The National Security Commission on Artificial Intelligence (2021) framed AI as central to both defense and long-term technological competitiveness. China likewise identifies AI as a strategic technology and seeks a leading role by 2030, a goal Cheng and Zeng (2023) examine in the context of global governance. The symmetry lies in the strategic priority assigned to AI; the asymmetry lies in political institutions, industrial structures, alliance networks, and the distribution of private and public capabilities.

Military applications can create first-mover incentives, but the magnitude of those advantages remains uncertain. Some AI capabilities diffuse quickly because algorithms, publications, and software are widely accessible. Others depend on expensive compute, secure data, classified integration, specialized hardware, and organizational knowledge. The balance between diffusion and exclusion is therefore decisive. If important military applications diffuse rapidly, organizational adaptation may matter more than exclusive possession. If access to critical compute or hardware remains concentrated, chokepoints can generate stronger strategic leverage.

Autonomous systems, surveillance, and cyber operations are central to the security dimension, but the argument does not depend on spectacular or weakly sourced anecdotes. The core issue is that AI can accelerate sensing, classification, decision support, and coordination while also introducing opacity, brittleness, adversarial manipulation, and automation bias. Faster decision cycles may pressure actors to act before information is fully verified, especially in crises. The strategic risk is therefore not simply that AI makes weapons more lethal; it may change the temporal and informational conditions under which escalation is managed.

These effects do not imply that AI will independently determine military outcomes. The technology remains embedded in command structures, doctrine, communications, training, and human judgment. A state that develops advanced AI but cannot integrate it into reliable operational systems may gain less advantage than a technologically less spectacular competitor with superior organizational adoption. This is one reason why the rivalry should not be evaluated through headline demonstrations alone.

2.2. Cybersecurity, information operations, and crisis management

Cybersecurity links technological rivalry to strategic stability because cyber operations are hard to attribute, can occur below the threshold of armed conflict, and affect both civilian and military infrastructure. Xu and Lu (2021) show that China-U.S. cyber-crisis management is complicated by mistrust, strategic misperception, and the politicization of technology competition. AI may intensify these problems by increasing the scale of vulnerability discovery, social engineering, code generation, anomaly detection, and information manipulation, even though its practical effects vary by task and by the quality of human supervision.

The international-security problem is two-sided. AI can improve defense by detecting anomalies, prioritizing alerts, and supporting intelligence analysis. It can also lower the cost of producing convincing disinformation, automate some reconnaissance functions, and increase the speed of cyber operations. In a crisis, speed can be destabilizing if decision makers cannot distinguish an intentional attack from an accident, an autonomous action, or a manipulated signal. The risk is heightened when adversaries lack trusted communication channels or shared expectations about acceptable behavior.

Generative AI also expands the information environment in which strategic competition occurs. Wang and Downey (2025) demonstrate that public narratives about generative AI differ across national media systems, reminding us that AI is not only a material capability but also an object of political imagination. Narratives of threat, national decline, opportunity, or technological destiny can themselves influence policy by legitimizing investment, restriction, and securitization.

The most credible path toward risk reduction is therefore not a broad claim that cooperation will end competition. It is the development of limited crisis-management mechanisms: technical dialogue, incident communication, confidence-building measures, and clearer expectations regarding high-consequence systems. Xu and Lu (2021) stress precisely the need for confidence-building and academic exchange in the cyber domain. Such mechanisms do not require strategic trust; they are often most valuable when trust is low.

3. Economic competition, investment, and supply chains

The economic dimension of the rivalry is not separate from security. AI is a general-purpose technology that affects production, services, scientific research, logistics, and digital platforms. Its value depends on complementary investments in computing infrastructure, energy, data, software, and human capital. Khan et al. (2024) survey AI’s broad economic implications, while the Stanford AI Index offers a more nuanced comparison of investment and model development. Together, these sources support a conclusion of intense competition but do not justify speculative estimates about AI’s direct contribution to national GDP.

Private investment illustrates the current asymmetry. Stanford HAI reported that U.S. private AI investment in 2024 was far higher than China’s, even as China remained a major source of research publications, patents, and competitive models. This pattern matters because it shows innovation power is multidimensional. The United States benefits from deep capital markets, leading firms, cloud providers, research universities, and global talent networks. China benefits from scale, industrial capacity, extensive digital platforms, policy coordination, and a large research base. Neither advantage is reducible to a single ranking.

