The trigger differs: a customer demanding that data stays in Europe, a board that no longer wants to depend on one supplier, a dollar invoice that keeps rising, or uncertainty about legislation across the ocean. Whatever the trigger, the reflex “then we pull every plug” is rarely the best move. European cloud or your own hardware is not an all-or-nothing choice, but a route in which you decide per component what belongs where and in which order you shift.

Why the question is on the table now

Customers in healthcare, government, finance and education write into contracts and tenders where data lives and who can legally access it. Regulators look harder at dependence on suppliers outside the EU. And the bill counts: dollar invoices, egress cost and stacked managed services often make stable workloads more expensive than necessary. Whether a European provider or your own hardware is cheaper has to be calculated, not assumed.

What is often missing is nuance. The hyperscalers have European regions and so-called sovereign variants: data stays physically in Europe, European staff operate the environment and you manage your own keys. Truly sovereign it is not, because as long as the owner is American, American legislation can compel the parent company to cooperate. For a large share of data, residency with your own key management is nevertheless sufficient. So the question is not “American or European”, but which data and systems genuinely require European jurisdiction or your own management.

The advantages, honestly

Shifting deliberately wins something real on a few points. Not in theory, but in conversations with customers, auditors and finance.

  • Data residency and jurisdictionWith a European owner, data falls under European law, without an American parent company that can be compelled to cooperate. For customers with strict requirements that is the difference between a long discussion with lawyers and a short “agreed”.
  • Less dependenceA platform that can run with more than one supplier sits differently at the table at a price increase or contract renewal.
  • Predictable costInvoices in euros, no surprises on data transfer and, for stable load, a lower price per compute hour. Own hardware is the extreme case: high fixed, low variable cost.
  • Close and approachableProviders such as Hetzner, Scaleway, OVHcloud, STACKIT or Leaseweb work with European contracts, support in your own time zone and a supplier you can call.
Sovereignty is not a move. It is knowing what runs where, why, and how quickly you could move it.

The disadvantages, just as honestly

Underestimate these costs and you trade a legal risk for an operational one.

  • Fewer managed servicesFor managed databases, queues, identity, secrets, observability and AI services there is not always an equivalent. What you cannot consume as a service you run yourself, with Kubernetes, operators and people who manage it.
  • Maturity and ecosystemThe breadth of services, integrations and documentation is larger at the hyperscalers, though it varies per provider: Scaleway and OVHcloud offer serious identity management with policies and conditions, while smaller parties often stop at project-level roles. You notice during an audit, an incident or when a new team member joins.
  • Own hardware is not cloudCapacity planning, hardware replacement, physical security, networking and someone reachable at night. Only cheaper with stable load and organised operations. Otherwise you are buying a problem.
  • Migration and double costTwo environments for a while, recovery retested and less speed in product development. Those costs belong in the business case, not in a footnote.

Not every plug: choose per component

Most organisations do not need to move their platform. They need to know which components belong where: which data is sensitive or contractually bound, which systems are easy to switch and which are deeply woven into the rest?

AI is a concrete example. Many teams started with an American model provider because at the time it was the only serious option. For summarising, classifying, searching documents and most assistant tasks, European models from Mistral for example are now good enough. If your application calls the model through one interface, you switch with a configuration change and a short quality check. The same goes for object storage through the S3 API, backups, email and monitoring. Every shift is a deliberate, reversible step; that is how you gradually bring more under European management without product development coming to a halt.

  • Must be in Europe or on own hardware: personal and customer data with contractual or legal requirements
  • Can stay: generic, non-sensitive workloads and services without a mature alternative
  • Shifts gradually: AI provider, storage, backups, mail, monitoring and tooling connected through standards

Portability is the real investment

Whether you move or not, the investment that always pays off is portability. Containers and Kubernetes make workloads movable, Infrastructure as Code makes an environment reproducible at any provider, GitOps keeps the delivery path the same and open standards such as Postgres, the S3 API and OpenTelemetry keep the heart of your platform free from one supplier. The goal is not to move. The goal is to be able to move: a platform you can move never has to be moved under pressure.

Do not start with the hardest workload, but with the plug that is easiest to switch.

A realistic route in four steps

Start small, build portability where you improve anyway, and move heavy components only with a business case.

  1. 1

    Inventory and classify

    Per system: which data, which requirements from customers and legislation, and how large the dependence on the current provider is. Usually most of it can simply stay.

  2. 2

    Make portable what you improve anyway

    Infrastructure as Code in order, containers, GitOps and standard integrations, as part of platform work that is on the agenda regardless.

  3. 3

    Start with what is easy to switch

    AI provider, object storage, backups, mail and monitoring: loosely coupled and quick to change. Every step immediately proves whether your portability really works.

  4. 4

    Move heavy components only with a business case

    Own hardware for stable load and organised operations, European cloud for sensitive systems. For most companies hybrid is a perfectly good destination, not an intermediate phase.

When you are better off leaving it

A small team without operating capacity, without customers demanding data residency and with a platform that leans on managed services has little to gain from a move. A European region at the current provider, good encryption with your own key management and a portable platform deliver more. Sovereignty on paper, but a platform you can barely operate, is a bad trade: manageability first, location second.

What it delivers

Choosing per component keeps you in control without bringing the organisation to a standstill. Sensitive data lands in Europe or on your own hardware, generic workloads stay where they run well and cheaply, and every shift is a deliberate, reversible step. In the end the gain lies less in the location than in the portability.