Viewing Sino–U.S. Offensive and Defensive Dynamics in Computing Power and Network Structural Power through the Framework of “Weaponized Interdependence”

Abstract

Contemporary globalization, empowered by artificial intelligence (AI) technologies, has ushered in an unprecedented period of prosperity. Yet this prosperity is built upon highly asymmetric structures of interdependence and therefore contains profound structural vulnerabilities. This article focuses on the emerging perspective of “weaponized interdependence” and raises the following questions: in an AI-driven global economy, how are foundational elements such as computing power, networks, and critical materials systematically weaponized; why does such weaponization exhibit structural inevitability; and within the framework of Sino–U.S. competition over computing power and networks, how can the disadvantaged party achieve limited equilibrium through mutual constraint.

Combining theoretical analysis with case studies, this article draws on Farrell and Newman’s “chokepoint effect” and “panopticon effect” [1][2], Susan Strange’s concept of structural power [6], and Keohane and Nye’s theory of asymmetric interdependence [7]. Core cases include the 2022 chip export controls, the 2024 Red Sea submarine cable disruptions, export controls on gallium and germanium since 2023, the “Clean Network” initiative, and Executive Order 14105. On this basis, the article constructs a three-layer analytical framework of “computing power resources – countermeasures – foundational constraints.”

The findings are threefold. First, the apparent prosperity of the AI-era global economy conceals structural vulnerabilities jointly shaped by supply concentration, channel monopolization, and market inequality. Second, inequalities in resources, power, and circumstances render weaponized interdependence structurally inevitable, manifesting in the AI domain as overlapping supply-based, channel-based, market-based, and narrative-based forms. Third, in Sino–U.S. competition over computing power and networks, both sides actively employ their advantageous positions within network topology, producing a “symmetrical weaponization” of structural power; mutual constraint thus becomes a structural choice for maintaining fragile stability in the short term.

The article further argues that while mutual constraint may help prevent unilateral “crushing victories,” its long-term cumulative effects risk pushing the global technological system toward the brink of “mutual assured destruction.” This underscores the urgent need for multilateral mechanisms and forms of technological arms control to establish minimal “guardrails” for structural competition in the AI era.

Keywords: weaponized interdependence; AI economic vulnerability; structural power; computing power networks; mutual constraint

I. Introduction

At present, the global economy is undergoing a new wave of prosperity driven by artificial intelligence (AI). The exponential growth of data flows, the global expansion of computing infrastructure, and breakthroughs in large-model technologies have jointly constructed a seemingly tightly interconnected digital economy. Beneath this surface-level prosperity, however, lie profound structural vulnerabilities. These vulnerabilities do not stem from technology itself, but rather from the fact that the global interdependent networks supporting the AI economy are being systematically “weaponized.”

Since the outbreak of Sino–U.S. trade frictions in 2018, and especially following the U.S. Department of Commerce Bureau of Industry and Security (BIS) issuance of the Interim Final Rule targeting advanced computing chips and semiconductor manufacturing equipment in October 2022, global technology supply chains have undergone dramatic restructuring [3]. In early 2025, the U.S. government further proposed “AI Technology Diffusion Control Regimes,” seeking to construct a transnational restriction system targeting high-performance chips, algorithmic weights, and related manufacturing equipment. Although this regime has experienced policy fluctuations—such as the proposal and temporary suspension of the “AI diffusion rules”—its core logic is already clear: leveraging asymmetric structural advantages within global supply networks to impose political coercion on competitors. By contrast, China’s hub position in critical materials such as gallium and germanium, as well as in submarine cable construction, provides it with potential resources for reverse weaponization [9][10].

This condition of “mutual development and mutual constraint” has become a new normal in international relations. On the one hand, the deepening of globalization-driven specialization has intensified interdependence among states in areas such as computing power, data, and materials; on the other hand, these interdependencies are increasingly politicized and securitized, becoming instruments of geopolitical competition. From embargoes on high-end GPUs to retaliatory export controls on rare earths, from disputes over submarine cable routing to restrictions on cross-border data flows, conflict in the AI era is no longer confined to traditional territorial or military domains but has extended into the physical and logical infrastructures underpinning the digital economy. Against this backdrop, the structural contradictions in Sino–U.S. relations are further magnified: neither side can fully decouple, yet neither can return to an era of “pure cooperation” [11].