Investment also interacts with state policy. When governments classify AI-related inputs as strategic, commercial transactions acquire national-security significance. Semiconductors are the clearest example because advanced computing depends on a highly internationalized but concentrated supply chain. Efforts to control access to leading chips, manufacturing equipment, and design tools can preserve short-term advantages, but they also create incentives for substitution, stockpiling, domestic capacity building, and alternative supplier networks. The strategic effect depends on whether controls delay competitors more than they accelerate the creation of parallel ecosystems.

Barkstrom (2020) interprets the technology race as a source of both investment risk and innovation pressure. That duality remains useful. Strategic competition can mobilize resources, accelerate research, and stimulate industrial policy, but it can also duplicate infrastructure, reduce scientific exchange, raise compliance costs, and fragment markets. The economic consequences are therefore distributive: some firms and states benefit from re-shoring, diversification, or new demand, while others face higher costs or reduced access.

Geopolitical shocks since 2020 have heightened policy attention to resilience and strategic dependence. Pandemic disruptions, war-related energy and commodity shocks, and U.S.-China technology restrictions are distinct processes and should not be collapsed into a single causal chain. Taken together, however, they have made supply-chain vulnerability a central policy concern. AI competition enters this broader resilience agenda because frontier AI depends on concentrated, capital-intensive infrastructures.

3.1. Research leadership, talent, and innovation ecosystems

Research output often serves as a proxy for national AI power, but publication counts and technological leadership are not the same. Min et al. (2026) find that China has overtaken the United States in the volume of AI-related publications while the United States retains advantages on several quality measures. Stanford HAI reports a comparable pattern: China is strong in publications and patents, while the United States has been the dominant source of notable frontier models. A rigorous comparison must therefore preserve this multidimensional picture rather than selecting whichever indicator supports a predetermined national ranking.

Talent further complicates national comparisons. AI research is unusually international, and many leading researchers study or work across borders. Universities, laboratories, and firms depend on transnational networks even as governments seek to protect strategic knowledge. Excessive restrictions can reduce knowledge leakage, but they can also weaken the openness that supports innovation. The strategic problem is therefore one of calibration: states attempt to protect high-value capabilities without undermining the research ecosystems from which those capabilities emerge.

This tension explains why technological interdependence persists even during political rivalry. The two countries can simultaneously compete over critical technologies, restrict selected flows, and remain connected through markets, scientific knowledge, supplier relationships, and global demand. Decoupling is consequently better understood as sector-specific and variable rather than as a completed separation of two national economies.

3.2. Supply-chain bifurcation and the politics of dependency

The strongest evidence for order-level change appears where policy transforms the architecture of interdependence. Export controls, investment screening, procurement rules, trusted-supplier strategies, and industrial subsidies all change who can access strategic inputs. Ding and Dafoe’s (2021) dependency logic is especially useful here: an asset becomes strategically salient when reliance on an external supplier creates vulnerabilities that market actors do not fully internalize.

Such measures can produce a partial bifurcation of technological ecosystems, but bifurcation should be treated as a tendency rather than an accomplished fact. Firms have incentives to preserve large markets and interoperable standards. Third countries often seek diversified rather than exclusive relationships. Components may cross borders through complex supplier networks even when governments encourage national or allied sourcing. The global technology system is therefore fragmenting unevenly, with some layers highly securitized and others remaining internationally integrated.

The distributional effects extend beyond the two major powers. Countries without advanced semiconductor industries, hyperscale cloud infrastructure, or frontier-model firms may become more dependent on imported systems. At the same time, they may gain bargaining leverage when the United States and China compete to provide infrastructure, finance, training, or market access. The strategic question for third states is thus not simply which bloc to join, but how to avoid lock-in while obtaining technological benefits.

4. Political ramifications: governance, standards, and alliances

AI competition has a political dimension because the two powers seek not only technological capability but also influence over norms, standards, institutional practices, and acceptable uses. Cheng and Zeng (2023) interpret China’s participation in global AI governance as part of a broader shift from norm-taking toward norm-shaping. U.S. strategy similarly treats alliances and shared technological standards as instruments of competitiveness. The contest is therefore partly about who helps define the governance environment in which AI is designed, traded, deployed, and audited.

