Opinion – AI Is Becoming the New Infrastructure of International Order
Martin Wählisch
Artificial intelligence is becoming one of the defining infrastructures of the twenty-first century. Recent developments make this increasingly clear. The United Nations released the first report of its Independent International Scientific Panel on Artificial Intelligence, laying the foundations for a permanent scientific institution to inform global AI governance. NATO reaffirmed artificial intelligence as a core enabler of future defence capabilities and operational readiness. At the same time, governments from Washington to Riyadh continued investing billions in AI infrastructure, recognising that compute, models and data have become strategic assets. These are not isolated policy developments. Together, they reveal a deeper transformation: artificial intelligence is becoming one of the fundamental infrastructures of international order.
Get E-International Relations delivered to your inbox, free of charge. As you sign up, consider becoming a paid subscriber to support our work.
International order has always been sustained not only by balances of power but by the infrastructures that enable states to govern, trade, communicate, defend themselves and cooperate. The railways and telegraph networks of the nineteenth century compressed time and space, allowing empires to coordinate administration, commerce and military power across continents. Industrial production transformed the balance of power, while oil and electricity underpinned the geopolitical order of the twentieth century. The internet then rewired global commerce, communication and diplomacy. Artificial intelligence belongs in this historical lineage. It is no longer simply another technology to regulate or another industry to promote. It is becoming a foundational infrastructure upon which states, markets and societies increasingly depend.
International Relations has always evolved alongside changes in the sources of power. Classical realists such as Hans Morgenthau understood power largely through military capability and statecraft. Kenneth Waltz shifted attention to the distribution of capabilities across the international system, while Robert Keohane and Joseph Nye demonstrated that institutions and complex interdependence could themselves become sources of influence. Perhaps the most enduring insight came from Susan Strange, who argued that the deepest form of power is structural: the ability to shape the systems within which others operate rather than simply influence individual decisions. Her four structures of power, security, production, finance and knowledge, remain strikingly relevant. Artificial intelligence increasingly cuts across each of them. It is transforming military capability, reorganising production, reshaping financial systems through automation and prediction, and redefining the creation and diffusion of knowledge through frontier models. AI should therefore be understood not simply as another source of power but as a new form of structural power that connects and reinforces them all.
At first glance, AI is in many respects a textbook case of structural power, and there are seemingly elements of IR theory to make sense of current developments. Yet treating AI simply as another source of structural power overlooks what distinguishes its emergence as infrastructure. Unlike many earlier strategic infrastructures, AI operates under conditions that fundamentally reshape how structural power is exercised.
First, control over key layers of the AI stack – from advanced semiconductor production to frontier compute – is concentrated in a remarkably small number of firms and jurisdictions. Frontier capabilities are overwhelmingly developed by private companies, while most states depend upon systems they cannot build, inspect or meaningfully adapt. Access to AI therefore becomes less a question of technological adoption than of dependence on infrastructures controlled by others.
Second, AI infrastructures are characterised by a level of opacity uncommon among earlier strategic systems. A bridge can be inspected and a financial ledger audited. By contrast, the outputs of frontier AI systems often cannot be independently verified without recourse to other complex computational systems. Trust consequently becomes embedded within the infrastructure itself as a technical requirement rather than remaining solely a political question of oversight and accountability.
Third, AI resists many of the conventional mechanisms through which strategic technologies have historically been governed. Nuclear technologies, for example, can be distinguished by measurable indicators such as enrichment levels that separate civilian from military use. AI offers no comparable handle for control. The same model weights may support language translation, scientific discovery and military targeting alike. Moreover, AI systems increasingly participate in judgment rather than merely processing information, shaping what analysts see, which options reach decision-makers and, ultimately, how political choices are made.
Taken together, these characteristics redirect analytical attention from who possesses the most capable AI systems to who controls the infrastructures upon which those capabilities depend. Concentrated control over computational resources, cloud infrastructure, frontier models and semiconductor supply chains increasingly translates into geopolitical leverage. AI is therefore becoming infrastructure not simply because it is widely used, but because it increasingly structures the distribution of power, dependence and autonomy within the international system.
This perspective becomes particularly clear in the widely discussed rivalry between the United States and China. Much commentary presents AI as a race to develop the most capable frontier models. Yet the more consequential competition concerns who will shape the infrastructure through which AI operates globally. Leadership will depend not only on technological breakthroughs but also on control over cloud computing, advanced semiconductors, technical standards, governance frameworks, research ecosystems and trusted international partnerships. The contest is not simply over innovation; it is over the infrastructure within which innovation takes place. Competition over innovation seeks advantage within an existing technological order, whereas competition over infrastructure seeks to shape the conditions under which innovation takes place. Those who control the technology architecture influence not only their own capacity to innovate but also who can innovate, on what terms and at what cost. This shifts the locus of power from relative technological performance to the ability to structure the strategic environment itself, making AI infrastructure a source of structural rather than merely competitive advantage.
