The economy of artificial intelligence is not organized around tangible goods or simple traditional factors of production. Its structural axis is constituted by data, understood not as neutral information, but as a strategic resource capable of generating cumulative advantages, power asymmetries, and new forms of systemic dependence. In this sense, data today perform a function analogous to that played by coal, oil, or steel in other historical phases, with one decisive difference: their extraction, circulation, and valorization occur in a continuous, diffuse, and often invisible manner, making the resulting power relations less legible.

The first structural element is the control of information flows. Data acquire value only when they can be collected systematically, integrated into large coherent sets, and transformed into predictive capacity. This process is not evenly distributed. According to estimates by UNCTAD, more than 60 percent of the economic value generated by digital data globally is concentrated in an extremely small number of large platforms, located primarily in North America and East Asia. Power resides not so much in the legal ownership of data as in the capacity to control the infrastructures that regulate their flow, standardization, and access. Those who control these nodes implicitly define what is visible, measurable, and monetizable.

From this configuration derives a process of asymmetric accumulation that tends to reinforce itself. Artificial intelligence systems improve as available data increase, and performance improvements enable the attraction of additional information flows. This dynamic is well documented also in the industrial domain: according to the OECD, data-intensive firms show productivity levels significantly above the average, with gaps that grow over time rather than narrowing. Initial advantage thus becomes a structural barrier to entry, making competition increasingly difficult not at the level of isolated innovation, but at the level of access to the informational raw material.

Unlike other resources, data are not consumed through use. Each processing cycle increases their potential value, strengthening the position of those who control them. This grants large data-centric actors a capacity to orient markets that goes beyond operational efficiency. Economic power manifests itself in the ability to anticipate behaviors, shape preferences, and influence collective decisions. In systemic terms, data become a form of capital that generates increasing returns and is difficult to redistribute through traditional market mechanisms.

This configuration makes the issue of digital sovereignty unavoidable. When data become economic infrastructure, their control acquires institutional significance. According to the European Commission, more than 70 percent of industrial data generated in Europe are processed on extra-European infrastructures, creating a structural dependence that concerns not only cybersecurity, but the capacity to orient industrial policies, innovation models, and long-term strategic choices. Sovereignty does not coincide with digital autarky, but with the ability to define conditions of access, use, and valorization consistent with collective objectives.

The thought of Michel Serres offers a particularly suitable interpretative key for this scenario. Serres described contemporary power as the capacity to govern flows rather than to possess objects. In the data economy, this intuition finds empirical confirmation. Power is not static, but relational; it does not reside in blocking information, but in channeling it. Governing data means inhabiting passages, deciding which connections are possible and which are rendered impracticable.

In the corporate world, this transformation imposes a profound revision of decision-making models. The massive use of data promises to overcome human cognitive limits, but introduces a new opacity. Algorithms trained on enormous volumes of information produce results that are difficult to interpret, shifting decision-making power toward those who control system architectures rather than those who formally take decisions. Leadership is no longer only the capacity to choose, but the capacity to interrogate the informational premises that make certain choices possible and others invisible.

The asymmetric accumulation of data also produces significant social effects. Communities, workers, and users who daily contribute to the generation of information flows rarely participate in the distribution of the resulting value. According to World Bank analyses, a growing share of global digital wealth is extracted without a corresponding local return, fueling new forms of inequality that no longer follow traditional lines of fracture between capital and labor, but those between inclusion and exclusion from high-value informational circuits.

Digital sovereignty, in this context, is not a symbolic claim but a matter of systemic balance. Defining rules on data use means influencing the distribution of economic power and the capacity of institutions to orient development. Data regulations do not act solely as instruments of protection, but as levers of industrial and social policy, capable of shaping the future geography of the global economy.

Data-driven decision-making processes ultimately pose a profound cultural challenge. Informational abundance can generate an illusion of completeness, obscuring the fact that every datum is the result of a selection and a context. Organizations that entrust their strategy exclusively to the predictive power of models risk losing the ability to recognize what remains outside the measurable perimeter. Data do not eliminate judgment, but increase its responsibility.

Considering data as a strategic raw material therefore means recognizing that the economy of artificial intelligence is neither neutral nor spontaneous. It is the product of infrastructural, institutional, and political choices that determine who accumulates value, who exercises influence, and who remains dependent. In this new global geography, the challenge is not to renounce data, but to govern their use with systemic awareness. Like every decisive raw material, data require a policy, not only a technique. It is in this distinction that a significant part of the future of the economy and institutions in the era of artificial intelligence will be decided.

Global AI Observatory