U.S. and Chinese governance models are sometimes reduced to a contrast between a minimally regulated market system and a tightly controlled state system. That dichotomy no longer captures the complexity of either case. U.S. governance combines federal and state regulation, sectoral rules, procurement requirements, safety initiatives, litigation, and private standards. China combines industrial promotion with extensive state oversight of platforms, data, algorithms, and content. The systems differ substantially in political accountability and rights protection, but both show that innovation and regulation occur simultaneously rather than as mutually exclusive processes.

At the international level, governance is also becoming more multilateral. The United Nations General Assembly adopted Resolution 78/265 in 2024 on safe, secure, and trustworthy AI systems for sustainable development without a vote. This does not eliminate strategic rivalry, but it shows that global AI governance cannot be reduced to bilateral U.S.-China competition. The European Union, India, Japan, the United Kingdom, middle powers, standard-setting bodies, firms, and transnational expert communities all influence the emerging regime.

The order-shaping effect of U.S.-China rivalry therefore operates through competition for coalition and rule-making capacity. States may cooperate with the United States on advanced semiconductors while trading heavily with China; they may adopt one provider’s telecommunications infrastructure while supporting multilateral principles that differ from both major powers; and they may seek domestic data-sovereignty rules that limit dependence on foreign platforms. Alignment is issue-specific, not necessarily bloc-wide.

4.1. Alliances and technological coalitions

Traditional alliances are increasingly supplemented by technology partnerships. Security relationships now involve semiconductors, cyber resilience, critical infrastructure, research security, cloud services, and standards alongside conventional military cooperation. The depth of such cooperation varies across partners and sectors, but the broader trend is toward integrating technological security into alliance coordination.

A stronger interpretation is that alliance politics creates networks of trusted access. These networks can improve resilience and coordination, but they can also deepen exclusion. The more sensitive a technology becomes, the more governments seek to distinguish trusted from untrusted suppliers and partners. This introduces political criteria into decisions previously treated as commercial. As a result, the international economy increasingly intertwines security screening and technological exchange.

China also uses economic scale, infrastructure provision, standards participation, and commercial platforms to build international technological relationships. Ünver and Ertan (2022) caution against assuming regime type mechanically determines whether a country imports U.S. or Chinese AI products; their analysis found more complex adoption patterns. This is especially important for avoiding a simplistic division between democratic states using U.S. technology and authoritarian states using Chinese technology. Cost, financing, technical compatibility, vendor availability, diplomatic relations, and domestic capacity all shape choices.

The political consequence is a more networked form of competition. States exercise influence through access to platforms and infrastructure, control of bottlenecks, interoperability, financing, training, standards, and security assurances. This differs from a classical territorial sphere of influence even when it produces comparable pressures. States can be deeply embedded in multiple technological networks at once.

4.2. Competing visions of governance without deterministic blocs

Normative competition nevertheless remains real. U.S. and allied policy discourse often emphasizes openness, rights, safety, accountability, and private innovation, whereas Chinese governance places greater weight on state capacity, social stability, security, and administrative control. These differences shape regulatory choices and international messaging, but actual policy practice is more mixed than ideal types suggest. Both sides regulate, subsidize, secure, and promote AI; the balance among those functions differs.

Tacheva and Ramasubramanian (2023) add another dimension by drawing attention to the global political economy of AI: extraction of data and resources, labor hierarchies, surveillance, and unequal value distribution. Their analysis does not show that developing countries naturally align with one major power, but it helps show that the AI order is structured by inequalities that transcend bilateral rivalry. A high-level account of international order must therefore examine who bears infrastructure costs, who controls platforms, who captures rents, and whose norms are embedded in systems.

Wang and Downey (2025) further demonstrate that societies construct different imaginaries of generative AI. These narratives matter because governance is partly legitimated through expectations about progress, risk, employment, security, and national prestige. Strategic competition can amplify technological nationalism, but it can also generate demands for international safeguards when the costs of uncoordinated development become more visible.

5. Global implications and future trajectories of U.S.-China AI rivalry

The implications of U.S.-China AI rivalry extend beyond the bilateral relationship and the immediate competition for technological capabilities. They affect the position of third states, patterns of technological dependence, the organization of global governance, and the possible future trajectories of international competition. This section therefore examines, first, how developing countries navigate the strategic pressures and opportunities created by competing technological ecosystems and, second, the main scenarios through which U.S.-China AI rivalry may evolve, from managed competition to fragmentation and limited cooperation.

