Artificial intelligence does not represent a simple technological extension of public action, but a structural factor that redefines the material conditions of the exercise of state power. In the post AI economy, a state’s ability to govern markets, institutions and social processes increasingly depends on access to, control over and integration of advanced computational infrastructures. Like electricity in the twentieth century or transport networks in the nineteenth century, digital infrastructure becomes an essential component of state capacity, directly affecting sovereignty, continuity of public action and the ability to shape the economic future.

The transformation is first and foremost infrastructural. Cloud computing, data centers, artificial intelligence models and data networks now constitute the operational substrate on which fundamental public functions rest. According to estimates by the International Energy Agency, by 2030 data centers will account for more than 3 percent of global electricity consumption, with marked geographic concentration in a few areas of the world. At the same time, analyses by the McKinsey Global Institute indicate that more than 60 percent of AI applications used by governments rely on cloud infrastructures managed by a limited number of large global operators. This infrastructural concentration introduces a new form of technological dependence that affects not individual services, but the very capacity of the state to function autonomously.

Dependence emerges with particular clarity in critical sectors. Tax systems based on predictive analytics, health platforms using automated triage models, security and urban mobility management tools founded on real time data analysis become progressively indispensable. However, when such systems are developed, maintained or hosted by external private actors, often subject to different jurisdictions, state sovereignty takes on a fragile character. World Bank reports show that many states lack the internal expertise required to fully audit the models in use, as well as the capacity to intervene rapidly in the event of service interruptions or unilateral changes in usage conditions. Public power thus risks being transformed into an advanced user of external infrastructures, rather than the architect of its own decision making space.

This fragility is not only technical, but economic and political. Competition among states in the AI era is played out through the ability to build complete ecosystems that include computing power, data supply chains, scientific expertise and integration into decision making processes. According to OECD data, more than 80 percent of global public investment in AI infrastructures is concentrated in fewer than ten countries. These states not only attract capital and talent, but define technical standards, governance practices and usage models that tend to spread globally. Other countries find themselves in a position of functional dependence, compelled to adopt solutions designed elsewhere and to adapt their institutions to infrastructural logics they do not fully control.

The economic consequences of this asymmetry are significant. Privileged access to computational capacity makes it possible to experiment with public policies based on advanced simulations, reduce uncertainty in macroeconomic management and intervene more precisely in markets. By contrast, states lacking adequate infrastructure operate under conditions of permanent informational delay, with direct effects on the competitiveness of domestic firms and on the ability to attract investment. In an economy in which value is increasingly linked to forecasting and coordination, the absence of computational power translates into a loss of strategic capacity.

In this context, the concept of functional sovereignty emerges. This does not mean pursuing an improbable technological autarky, nor renouncing international cooperation. Functional sovereignty instead refers to the capacity to guarantee continuity, reliability and control over essential state functions, even when these are based on complex and global technologies. It means being able to understand how the systems in use function, assess their effects, modify or replace them without compromising public operations. Some recent experiences, such as European investments in high performance computing infrastructures or national public cloud programs in East Asia, show how states are beginning to recognize digital infrastructure as a long term strategic asset.

Reflection on computational sovereignty recalls a broader conception of the nature of the state. Hannah Arendt observed that power does not reside simply in the capacity to command, but in the ability to act in a coordinated and durable manner over time. In the economy of artificial intelligence, this possibility passes through infrastructure. Without adequate computational foundations, public action becomes fragmented, reactive and dependent on external logics. AI does not automatically weaken the state, but tests its deep architecture, making visible the lines of fracture between those who design infrastructures and those who use them.

The parallel with the corporate world offers useful insights. Firms that have fully outsourced their critical digital functions have often discovered that they lost strategic control, process visibility and adaptive capacity. By contrast, those that invested in their own infrastructures, while cooperating with external partners, have preserved real decision making autonomy. The same logic applies to the state. Public leadership in the digital era is not measured by the quantity of technologies adopted, but by the capacity to integrate them into a coherent vision that preserves control over essential functions.

Artificial intelligence, ultimately, is not merely a tool of administrative efficiency, but a new terrain of economic and institutional power. To speak of state capacity in the AI era means recognizing that computational infrastructure has become a necessary condition of political and economic autonomy. In a world driven by data and prediction, the true difference among states lies not in the speed with which they adopt new technologies, but in their ability to remain invisible yet decisive architects of their own digital power.

Global AI Observatory