To analyze these dynamics, this article defines two core concepts. The first is weaponized interdependence, originally proposed by Farrell and Newman (2019) to describe how states exploit their central positions in global networks—such as the SWIFT financial system or internet infrastructure—to exert coercive power over other actors [1][2]. This article extends the concept from the financial and communications domains to the AI domain, referring to strategic behavior that leverages asymmetric advantages in computing power supply chains (chips, EDA software), data infrastructure (cloud services, submarine cables), and technical standards to cut off opponents’ access to critical resources (the chokepoint effect), or to exploit network centrality to obtain intelligence on adversaries (the panopticon effect).

The second concept is AI Technology Diffusion Control Regimes, which in this article refers specifically to transnational restriction systems built upon the U.S. Export Administration Regulations (EAR), including the Entity List, the Foreign Direct Product Rule (FDPR), and emerging “AI diffusion rules.” These regimes aim to delay or block competitors’ AI technological progress by controlling the cross-border flow of high-end computing chips, semiconductor manufacturing equipment (SME), and core algorithmic weights. This is not merely a trade policy, but a systematic instrument for projecting structural power.

This article adopts a three-layer analytical framework of “computing power resources – countermeasures – foundational constraints” to systematically analyze weaponized interdependence in the AI era.

II. The Alienation of Interdependence: Vulnerability and Manifest Risks in the AI Economy

2.1 Surface-Level Vulnerability of the Economy and AI

The prosperity of the global AI economy is built upon a profound “structural imbalance” that pushes efficiency maximization to the extreme at the expense of resilience. On the surface, global data flows and demand for computing power are growing exponentially, constructing a seemingly tightly interconnected digital economy. Yet the global supply chain networks supporting this prosperity exhibit extreme spatial concentration. The production of high-end AI chips is highly dependent on a small number of foundries such as TSMC, as well as on an extremely limited number of equipment suppliers such as ASML; the global financial settlement system similarly relies on single-point dependencies on the U.S. dollar and the SWIFT system. While such extreme concentration generates economies of scale, it also creates numerous “obligatory passage points” [6]. As Farrell and Newman point out, this highly centralized network topology grants central nodes a powerful “chokepoint effect,” whereby even minor geopolitical disturbances can trigger systemic paralysis.

Notably, this imbalance is particularly pronounced in the Sino–U.S. context. In computing power and high-end chip sectors, the United States and its allies dominate core segments such as design, EDA software, and manufacturing equipment, whereas China holds advantages in processing capacity and output in critical materials and certain infrastructure domains [9][10]. This upstream–downstream role differentiation places both sides within the same network but exposes them to fundamentally different structural risks. Historical precedents have repeatedly foreshadowed the eruption of such vulnerabilities: the 2018–2020 Sino–U.S. trade frictions shattered illusions of unfettered global free trade; financial sanctions following the 2022 Russia–Ukraine conflict demonstrated the potency of weaponization; and escalating chip export controls since 2023 have directly exposed these vulnerabilities at the core of the AI industry. This prosperity built on “quicksand” leaves the global economic system increasingly brittle in the face of geopolitical shocks, exhibiting systemic fragility in which “a single pull moves the entire body.”

2.2 Manifest Risks in the AI Economy

The weaponization of computing power resources directly disrupts the logic of technological iteration in the AI industry, transforming pure technological competition into a war of resource attrition. As the marginal returns of Moore’s Law diminish, improvements in AI model performance increasingly rely on the accumulation of massive parallel computing capacity. However, U.S. export control policies have artificially distorted this technological trajectory. With high-end GPUs such as NVIDIA H100 and A100 placed on embargo lists, sanctioned countries’ AI firms are forced to turn to lower-performance substitutes or acquire chips at high cost through complex gray channels. Research indicates that since 2024, the cost of acquiring computing power for Chinese AI firms has risen by an average of 40–60 percent, while the absence of cutting-edge hardware has significantly lengthened large-model training cycles—an existential blow to an industry driven by rapid iteration [4]. This “Whack-a-Chip” style of hardware blockade not only increases compliance costs and operational risks, but also seeks to lock in a hard ceiling on latecomers’ computing power, thereby fundamentally constraining the pace of their AI development.

Beyond these “hard” constraints on computing power, “soft” barriers in data flows and physical infrastructure are reshaping the global digital landscape, intensifying the “data island” effect. Data is the fuel of AI training, especially for large language models that depend on massive, diverse, and multilingual datasets. Yet under the banner of national security, countries are increasingly enacting data localization laws that restrict cross-border data flows, creating a fundamental tension with the globalization required for AI development. Meanwhile, submarine cable networks carrying approximately 99 percent of global data traffic face dual risks of physical disruption and route control. The combination of physical infrastructure vulnerability and institutional data barriers is accelerating the fragmentation of the global internet into mutually disconnected regional networks. Together, these manifest risks form a multidimensional containment network that places unprecedented structural obstacles before late-developing countries seeking AI advancement under conditions of constrained resources and network fragmentation.

