The economy of artificial intelligence is structured around infrastructures that no longer perform a merely technical function, but constitute the material backbone of contemporary economic power. In this configuration, growth does not eliminate dependencies, but concentrates them. Digital infrastructures, particularly advanced cloud and computational systems, become points of coagulation for value, decision making, and systemic vulnerability. Global competition thus shifts from control of traditional resources to the ability to govern invisible architectures whose functioning conditions entire productive and institutional sectors.
The cloud represents the core of this transformation. According to data from the Synergy Research Group, more than 65 percent of the global market for infrastructural cloud services is controlled by three operators. This concentration is not only a matter of market share, but a structural element that defines technological standards, pricing models, and access to innovation. Artificial intelligence, by its data and computation intensive nature, further intensifies this dependency. Training the most advanced models requires computational resources that only a few platforms can provide on a continuous basis, making access to infrastructure a preliminary condition for competitiveness.
The centralization of computational capacity produces a cumulative effect. Advanced skills, data flows, and development opportunities tend to cluster around the same infrastructural nodes, reinforcing the center and progressively emptying the periphery. This process recalls dynamics already known in industrial economies, but with a substantial difference: speed. In the case of AI, concentration occurs much more rapidly, leaving little room for gradual adaptation strategies. In the corporate world as well, the centralization of key functions improves efficiency in the short term, but exposes organizations to critical dependence on a small number of control points.
From this configuration derive systemic risks that are difficult to neutralize. Service disruptions, geopolitical tensions, or regulatory constraints affecting central infrastructures can rapidly propagate along global value chains. In 2023, a prolonged outage at a major cloud provider caused simultaneous disruptions across financial, logistics, and healthcare sectors in multiple regions of the world, demonstrating how centralization transforms a technical incident into a systemic shock. Artificial intelligence, by increasing the interconnection between decision making and operational processes, further amplifies this exposure.
The geopolitical response to such fragility takes the form of fragmentation. States and economic blocs attempt to reduce dependency by promoting alternative infrastructures or data localization policies. However, this push does not eliminate the problem, but reorganizes it. Partially interoperable ecosystems emerge, in which access to infrastructure becomes a political as well as economic variable. For global enterprises, this means operating in an environment where infrastructural continuity is no longer guaranteed solely by financial capacity, but depends on unstable geopolitical balances.
In this context, the reflection of Ivan Illich appears particularly relevant. Illich showed how institutions, once they exceed certain thresholds of scale and complexity, tend to invert their original function, transforming from instruments of emancipation into devices of dependency. The digital infrastructures of AI embody this paradox: designed to offer flexibility and widespread access, they end up concentrating power and decision making constraints in a few central nodes. Infrastructure does not impose, but orients, rendering some choices feasible and others simply impracticable.
For enterprises, this configuration poses a strategic issue that goes beyond efficiency. The choice of infrastructure is no longer neutral, but defines the perimeter of future autonomy. According to McKinsey analyses, more than 70 percent of companies that have migrated critical functions to centralized cloud systems report significant difficulties in restoring alternative solutions within short time frames. The reversibility of infrastructural decisions thus becomes a key indicator of resilience, often neglected in short term evaluations.
Infrastructural dependency also raises a problem of responsibility. When essential services are outsourced to a small number of global actors, the chain of responsibility fragments. In the event of failure or disruption, the ability to attribute decisions and consequences weakens, generating grey areas of governance. This phenomenon is not marginal: according to the European Union Agency for Cybersecurity, lack of clarity in infrastructural responsibility is one of the main sources of systemic risk in advanced digital economies.
Geopolitical fragmentation adds a further layer of complexity. Operating on infrastructures subject to divergent regulatory regimes requires a level of coordination that not all organizations possess. Global competition thus shifts from technological innovation alone to the management of infrastructural complexity. This dynamic recalls, in amplified form, the historical difficulties faced by multinational enterprises in reconciling global integration with local adaptation.
Dependence on centralized digital infrastructures is not a temporary deviation, but a structural feature of the artificial intelligence economy. The challenge is not to eliminate it, but to render it governable. This implies an explicit assessment of acceptable dependencies and those that compromise the capacity for choice. Efficiency and resilience thus enter into a permanent tension that cannot be resolved once and for all, but only managed over time.
In this unstable space, artificial intelligence makes visible a deeper transformation. The center, precisely as it concentrates capacity and value, becomes a point of systemic fragility. The new geometry of power is not based on dispersion, but on architectures that require a higher degree of strategic awareness. To inhabit this fragility means recognizing that the future of the economy will depend not only on the accumulation of computational power, but on the ability to build infrastructures that do not transform the center into a permanent vulnerability.
Global AI Observatory