5.1. Developing countries and the reconfiguration of technological dependence

Developing countries occupy a central but frequently oversimplified position in debates about U.S.-China AI rivalry. Treating them primarily as passive recipients of technology or as likely members of one technological bloc understates their agency and overstates the coherence of both the Global South and the major-power offerings available to it. Countries differ widely in digital infrastructure, market size, regulatory capacity, language resources, energy availability, education systems, and geopolitical preferences.

Aderibigbe et al. (2023) emphasize that AI adoption in developing countries is constrained by infrastructure, skills, connectivity, financing, and institutional capacity while also offering opportunities in health, education, agriculture, and public services. These constraints interact with geopolitical competition. Access to cloud computing, data centers, undersea cables, telecommunications networks, semiconductors, and training can improve national capabilities but may also create lock-in if contracts, standards, or proprietary systems are difficult to replace.

The relevant strategic choice is therefore often diversification. A state may use Chinese telecommunications equipment, U.S. cloud services, European regulatory templates, open-source models, and domestic data rules at the same time. Such combinations are not anomalies; they are rational attempts to avoid excessive dependence. Malik’s (2025) discussion of India, despite the limitations of that single-country study, illustrates how a third state can seek strategic autonomy while benefiting from multiple technological relationships.

Competition can also improve bargaining positions. When multiple providers seek market access, governments may negotiate better financing, localization, training, or data-governance conditions. But bargaining power is uneven. Small economies with limited technical expertise may lack the capacity to evaluate security, interoperability, long-term maintenance costs, or contractual dependencies. The international-order question is therefore not whether developing countries will choose the United States or China in the abstract, but whether they can preserve meaningful technological agency.

Global governance matters here because bilateral competition does not automatically provide public goods. Shared principles for safety, interoperability, development, and capacity building can reduce the pressure to accept exclusive technological dependence. The 2024 UN General Assembly resolution matters because it frames AI governance as a collective development issue rather than solely a strategic contest among major powers.

5.2. Future scenarios: competition, fragmentation, and limited cooperation

The future trajectory of U.S.-China AI relations is unlikely to be linear or uniform across technological domains. Rather than projecting a single outcome, this section distinguishes several plausible pathways shaped by the interaction between strategic rivalry, economic interdependence, technological dependencies, and the possibility of selective cooperation. It first examines three broad scenarios — managed technological competition, accelerated ecosystem fragmentation, and competitive coexistence with issue-specific cooperation — before turning to a cross-cutting problem that affects all three: the risk that AI-enabled speed, automation, and ambiguity may intensify escalation dynamics.

5.2.1. Competing trajectories: managed competition, fragmentation, and selective cooperation

The future of U.S.-China AI relations is unlikely to fit a single historical analogy. Goldstein (2020) warns against assuming that a new Cold War would reproduce the U.S.-Soviet confrontation. The economic interdependence between the United States and China, the central role of private firms, and the transnational structure of AI research make the current rivalry distinct. Vuving (2020) likewise identifies bifurcation and intensified great-power competition as important trends in the Indo-Pacific strategic environment, but these trends coexist with dense commercial connections.

A first scenario is managed technological competition. In this pathway, the two powers continue to compete intensely over frontier capabilities, semiconductors, investment, and standards while maintaining sufficient commercial and diplomatic interaction to avoid full technological separation. Selective controls expand in sensitive domains, but global firms and third states preserve substantial cross-system ties. This scenario aligns with the current mix of rivalry and interdependence.

A second scenario is accelerated ecosystem fragmentation. Here, export controls, sanctions, domestic substitution, data localization, standards competition, and alliance-based procurement progressively reduce interoperability. The world does not divide neatly into two blocs, but the costs of operating across systems rise. Smaller states face stronger pressure to choose trusted suppliers or maintain costly parallel infrastructures. Fragmentation may improve resilience for some strategic assets while reducing global efficiency and scientific exchange.