2.3 Typical Case Analysis

2.3.1 The Systemic Evolution of Chip Embargoes: From Trade Restrictions to Precision Strikes

U.S. chip export controls against China have evolved from isolated trade restrictions into a systematic form of “precision surgical strikes,” rooted in the logic of leveraging upstream monopolistic power to sever downstream technological trajectories. This process began with the BIS Interim Final Rule of October 7, 2022, which not only restricted physical chip exports but, more destructively, implemented long-arm jurisdiction through the Foreign Direct Product Rule (FDPR), prohibiting any firm worldwide using U.S. technology—even when manufacturing in third countries—from supplying advanced chips to designated Chinese entities [13]. By 2025, these controls had further escalated into an “AI Technology Diffusion Control Regime” seeking to regulate cross-border transfers of large-model weights and even remote access to cloud-based computing power. Despite tactical fluctuations in implementation—such as permitting taxed exports of the H200—the strategic intent has remained constant: to paralyze competitors’ AI training capabilities by controlling the most irreplaceable chokepoints in the semiconductor supply chain, including EDA software, lithography machines, and high-end GPUs. This process not only exposes the vulnerability of the global semiconductor supply chain to political will, but also signals the collapse of the myth of “technological neutrality,” marking the formal entry of the global technology system into a security-driven era of bloc alignment.

2.3.2 Physical Infrastructure Vulnerability: Lessons from the Red Sea Cable Disruptions

Beyond cloud-level competition, submarine cable networks—the physical backbone of the global internet—are increasingly becoming direct targets of geopolitical conflict. In February 2024, multiple critical submarine cables in the Red Sea region (AAE-1, EIG, SEACOM) were severed. Although officially attributed to accidental anchor dragging caused by Houthi attacks on vessels, the incident carried profound geopolitical implications. Approximately 25 percent of data traffic between Asia and Europe was disrupted, leading to significant increases in internet latency in certain regions and severely affecting latency-sensitive financial transactions and cross-border cloud services [13]. This incident vividly illustrates that the “cloud” rests on fragile physical entities—cables no thicker than garden hoses that carry the lifeblood of the global digital economy. For states pursuing digital sovereignty, submarine cable security is no longer a matter of redundancy alone, but a strategic choke point of national economic security. The Red Sea cable incident underscores that the command of the digital economy still lies at physical bottlenecks, rendering any digital strategy that neglects physical infrastructure security fundamentally untenable.

2.3.3 The Structural Logic of Critical Materials Countermeasures: Reverse Projection of Asymmetric Power

Faced with upstream technological blockades, downstream states are not without recourse; monopolistic positions in critical materials provide strategic opportunities to construct “reverse chokepoints.” In response to chip embargoes, China imposed export controls on gallium and germanium beginning in August 2023—two rare metals indispensable to semiconductors, radar systems, and optical communications equipment. Following the controls, China’s gallium exports fell to zero in certain months, while global market prices surged by more than 150 percent, imposing significant material shortages on Western semiconductor and defense firms dependent on Chinese supply chains [14]. The logic of this countermeasure is not mere retaliation, but the demonstration of “asymmetric destructive capacity”: while unable to manufacture cutting-edge chips, China can control upstream lifelines essential to chip production. By converting resource advantages into geopolitical leverage, China has successfully constructed “reverse deterrence” within an asymmetric interdependence network. This “materials-for-technology” bargaining logic reveals the fluidity of power in interdependent networks: control over key nodes enables even technologically downstream states to exert asymmetric balancing forces in specific domains, compelling adversaries to weigh the supply chain costs of escalation.

III. Mechanisms of Structural Power Generation: The Migration of Power from Finance to Computing Power

3.1 Theoretical Foundations: Chokepoint Effects and Panopticon Effects

The core mechanism of weaponized interdependence lies in the reconstruction of structural power—that is, power derived not from direct material coercion, but from reshaping the choice environment of others. Within Farrell and Newman’s framework, such power is projected through two primary effects in asymmetric networks: the “chokepoint effect” and the “panopticon effect.” The chokepoint effect grants central nodes the ability to sever flows, as exemplified by U.S. monopolistic control over EDA software and semiconductor manufacturing equipment to cut off access to advanced process computing power; the panopticon effect grants privileged access to information, such as monitoring global financial flows through SWIFT or enforcing “know your customer” (KYC) rules via cloud platforms like AWS and Azure to infer adversaries’ AI computing intentions and R&D progress [2]. These effects are deeply intertwined in the AI era: panoptic surveillance enables precise identification of targets for chokepoint enforcement, while chokepoint threats force actors into monitored channels, further reinforcing surveillance capacity.

