For years, the cloud was presented as an almost obvious trajectory: to migrate, gradually shedding private infrastructures to join the public cloud.
This framing lost relevance as information systems grew more complex and sovereignty, security and innovation concerns drew closer together.
In this context, hybridization is not a cautious compromise between the old world and the new; it signals a more nuanced approach, acknowledging that not all uses are meant to be addressed in the same environment.
For leaders, the question is therefore no longer to pick a side between private cloud and public cloud, but to design a path that enables access to advanced technologies without losing control over data and dependencies.
Vendor Diversification and Control: Two Distinct Logics
Confusion between multicloud and hybrid cloud remains common. Multicloud most often follows a diversification logic: multiplying providers, avoiding putting all eggs in one basket, gaining access to complementary services, or preserving leverage in negotiations.
This approach can be useful, but it does not, on its own, guarantee more freedom. The idea that workloads could be moved simply from one operator to another remains theoretical: platforms have their technical lock-ins, while contracts create economic dependencies.
Hybridization responds to a different logic: choosing the right environment for the right use, depending on the data’s criticality and the level of trust required, instead of pitting several clouds against each other.
AI Imposes a Trade-Off Between Power and Data Protection
The rise of AI makes these trade-offs more tangible. The infrastructures needed, notably around GPUs, are costly, evolve rapidly, and amortize over short cycles. For a company whose business is not IT, financing GPU infrastructure and computing capacity that evolves so quickly is unlikely to be a sustainable strategy.
Nevertheless, access to these capabilities becomes indispensable. Organizations want to develop models, process data, and create new services, but not all data can be exposed in any environment, especially when it concerns sensitive or regulated activities.
That is where hybridization makes all the difference: some companies have already launched data- or AI-driven use on American cloud platforms and are seeking to relocate a portion of these applications into a trusted environment; others, notably in banking, have migrated little to the public cloud, but must find complementary capabilities to their private cloud.
The challenge remains the same: to access innovation without abandoning control.
From Sensitive Use to the Trusted Perimeter
It is precisely when uses touch data, AI, or regulated environments that the notion of trust ceases to be theoretical. Because it goes beyond simply where data is located, it must rest on verifiable guarantees, whether it is the HDS requirements for certain health-related uses or SecNumCloud for the most sensitive environments, along with protection against certain extraterritoriality risks, security, and reversibility.
This reversibility cannot be merely a contractual clause. It depends on architectural choices, licenses, components used, and the organization’s ability to understand what it entrusts to its suppliers, because outsourcing has never meant relinquishing control.
This trust-based logic does not imply turning away from major cloud platforms. It rather leads to better defining their role within the overall architecture. Hyperscalers retain a major place for many industrialized services, and large organizations will naturally continue to rely on them for part of their use cases.
But this model is not intended to cover every situation. For sensitive, regulated, or mission-critical uses, hybrid cloud enables composing multiple environments without pitting innovation against control.
Eric Haddad is CEO of Numspot
