The Cloud Is No Longer a Binary Choice
For years, the cloud question boiled down to “migrate or stay on-prem.” That era is over. The reality for businesses is now hybrid and multi-cloud: according to Flexera (2025), organizations on average use 2.4 distinct public clouds, and they haven’t always chosen them.
This proliferation isn’t just a strategic choice: it’s often an unavoidable reality. Mergers and acquisitions accumulate heterogeneous systems, business units adopt SaaS solutions hosted on different clouds, and technical departments favor the best service rather than consolidation. The result: a fragmented landscape that leaders must take back under control rather than endure.
Three models now structure the trade-offs: the hybrid cloud (a mix of private, public, and on-premises environments), the multicloud (simultaneous use of multiple public providers), and the sovereign cloud (hosting that guarantees European legal control). They are not mutually exclusive: most real architectures combine them.
For a leader, the risk isn’t so much about “missing the cloud shift” — it is largely taken on — but ending up with an incoherent and costly architecture, assembled by accumulation rather than by decision. Reasserting control requires understanding what these models entail and the pressures that justify choosing between them.
Three Pressures That Forcing the Arbitration
The first pressure is costs. The ease of consuming public cloud can backfire: bills that drift away, hard to forecast. Many organizations discover that hosting everything with a hyperscaler costs more than expected, which renews interest in hybrid architectures where each workload is placed where it is most economical.
This finding isn’t incidental: cloud cost is, for the third year in a row according to Flexera, the top concern of IT leaders, cited by 84% of them. The initial reflex of “all to the public cloud” gives way to a more measured approach, where some predictable workloads are sometimes repatriated to private or sovereign cloud when economics justify it.
The second pressure is performance and latency. Not all applications tolerate the distance to a central data center. Local processing needs — industry, retail, real-time applications — push toward hybrid architectures that incorporate edge computing, where data is processed as close as possible to its source.
The third pressure, now decisive in France and Europe, is sovereignty. With the AI Act, stronger NIS2, and geopolitical tensions, the question of who can access data is central. According to an IDC survey relaunched by VMware, 84% of European organizations are using or planning sovereign cloud solutions within twelve months.
Sovereignty: A Legal Issue Before a Technical One
The heart of the sovereignty topic isn’t the physical location of servers but juridical extraterritoriality. The 2018 CLOUD Act in the United States gives U.S. authorities the power to access data held by any American company, regardless of where the data is stored.
In practice, a French company hosting sensitive data in a “Paris” region of an American hyperscaler remains potentially exposed to a U.S. requisition. This risk is what ANSSI’s SecNumCloud certification aims to mitigate, by imposing immunity from non-European law and ensuring European ownership/control. The European sovereign cloud market, estimated at €12.4 billion in 2026 (Markess by Exaegis), testifies to the scale of this concern.
Nevertheless, sovereignty is not a binary all-or-nothing choice. All data do not require the same level of protection: a public-facing site and a patient file do not fall under the same regime. The arbitration is about classifying data and placing each item in the appropriate environment.
Arbitrate Rather Than Choose a Camp
The right approach isn’t to name a single winner model, but to compose according to needs. A few guiding principles shape the arbitration:
- Classify data and workloads by criticality, sensitivity, and performance requirements — this is the prerequisite for any placement.
- Reserve sovereignty for sensitive data (health, regulated data, secrets) and reserve the public cloud for elasticity and innovation.
- Preserve reversibility: avoid locking yourself into a single provider, to maintain maneuvering room.
- Control costs from the outset, placing each workload where its cost/performance ratio is optimal.
The challenge for a leader is thus not to answer “public or sovereign?” but to build a coherent placement strategy where every data item and every application finds its appropriate environment. The cloud stop being a cost borne passively and becomes a managed lever, balancing agility, compliance, and cost control. It is this balance, more than adherence to a single model, that distinguishes organizations that actively manage their cloud from those that endure it.
This strategy is not fixed. The context evolves — regulations, geopolitical tensions, narrowing price gaps between sovereigns and hyperscalers — and the architecture must be able to follow. That is why reversibility (the ability to switch providers or repatriate workloads) becomes a strategic criterion: it ensures that today’s choices will not trap us in a dead end tomorrow. A savvy leader arbitrates not only on cost and sovereignty, but also on future maneuvering freedom.
This content is published by Mentioned