Structural power generation also depends on the synergistic operation of three forms of power: supply-based power, channel-based power, and market-based power. Supply-based power derives from monopolies over irreplaceable resources such as high-end GPUs and rare earths; channel-based power from control over circulation networks such as submarine cables and cross-border data regimes; and market-based power from access barriers to massive consumer markets [6][10]. Unlike the financial domain’s reliance on single-channel control (e.g., SWIFT disconnection), AI-era power structures are multidimensional, combining hard constraints (hardware embargoes) with soft constraints (standards-setting and ethical norms). This multidimensionality transforms AI competition from linear confrontation into systemic contestation.

3.2 Asymmetry and Inevitability: The Double Helix of Cooperation and Conflict

Weaponized interdependence is not an accidental policy choice, but a structural inevitability rooted in extreme asymmetries of resources, power, and circumstances. In the global AI industrial chain, key factors—computing power, data, algorithms, and capital—exhibit power-law distributions, providing fertile ground for power extraction. When one actor holds overwhelming advantages at critical nodes (such as sub-7nm fabrication processes), the temptation to convert economic advantage into geopolitical leverage becomes nearly irresistible. As Keohane and Nye observe, asymmetric interdependence itself constitutes a source of power, and under great-power competition, such asymmetries are rapidly securitized into strategic instruments [7].

More fundamentally, a “cooperation–conflict” double-helix dynamic is at work. While globalization-era interdependence was expected to raise conflict costs and promote peace, deep cooperation in the AI domain paradoxically creates more points of vulnerability. Each unified technical standard, submarine cable connection, and open-source community enhances diffusion while simultaneously creating new levers for weaponization. This cumulative “stacking blocks ever higher” effect generates a dangerous dynamic: as systems grow more complex and nodes more concentrated, mutual distrust drives preemptive control of key nodes, trapping actors in a security dilemma spiral in which “more sanctions yield less security, and more countermeasures yield more confrontation.” This reveals the alienation of interdependence in the absence of institutional guardrails—it ceases to be a stabilizer and becomes a catalyst for conflict.

3.3 Amplification Effects in Applying These Mechanisms to AI

When structural power migrates from finance to the AI domain, its coercive force is significantly amplified due to AI’s distinctive features of technological lock-in and network effects. First, AI industries exhibit strong technological lock-in. NVIDIA’s CUDA ecosystem, accumulated over more than a decade, has become foundational infrastructure for AI developers. This integrated hardware–software lock-in is far more difficult to escape than financial dependence, as switching costs encompass not only capital investment but also community migration and codebase restructuring [11]. Once weaponized, such ecosystems impose “shock therapy” disruptions rather than mere cost increases.

Second, winner-takes-all network effects further amplify central-node power. Through positive feedback loops among data, algorithms, and computing power, leading platforms and infrastructure providers reinforce their dominance over time, rendering market-based catch-up increasingly difficult. This allows hegemonic states to achieve disproportionate impact with relatively low administrative costs, producing “four ounces to move a thousand pounds” effects. Moreover, as a general-purpose technology with blurred civil–military boundaries, AI cloaks weaponization in heightened legitimacy; export controls justified by “national security” face far less normative resistance than overt protectionism. Consequently, structural power in the AI domain is not only more destructive, but also more concealed and resilient.

IV. Strategic Offense and Defense: Structural Competition over Computing Power and Network Nodes between China and the United States

4.1 Beyond Countermeasures: Symmetrical Weaponization of Structural Power

In the strategic chessboard of AI competition, Sino–U.S. rivalry has transcended linear “sanctions versus counter-sanctions” logic, evolving into a contest of “symmetrical weaponization of structural power.” Traditional perspectives often frame one actor as aggressor and the other as defender, but in a highly interconnected AI network, both sides actively exploit advantageous positions in network topology to construct integrated offensive–defensive strategies. This competition is no longer zero-sum but dynamically balanced: each offensive move (such as chip embargoes) triggers defensive offensives (such as material controls), producing a spiraling dynamic. Great-power rivalry thus shifts from territorial or resource contests to struggles for control over “chokepoints” in global technological networks. While the United States seeks to maintain technological hegemony through supply- and channel-based power, China leverages market power and infrastructure construction capacity to build asymmetric counterbalances, transitioning from passive retaliation to proactive defense [9].