The implications are equally significant for security and defence. NATO increasingly treats AI as a foundational capability embedded across intelligence, command and control, logistics, cyber defence and autonomous systems. AI is no longer viewed as another weapons technology but as part of the digital infrastructure upon which future military effectiveness depends. Strategic competition therefore extends beyond tanks, ships and aircraft to encompass compute capacity, semiconductor supply chains, advanced algorithms and trusted digital ecosystems. Military power is becoming inseparable from technological infrastructure. The focus of strategic competition is, hence, shifting from who possesses the most capable weapons to who controls the infrastructures upon which military capability increasingly depends.
Europe illustrates how power in the AI era has become increasingly multidimensional. Through the AI Act and its broader digital regulatory agenda, the EU has established itself as the world’s leading governance actor in AI. Yet it remains dependent on frontier models, hyperscale cloud providers and advanced semiconductor ecosystems largely developed elsewhere. This divergence between regulatory authority and technological capability demonstrates that influence in the AI era cannot be measured through innovation alone. It also depends on who establishes rules, develops technical standards, builds trusted assurance mechanisms and creates confidence in the infrastructures upon which AI deployment relies. Regulatory authority and infrastructural control are therefore related but distinct dimensions of structural power.
Perhaps the most significant implication concerns the nature of sovereignty itself. States remain the principal actors in international politics, yet their ability to exercise sovereign authority increasingly depends on access to privately owned computational infrastructure. Few governments possess the compute, cloud capacity or frontier models necessary to develop advanced AI independently. Instead, they rely upon a small number of technology firms that own and operate critical components of the global AI ecosystem. This creates new forms of strategic interdependence in which public authority increasingly relies upon private technological capability.
This explains why countries such as Singapore, Saudi Arabia, the United Arab Emirates and Azerbaijan have placed AI at the centre of long-term national strategies. They are not simply investing in another high-growth sector. They are embedding AI across public administration, industrial policy, healthcare, education and economic development because they recognise that AI is becoming part of the institutional infrastructure of the modern state. Their objective is not merely technological adoption but the development of national capacity through foundational digital infrastructure.
The emergence of AI as a foundational infrastructure of international order invites renewed attention to structural theories of power within International Relations. Rather than asking primarily which states possess the greatest capabilities, scholars must increasingly ask who controls the infrastructures through which those capabilities are generated, exercised and constrained. This does not replace traditional analyses of capabilities. Rather, it complements them by recognising that capabilities increasingly depend upon infrastructures whose ownership, governance and operation shape patterns of power, dependence and autonomy. Building on Susan Strange’s conception of structural power, AI demonstrates that control over foundational digital infrastructure is becoming an increasingly important source of international order in the twenty-first century.
Martin Wählisch is Associate Professor of Transformative Technologies, Innovation, and Global Affairs at the University of Birmingham, with joint appointments in the School of Government and the School of Computer Science. He is a member of the Centre for AI in Government (CAIG) and a Fellow of the Institute for Data and AI (iDAI), where he co-leads the Governance, Economy and Society Group. Previously, he served for over a decade with the United Nations, and advises governments and international organisations on AI governance and emerging technologies.
The shift from capability to infrastructure is the right analytical move, and it holds up better if pushed one layer further down. Infrastructure is not only ownership and governance. It is also physical, and its geography is considerably narrower than the language of cloud and stack suggests.
Below the model layer sits a set of very few nodes: EUV lithography from a single Dutch supplier, leading-edge logic concentrated in Taiwan, high-bandwidth memory in South Korea, advanced packaging capacity in a handful of plants, and, increasingly, grid interconnection as the binding constraint on new compute. Sovereign AI programmes of the kind pursued by Singapore, the UAE and Saudi Arabia buy accelerators and build national data centres, which changes who holds the asset. It does not change who can make it. The vulnerability sits upstream, at fabrication, tooling and power, and buying more of the finished good does not reach it.
That matters because the redundancy in this system is more correlated than it looks. A distributed cloud footprint reads as resilience on a map, but if the fabs, the packaging lines and the memory supply fall inside the same contingency, geographic dispersion is bookkeeping rather than insurance.
The European case makes the same point from the other side. The AI Act is real leverage, but it operates at the demand layer, and demand-layer instruments bind only while the regulated firm wants access to the market. Autonomy is a full-stack property. Holding rules without holding compute produces influence over conduct, not control over dependence.
Strange's knowledge structure is the right frame. Its physical floor, though, behaves less like a network and more like a chokepoint system, and the analytical vocabulary developed for maritime straits transfers with surprisingly little loss. A chokepoint is not simply a narrow place. It is a point where volume, substitutability and transit time collapse simultaneously, so that the cost of a closure is set by how long the rest of the system can run without it. Hormuz, Malacca and Bab el-Mandeb qualify on those terms. So do EUV tooling, sub-5nm logic and advanced packaging, on shorter timelines and with even thinner alternatives. Reading AI infrastructure this way changes the governance question from who writes the rules to who can hold a passage closed, and for how long before the dependent parties are forced to concede.
AI militarization, especially with ML is risky because it can trigger actions without human judgment, here we go again the problem with North South divide as well