For almost fifteen years, one belief seemed settled: the future of enterprise IT lay in the cloud. The cloud promised agility, scalability, innovation, and a reduction in operational complexity. For many organizations, it became the natural foundation of their digital transformation. Yet the rise of artificial intelligence is now reshaping the landscape.
For the first time in a long while, senior executives and CIOs are revisiting questions once thought settled: where should tasks be executed? Where should data be hosted? How can costs be controlled? How can dependence on hyperscalers be reduced? How can the sovereignty of the most strategic digital assets be guaranteed?
AI does not overturn the cloud. However, it clearly puts infrastructure at the center of strategic decision-making.
The return of a topic once considered closed
In recent years, infrastructure gradually became an execution concern. Business leaders spoke about user experience, digital products, or data. IT teams tended to keep the platforms running.
AI fundamentally alters this dynamic.
Unlike traditional applications, AI models demand substantial computing resources. Training, as well as inference, require GPU capacity, storage, high-bandwidth networks, and architectures that can scale rapidly.
Added to this are new components: vector databases, data pipelines, orchestration platforms, specialized models, or agent infrastructures.
Put differently, every new AI use case now has a direct translation into the required infrastructure.
The economic equation is changing
And one of the first visible effects concerns costs. For a long time, the cloud allowed transforming hardware investments into more flexible operating expenses. That logic remains valid in many scenarios.
But AI shifts this equation: a company deploying dozens of agents, querying multiple models, processing millions of documents, or running continuous inferences quickly discovers that compute power becomes a strategic line item in the budget.
The challenge is no longer solely about the price of a server or a virtual machine. It now requires balancing between several business models: using a managed service, running an open-source model on your own infrastructure, pooling GPU resources, or intelligently distributing workloads according to their criticality.
These trade-offs become strategic decisions as much as technical ones.
The future will be hybrid
Should we conclude that the shift toward the cloud will reverse? Probably not. Yet the debate is no longer simply cloud versus on-premise. The future will be more nuanced.
Some workloads will naturally continue to run in the public cloud. Others will make more sense on private, sovereign, or specialized infrastructures. Some companies will favor hybrid models capable of dynamically weighing costs, performance, privacy, or regulatory constraints.
The objective will no longer be to centralize everything in one place, but to place each workload where it creates the most value.
This evolution places CIOs at the center of decision-making. The choices made today regarding architecture, data governance, sovereignty, compute capacity, or interoperability will directly shape a company’s ability to industrialize its AI initiatives tomorrow. In other words, infrastructure ceases to be a mere enabler of innovation and becomes one of its main drivers.
The recent arc of digital technology shows that major technological ruptures always redefine the fundamentals of enterprise IT. AI is no exception.
The real change may not be that AI becomes ubiquitous. It is that it compels companies to reconsider what they had come to accept as given: their infrastructure.
* Thomas Thelliez is Chief Technology Officer at Positive.