4.2 Supply-Based Power Offense and Defense: Computing Power Blockades and Material Countermeasures

U.S. supply-based power offensives manifest as comprehensive blockades of the physical foundations of AI computing power. Through successive BIS export control regimes, the United States restricts not only direct exports of high-performance chips such as the NVIDIA A100 and H100, but also leverages monopolies over EDA software and semiconductor manufacturing equipment to erect a “technological iron curtain.” The strategic intent is to employ chokepoint effects to sever China’s upward technological mobility, confining it to mature-process domains. The 2025 implementation of AI Technology Diffusion Control Regimes further extends this blockade to algorithmic weights and cloud-based computing power, aiming for multidimensional suppression [12].

China’s response eschews fully symmetrical confrontation in chip fabrication, instead exploiting its hub position in critical material supply chains. Export controls on gallium, germanium, and rare earth technologies constitute effective countermeasures against U.S. supply-based coercion. This strategy’s sophistication lies in its exploitation of structural vulnerabilities in the global semiconductor chain: while the United States dominates top-tier design and equipment, its foundation remains dependent on stable rare-metal supplies. By converting asymmetric dependence into symmetrical harm potential, China increases the systemic costs of U.S. sanctions. This demonstrates that even technologically downstream states can wield structural deterrence by controlling seemingly peripheral nodes [8].

4.3 Channel-Based Power Offense and Defense: Digital Containment and Infrastructure Breakthroughs

In channel-based power competition, the United States pursues a “digital containment” strategy aimed at excluding China from core global data and computing networks. This strategy is epitomized by the “Clean Network” initiative, through which the United States pressured allies to exclude HMN Tech from key submarine cable projects such as SEA-ME-WE 6 and reroute trans-Pacific cables away from Chinese territory. This is not merely physical isolation, but a restructuring of data flow pathways designed to deprive China of global data hub status. Proposals to impose KYC rules on cloud computing services further aim to establish a global digital panopticon, monitoring and restricting Chinese entities’ access to U.S.-based computing power [15].

China’s responses exhibit resilience and substitutability. Domestically, the Data Security Law and Personal Information Protection Law establish data sovereignty barriers to prevent core data leakage. Externally, China promotes the “Digital Silk Road,” supporting firms such as Hengtong Optic-Electric in constructing alternative infrastructure like the PEACE cable along Belt and Road routes. While these self-built channels cannot fully replace existing systems in the short term, they structurally erode U.S. channel monopolies. Over time, this competition may transform global networks from single-center to multi-center configurations, reducing the destructive potential of weaponized nodes [10].

4.4 Market-Based and Narrative Power Offense and Defense: Access Barriers and Legitimacy Competition

Market-based power competition centers on capital flows and market access. U.S. Executive Order 14105 restricts U.S. investment in Chinese semiconductor, quantum computing, and AI sectors, seeking to sever capital lifelines [16]. China counters by leveraging its vast consumer market, conducting cybersecurity reviews of U.S. firms such as Micron, signaling that market access is conditional. This contest ultimately concerns control over the gravitational center of the global tech ecosystem—who defines security boundaries and who determines participation in future technological orders.

Simultaneously, narrative power competition intensifies. The United States frames its actions as “de-risking” rather than decoupling, legitimizing controls as national security imperatives and casting technological rivalry as ideological contestation between democracy and authoritarianism. China counters with narratives of digital sovereignty and development rights, criticizing “small yard, high wall” strategies and technological hegemony, appealing particularly to the Global South [8]. These narrative battles shape global governance norms and alliance alignments; long term, narrative legitimacy may determine whose standards and security concepts become embedded in future institutions.

V. Conclusion: Mutual Constraint as a Structural Choice

5.1 Structural Fragility beneath the Surface of Prosperity

This study reveals a troubling reality: AI-driven global economic prosperity rests on an extremely fragile structural foundation. This fragility is not a transient policy friction but arises from the alienation of global division-of-labor networks. In pursuing efficiency maximization, extreme concentration of computing power, data, and capital has created central nodes wielding disproportionate power. As Farrell and Newman predicted, once political will captures these nodes, chokepoint and panopticon effects are rapidly weaponized [1][2]. Sino–U.S. competition over chips, submarine cables, and materials demonstrates that interdependence has ceased to guarantee peace and instead serves as a medium for conflict. Structural inequalities inevitably drive advantaged states to extract geopolitical rents, provoking countermeasures that push the global technology system toward fragmentation.

