The rise of artificial intelligence is reshaping the cards for digital infrastructure. Behind the promises of performance, another reality is imposed on general management…
controlling costs, supplier dependence and technological trade-offs. Digital sovereignty also involves the ability to regain economic control.
For years, infrastructure choices have primarily responded to a logic of speed. It was necessary to migrate, deploy, industrialize, sometimes without fully measuring the long-term economic consequences. This approach has made it possible to accelerate digital transformation, but it has also installed architectures that are costly, complex and sometimes difficult to reverse.
Today, this equation is changing. With AI, the question is no longer just about the available power, but about its real cost. Infrastructure is no longer a simple support; it is becoming a strategic position, capable either of fueling innovation or of absorbing a growing share of IT budgets.
Waste that has become visible
Lessons from the study Digital Sovereignty Trilemma are, in this respect, enlightening. European companies estimate that they waste on average 24% of their annual Cloud capacity. Part of this excess capacity is assumed to guarantee resilience. But the main thing is clearly identified weaknesses, namely unused resources, lack of visibility and insufficient governance.
This observation must be taken seriously. Waste is not only a subject of financial optimization. It reduces companies’ room for maneuver, weakens their investment capacity and further exposes them to supplier price volatility. As AI uses develop, this loss of control becomes a management risk.
The illusion of the cheapest
The increase in hosting costs linked to AI confirms this pressure. In many companies, it already leads to reallocating budgets initially planned for other projects. In other words, the innovation promised by AI can, if poorly anticipated, be financed to the detriment of other technological priorities.
In this context, the FinOps discipline is necessary, but it is not sufficient. Correcting the invoice a posteriori does not compensate for poorly made upstream architectural decisions. The real challenge is to put the economic analysis back at the moment of choice: where to run a workload, according to what cost model, with what level of portability, and with what future dependency?
This is the current paradox. The cheapest choice in the short term is not always the most rational in the long term. More and more companies are agreeing to invest more in certain infrastructures to meet sovereignty, compliance or reversibility requirements. This apparent additional cost may in reality constitute a form of protection against much heavier future costs.
Take back control
Economic sovereignty therefore presupposes three reflexes. First, systematize complete cost analyzes before any structuring decision. Then, design portable architectures, in order to limit technological confinement. Finally, address inefficiencies at the root, at the level of architecture and governance, rather than trying to compensate for them once the expense has been incurred.
The question posed to financial and IT departments is now simple: do AI investments prepare a lasting advantage, or do they install a cost model that is not very reversible? In the digital economy, sovereignty is no longer just about compliance or localization. It begins with the ability to keep control of its choices, its margins and its innovation trajectory. However, in AI projects, this control cannot be thought of without flexibility: companies are still moving forward in a changing landscape, marked both by technological acceleration and by limited visibility on real functional needs. The challenge is therefore not only to reduce the bill, but to maintain the freedom to adjust its architectures and investments as uses become clearer.
