In contemporary capitalism, economic power is progressively shifting from the visible sphere of ownership and exchange toward less immediately observable domains, linked to the capacity to model complex decision-making processes. Advanced artificial intelligence acts as a systemic infrastructure that enables the governance of information flows, the anticipation of behavior and the reduction of uncertainty on an extended scale. In this transition, power no longer derives solely from control over the means of production or financial markets, but from the ability to intervene upstream of decisions, influencing the very conditions under which such decisions become possible. The center of gravity of the economy thus shifts toward those who control data, computational capacity and predictive models.

The concentration of economic power in the algorithmic economy is not an accidental effect, but a structural dynamic. The costs of developing, training and maintaining the most advanced artificial intelligence systems are high and increasing. According to estimates by the OECD and the International Energy Agency, the training of large-scale models requires investments in computing infrastructure and energy consumption that are sustainable only for a limited number of global actors. This material constraint produces an upstream selection, favoring firms and organizations capable of sustaining continuous investment cycles. The result is not merely a reduction in competition, but its qualitative reconfiguration, in which the market tends to take the form of closed ecosystems governed by platform logics and cumulative rents.

In this context, inequality no longer follows exclusively the traditional lines between capital and labor or between large and small firms. A technological and epistemic fracture emerges. Organizations that integrate artificial intelligence as a strategic lever can optimize processes, anticipate trends and systematically reduce risk. Those that remain at the margins are forced to react belatedly or to depend on external infrastructures, accepting operational conditions defined by others. This asymmetry tends to be self-reinforcing. The more a system is used, the more data it generates, improving its performance and strengthening its initial advantage. Computational power becomes cumulative, making the entry of new actors increasingly difficult and transforming competition into a structurally unequal confrontation.

The effects on the distribution of value are significant. When decision-making capacity concentrates in the central nodes of algorithmic infrastructure, the remuneration of labor and peripheral activities tends to be compressed. Recent reports by the International Monetary Fund indicate that, in sectors with high intensity of automation and artificial intelligence, profit growth has significantly outpaced the growth of average wages. A growing share of the wealth produced remains concentrated among those who control decision systems, while along the economic chain forms of precarity and technological dependence spread. This phenomenon does not concern only workers, but also entire productive sectors and economic areas that can be excluded from accumulation circuits without an apparent decline in aggregate productivity.

The corporate world clearly reflects this transformation. The adoption of predictive systems makes it possible to reduce uncertainty and improve operational performance, but shifts decision-making power toward opaque models that are difficult to contest. Leadership increasingly coincides with the ability to interpret and validate algorithmic outputs rather than with the direct exercise of judgment. Authority moves from the person to the procedure, from deliberation to prediction. Studies by the Bank for International Settlements show how the widespread use of similar models in financial markets has contributed to growing homogeneity of behavior, increasing apparent efficiency but also the risk of systemic errors.

These dynamics raise questions that go beyond the economic domain. Karl Polanyi observed how market self-regulation, if not accompanied by institutional countermeasures, could disintegrate the social bonds on which the economy itself is founded. Artificial intelligence driven capitalism reproduces this dilemma in an updated form. Systems that are highly efficient from a technical perspective generate side effects that affect social cohesion and institutional legitimacy. The difference compared to the past lies in the lower visibility of these mechanisms, hidden behind neutral interfaces, objective metrics and specialized languages.

The power asymmetries generated by AI are also reflected in collective decision-making processes. Those who control advanced tools of analysis and forecasting exert indirect influence over the choices of governments, institutions and markets. Artificial intelligence thus becomes an implicit political actor, capable of orienting priorities and allocations without assuming formal responsibility. For firms, this scenario introduces a responsibility that transcends the corporate perimeter. The competitive advantage obtained through AI is not neutral, but produces systemic effects that affect the distribution of power and opportunities.

Questioning these asymmetries does not imply a rejection of technological innovation, but a reflection on the conditions of its economic and institutional legitimacy. A system that concentrates power without rebalancing mechanisms risks eroding trust, which represents a fundamental resource for long-term stability. In the algorithmic economy, the central issue does not concern only the efficiency of intelligent systems, but their capacity to integrate into an economic and social order that preserves cohesion, decision-making plurality and shared responsibility.

Global AI Observatory