5.2 From Asymmetric Dependence to Symmetrical Deterrence

By extending weaponized interdependence theory from finance to AI, this article identifies a multidimensional pattern of competition. Unlike single-channel SWIFT sanctions, AI conflict spans hardware supply, network channels, market access, and narratives. Crucially, even asymmetrically disadvantaged states can construct “reverse chokepoints” through upstream material control or market scale, generating symmetrical deterrence. This shift marks a transition from unilateral hegemonic coercion to reciprocal harm, enriching understandings of structural power and reframing Sino–U.S. structural contradictions [11].

5.3 De-weaponization Pathways and Institutional Guardrails

In an era where weaponized interdependence is normalized, full de-weaponization is unrealistic. A more viable path lies in limited counter-weaponization and institutional guardrails that bound competition. Supply-chain diversification and de-risking can mitigate chokepoint impacts at efficiency costs; channel redundancy through regional connectivity projects can expand strategic maneuverability; institutionally, “technological arms control” mechanisms—such as treating civilian submarine cables and basic computing infrastructure as protected zones—may establish red lines against extreme weaponization. Chinese scholarship on counter-weaponization and guardrails suggests such arrangements are pragmatic tools for minimal cooperation amid distrust [8].

5.4 Practical Implications: Seeking Guardrails amid Mutual Assured Destruction

Under escalating structural conflict, mutual constraint becomes a rational choice short of full decoupling. Analogous to Cold War nuclear deterrence, a form of “mutual assured economic destruction” is emerging, sustaining a fragile balance. Yet this is unsustainable: escalating sanctions erode global innovation efficiency and amplify systemic risks. Future pathways may require technological arms control or guardrails that define neutral boundaries for critical infrastructure, preventing “structural nuclear strikes.” In an era of low trust, mutual constraint may be the least-bad option—but preventing it from cascading into systemic collapse remains the foremost challenge for international politics and global governance.

References

Farrell, H., & Newman, A. L. (2019). Weaponized Interdependence: How Global Economic Networks Shape State Coercion. International Security, 44(1), 42–79.
Farrell, H., & Newman, A. L. (2021). The Janus Face of the Liberal International Information Order: When Global Information Networks Becomes a Tool of State Power. International Studies Quarterly, 65(4), 977–990.
Bown, C. P. (2020). How the United States Marched the Semiconductor Industry into its Trade War with China. Peterson Institute for International Economics Working Paper, 20-16.
Gupta, R., Walker, L., & Reddie, A. W. (2025). Whack-a-Chip: The Futility of Hardware-Centric Export Controls. arXiv preprint arXiv:2411.14425.
Li, X., & Zhang, Y. (2025). Escaping a Weaponized Network: China’s Reaction to the United States Escalating Technology Controls. Journal of Chinese Political Science, 30(1), 1–25.
Strange, S. (1988). States and Markets. London: Pinter Publishers.
Keohane, R. O., & Nye, J. S. (1977). Power and Interdependence: World Politics in Transition. Boston: Little, Brown.
任琳. (2024). 见招拆招：相互依赖武器化态势下的反武器化. 世界经济与政治, (4), 55–80.
丁一凡. (2024). 武器化相互依存下的中美算力与网络博弈. 学术前沿, (5), 23–35.
张宇燕. (2024). 全球经济治理中的结构性权力. 世界经济与政治, (2), 1–25.
王缉思, 李侃如. (2023). 中美关系的结构性矛盾与相互依赖. 国际问题研究, (3), 1–20.
U.S. Department of Commerce. (2025). Framework for Artificial Intelligence Diffusion and Export Controls. Federal Register, 90(15), 1234–1245.
Center for Strategic and International Studies (CSIS). (2024). Red Sea Cable Damage Reveals Soft Underbelly of Global Economy. CSIS Critical Questions.
Reuters. (2025). China suspends ban on exports of gallium, germanium, antimony to US. Reuters Technology News, Nov 09.
Reuters. (2024). US proposes requiring reporting for advanced AI, cloud providers. Reuters Technology News, Sep 09.
U.S. Department of the Treasury. (2024). Provisions Pertaining to U.S. Investments in Certain National Security Technologies and Products in Countries of Concern. Federal Register, 89(209), 8567–8580.