A third scenario is competitive coexistence with issue-specific cooperation. Even adversarial states can share interests in preventing catastrophic accidents, improving crisis communication, and setting expectations for high-risk systems. Cyber-crisis management offers a precedent for such limited cooperation (Xu & Lu, 2021). Similar logic could apply to autonomous weapons, nuclear command-and-control interactions, model safety, incident notification, or protection of critical civilian infrastructure. Cooperation would not end rivalry; it would seek to bound its most dangerous consequences.

The probability of each scenario varies by technological layer. Semiconductor manufacturing may be more securitized than academic publication; military AI more restricted than consumer applications; safety research more cooperative than model weights; and telecommunications procurement more politically contested than many enterprise software services. Future analysis should therefore avoid treating AI as a single policy domain with a single trajectory.

5.2.2. Escalation risks and the problem of speed

Escalation risk deserves particular attention because AI can compress decision time. Systems that generate recommendations, detect targets, classify threats, or automate cyber responses may create incentives to act quickly. Speed is not inherently destabilizing, but it becomes dangerous when it exceeds the capacity for verification, political control, or adversary signaling. False positives, adversarial manipulation, model error, and ambiguous attribution can then acquire strategic significance.

Escalation arguments require especially careful evidentiary grounding because technical mechanisms, strategic incentives, and crisis behavior cannot be inferred from the existence of AI alone. Goldstein (2020) provides the broader framework for understanding U.S.-China strategic rivalry, while Xu and Lu (2021) identify mistrust, misperception, and weak crisis-management mechanisms in the cyber domain. Taken together, these studies support a cautious inference: AI-related speed and automation may become destabilizing when they interact with ambiguous attribution, compressed decision time, and inadequate communication channels.

Risk reduction should focus on decision architectures rather than assuming that a ban on AI is realistic. Human authorization, testing, redundancy, auditability, secure communications, and doctrine can all shape whether AI accelerates instability or improves situational awareness. The most important variable is not merely whether AI is present, but how it is integrated into command systems and what safeguards surround its use.

6. Discussion: technological race or reshaping of international order?

The evidence supports a qualified answer to the research question. The U.S.-China AI rivalry is undeniably a technological power struggle, but it is not merely one. Its systemic significance arises because the objects of competition—compute, chips, data, cloud infrastructure, talent, standards, research networks, platforms, and military integration—are also infrastructures of contemporary economic and political order. Competition over them changes the conditions under which states, firms, and institutions participate in global systems.

The first mechanism is the redistribution of capability. AI contributes to economic productivity, scientific research, surveillance, military decision support, and cyber operations. Advantages in these areas can reinforce national power. Yet capability distribution is multidimensional. The United States can lead in frontier models and private investment while China leads in publication volume and patents. These cross-cutting indicators make a simple winner-takes-all narrative unsuitable for scholarly analysis.

The second mechanism is the reorganization of dependency. Advanced AI depends on strategic infrastructures that are concentrated geographically and institutionally. Policies aimed at securing semiconductors, cloud capacity, energy, or talent can reroute investment and trade. Dependency management is one of the clearest channels through which technological rivalry produces lasting order effects because it changes the architecture of access rather than only the ranking of capability.

The third mechanism is rule-making. AI governance increasingly concerns safety, data, standards, intellectual property, accountability, export control, military use, and human rights. The United States and China seek influence in these arenas, but neither acts alone. Multilateral institutions, the European Union, middle powers, firms, and technical bodies constrain a purely bilateral order. The emerging system is contested and plural rather than simply bipolar.

The fourth mechanism is third-state agency. A binary framework often assumes that developing countries will be sorted into American and Chinese technological spheres. The evidence suggests a more complex pattern. Many states seek to combine suppliers, preserve strategic autonomy, and exploit competition to improve access or financing. Their choices can in turn affect standards, market scale, and the legitimacy of governance models. Third states are not merely objects of order; they participate in producing it.

A maximalist claim that AI rivalry has already transformed the entire international order would be difficult to sustain as a completed empirical fact. Order transformation is ongoing and uneven. Many institutions and patterns of economic interdependence remain intact. What can be defended is a more precise argument: AI rivalry accelerates the securitization of technological interdependence and expands the domains in which technological infrastructure becomes an instrument of statecraft. This process reshapes important components of the international order without yet replacing it with a stable alternative.

