The economy of artificial intelligence introduces a structural rupture in the way sovereignty has historically been conceived. Decision making power is no longer organized primarily around control of territory or delimited resources, but along transnational technological infrastructures that cut across states, markets, and institutions. In this context, sovereignty does not disappear, but changes its nature: from a territorial prerogative to the effective capacity to orient complex systems operating beyond traditional legal borders.
The first critical node concerns strategic autonomy. In the age of AI, autonomy does not coincide with technological self sufficiency, but with the concrete possibility of choice. According to OECD data, more than 75 percent of global advanced cloud computing capacity is concentrated in fewer than ten industrial groups, with strong geographical localization. This concentration limits the decision making space of states and enterprises that depend on external infrastructures for critical functions, from data management to cybersecurity. Sovereignty, in this framework, becomes a graduated variable: not a binary attribute, but a set of negotiable margins.
The role of private actors is central in this metamorphosis. Large technology platforms do not operate as simple suppliers, but as architects of cognitive ecosystems that establish technical standards, interoperability models, and innovation timelines. In 2024, according to World Bank estimates, more than 60 percent of artificial intelligence projects in the public sector globally rely on infrastructures or models developed by private entities external to the state apparatus. Public decision making is thus exercised within predefined technical perimeters, which substantially condition the available options.
Industrial policies on AI attempt to respond to this asymmetry, but operate within a constrained space. Investments in semiconductors, data centers, and advanced research have increased significantly: the European Union has allocated more than 200 billion euros in public and private funds to strengthen its technological capacity over the next decade. However, these interventions do not eliminate systemic dependency, but aim to rebalance it. Technological sovereignty thus emerges as the capacity to reduce critical vulnerabilities, not as a return to total control.
The limits of national sovereignty become evident when observing informational and computational flows. Data, models, and digital infrastructures operate according to logics of scale that transcend state borders. Attempts to confine AI within exclusively national perimeters have shown high costs in terms of efficiency and innovation. Sovereignty therefore takes on a relational form, based on the capacity to negotiate interdependencies through technological alliances, shared standards, and multilevel governance. This is a logic already familiar to multinational enterprises, which balance global integration and local autonomy to preserve their decision making capacity.
In this context, the thought of Antonio Gramsci offers a particularly effective interpretive key. Gramsci showed how modern power is exercised through hegemony, that is, the capacity to render certain power arrangements natural. In the artificial intelligence ecosystem, hegemony manifests itself in the definition of technical standards, programming languages, and decision making architectures. Those who establish these elements do not merely impose operational solutions, but delimit the horizon within which choices become thinkable.
The parallel with the corporate world is direct. The most incisive leadership does not coincide with continuous intervention, but with the capacity to structure the context in which decisions emerge. Similarly, technological sovereignty is not measured by the quantity of regulation produced, but by the ability to influence the foundational conditions of innovation. AI driven decision systems amplify this dynamic: they reduce human cognitive limits, but can crystallize dependencies if the underlying architectures remain opaque.
Dependence on private infrastructures also raises a democratic issue. When essential functions are mediated by proprietary systems, transparency and accountability become difficult to guarantee. According to MIT Technology Review, less than 20 percent of AI systems used in public administration have fully operational independent audit mechanisms. Sovereignty, in this sense, also concerns the possibility of contesting automated decisions and attributing responsibility for them.
If artificial intelligence policies are to be structural, they must therefore move beyond the logic of technological catch up. The central issue is not replicating the capabilities of others, but deciding which forms of dependency are acceptable and which compromise decision making autonomy. This requires an integrated vision that brings together economics, law, and strategy, avoiding the illusion that sovereignty can be recovered simply by multiplying investments.
Sovereignty in computational economies appears as an unstable equilibrium. On the one hand, the need to participate in global ecosystems in order to remain relevant; on the other, the need to preserve spaces of real choice. This is not a tension destined to be resolved, but a permanent condition to be governed. In this space, sovereignty does not manifest as absolute control, but as the capacity for informed orientation and shared responsibility.
Artificial intelligence makes visible a profound transformation of contemporary power. Power no longer resides solely within borders, but in the infrastructures that cross them. Understanding this logic does not mean renouncing autonomy, but redefining it in terms compatible with an economy that operates beyond traditional space. It is in this understanding, more than in any formal proclamation, that the future of sovereignty in the age of artificial intelligence is decided.
Global AI Observatory