This analysis also clarifies why technological determinism should be avoided. AI does not autonomously produce a new world order. Political institutions, corporate strategies, alliance commitments, regulatory choices, economic structures, and organizational adoption mediate its effects. The same technology can strengthen surveillance in one setting, productivity in another, military logistics in a third, and transnational scientific collaboration in a fourth. The international consequences of AI are therefore contingent and politically constructed.

The rivalry is best understood as a struggle over the infrastructures and rules of future power. Its outcome will depend less on a single breakthrough than on the durable capacity to combine innovation with adoption, secure critical inputs, attract talent, coordinate allies, manage risks, and offer third countries credible technological partnerships. This shifts the analytical focus from the question of who “wins” AI to the configurations of capability, dependence, and governance that become institutionalized. It is at this level that technological competition becomes order formation.

Table 2. Competing interpretations of the U.S.-China AI rivalry

Interpretation	What it explains well	What it misses	Assessment
Technological power race	Investment, models, patents, military adoption	Rules, dependencies, third-state agency	Necessary but insufficient
Bipolar technological fragmentation	Export controls, trusted networks, strategic supply chains	Persistent interdependence and issue-specific alignment	Plausible in selected layers, not as a total system
Order reconfiguration through strategic interdependence	Capabilities + infrastructure + governance + alliances	Requires careful dating and sectoral disaggregation	Best fit with current evidence

7. Limitations and research agenda

This article is a structured analytical review, not a systematic review, bibliometric study, or original empirical investigation. It does not provide a reproducible search protocol for the entire literature on U.S.-China AI competition; its conclusions should therefore be interpreted as a theory-guided synthesis rather than exhaustive measurement. Future systematic reviews could strengthen reproducibility by specifying databases, date ranges, search strings, inclusion criteria, and a coding framework for the literature.

The comparison also faces a rapidly moving empirical environment. Investment, model performance, export controls, corporate structures, and regulatory frameworks can change within months. Quantitative indicators must be dated explicitly and should not be converted into timeless statements about national superiority. The article therefore favors mechanisms that are more stable than individual rankings: dependence on strategic infrastructure, organizational adoption, governance competition, and third-state alignment.

Finally, the categories United States, China, and developing countries conceal significant internal variation. Future research should disaggregate firms, agencies, sectors, and technological layers, and examine how specific third countries combine suppliers and governance models. Comparative case studies would be particularly valuable for testing whether the tendency toward technological fragmentation actually produces durable political alignment.

Conclusion

The U.S.-China rivalry in artificial intelligence cannot be reduced to a contest over which country develops the most powerful model. The technology matters because it is embedded in infrastructures, institutions, supply chains, military organizations, markets, and governance systems. These connections convert technical capability into economic and political power. They also explain why competition over AI increasingly extends to semiconductors, cloud computing, standards, research security, talent, export controls, and alliance coordination.

The analysis therefore supports the central proposition, but only in a qualified sense. AI rivalry is technological in its immediate instruments but systemic in its effects. It can reshape international order when it changes dependencies, alters access to strategic assets, reorganizes technological networks, and generates new forms of rule-making competition. Yet the evidence does not support a deterministic scenario in which one power simply replaces the other or the world divides automatically into two closed blocs.

The more plausible trajectory is a contested order characterized by selective decoupling, persistent interdependence, competitive coalition-building, and limited cooperation in areas where unmanaged risk would be costly to both sides. The United States and China possess different strengths, and those strengths operate across different layers of the AI ecosystem. Third countries, including developing states, retain agency and can sometimes use major-power competition to diversify partnerships, although unequal capacity creates real risks of dependency.

For international-relations scholarship, the main analytical lesson is that AI should be studied neither as an autonomous technological force nor as a simple proxy for national power. The decisive questions concern adoption, infrastructure, dependency, institutions, and governance. For policy, the implication is equally clear: preserving technological advantage is only one objective. Strategic stability will also require mechanisms that reduce misperception, protect critical systems, maintain channels for crisis communication, and allow multilateral governance to address risks that cannot be contained within national borders.

Seen in this perspective, the U.S.-China AI rivalry is already contributing to a redefinition of how power is produced, distributed, and constrained. Whether that process ultimately yields a fragmented order, a renewed form of competitive coexistence, or a more institutionalized system of technological governance will depend on choices made by governments, firms, international organizations, and third states. The international order is not being transformed by AI alone; it is being renegotiated through the political struggle over the infrastructures, rules, and uses of AI.