European Cloud Providers and Data Sovereignty
Data sovereignty is three questions wearing one label: where bytes rest, whose courts apply, and who can see plaintext. How European providers answer each.
Data sovereignty gets discussed as though it were a single property a provider either has or does not. In practice it is three separate questions that happen to share a word, and a provider can satisfy one of them completely while failing another. Procurement conversations go badly when the two sides are answering different questions.
If you are being asked to keep data in Europe, the first useful step is to find out which of the three questions is actually being asked, because the answers lead to different providers and different contracts.
Three questions wearing one label
Where does the data physically rest? Which building, in which country, holds the disks. This is the most concrete question and the easiest to answer.
Which legal system can compel disclosure? Whose courts and agencies can issue an order that the operating company must obey, and whether that jurisdiction reaches beyond its own borders to data held abroad by companies under its control.
Who can technically see the data? Which staff, contractors and subcontractors have a route to the plaintext, where those people sit, and whose law applies to them.
A workload can be stored exclusively in a German data centre, operated by a company whose ultimate parent sits outside the European Union, and supported by engineers on another continent. All three statements can be true at once. Deciding whether that is acceptable is a legal and risk judgement, not a technical one, and it is the judgement your regulator or your customer’s security team is really making.
Residency is the part you can verify
Pinning down physical location is straightforward, and you should do it concretely rather than accepting a marketing region name. Region labels describe a geography loosely; what you want is the country and ideally the metropolitan area of the facility.
Then check the services around the data, because residency leaks at the edges. Backups may replicate to a different region by default. Logs, metrics and traces often flow to a central observability endpoint that is not in your chosen location. Content delivery caches copies worldwide by design. Support tooling may hold diagnostic extracts elsewhere. Email notifications containing record identifiers pass through whatever mail infrastructure the provider uses.
Go through each of these explicitly for your deployment. The fundamentals of how region and zone boundaries work are in how cloud regions and availability zones work, and the distinction between a region and a replication target is where most residency surprises come from.
Corporate jurisdiction is where it gets contested
The harder question is which legal regime the operating company falls under. Some jurisdictions assert authority over data held by companies subject to their law regardless of where the servers sit. Where that is the case, choosing a local data centre operated by a subsidiary of a foreign parent may not change the legal exposure, even though it changes the physical one.
This is precisely why a layer of European-headquartered providers has become a procurement answer rather than just a cost answer. A company incorporated and controlled in the European Union, with no foreign parent, has a cleaner answer to the jurisdiction question, and that answer is what some buyers are paying for.
Be careful not to overstate it. Jurisdiction is a legal analysis that depends on your sector, your data and your counsel’s reading, and it changes as case law and treaties move. Treat provider ownership as one input to that analysis, documented in your records, not as a conclusion.
Follow the support path and the sub-processors
The third question is the one most often skipped, and it is the one that fails an audit.
Ask where support staff are located and whether a support engineer can access your data or only your account metadata. Ask what happens during an escalation: does a diagnostic bundle get uploaded somewhere, and where is that somewhere? Ask for the sub-processor list, which names the third parties the provider itself relies on, and read it. A provider operating entirely within Europe may still use a non-European service for ticketing, monitoring or payment processing.
Then ask whether you are notified before a sub-processor changes. A list is a snapshot; the notification clause is what keeps it accurate. Record the answers alongside your other compliance evidence, as discussed in cloud compliance, GDPR, HIPAA and SOC 2.
How the European providers differ on this axis
Several established European operators come up when residency and jurisdiction drive the decision. They are not interchangeable, and the differences that matter here are structural rather than feature-level.
Hetzner is a German company operating its own data centres in Germany and Finland, with additional presence in the United States. For a buyer whose requirement is German or European residency under European corporate control, the chain of ownership is short and easy to document. It is also known for low prices on dedicated and virtual servers, which means sovereignty here does not come with a premium. Listings are available as a Buy Hetzner Account.
OVHcloud is French, operates its own facilities across several countries, and builds much of its own hardware. Of the group it is the one most often found in enterprise and public sector procurement, because it offers a broader service catalogue alongside plain compute and has long engaged with European regulatory expectations. Its scale also means more region choice within Europe, which matters when a requirement names a specific country rather than the bloc.
Scaleway is French and tends to appeal to teams who want managed services, containers and object storage with a European legal footing rather than only virtual machines. If your residency requirement sits alongside a desire not to hand-roll every platform component, it is usually on the shortlist. See the Buy Scaleway Account listing for current options.
UpCloud is Finnish and positions around storage and network performance. For workloads where a database must stay in Europe and must also be fast, it is a credible answer to both constraints at once rather than a compromise on one.
Contabo is German and sits at the value end, offering generous resource allocations for the price. It fits bulk storage, archives and secondary environments that must remain in Europe but do not justify premium infrastructure. Where it is less suited is workloads with strict performance consistency requirements, which is a normal trade at that price point.
If you are weighing two of these directly, our Scaleway and OVHcloud comparison covers the service breadth difference, and the Buy OVH Account listing shows what is currently available.
Encryption shifts the question rather than answering it
Holding your own keys, outside the provider, changes the shape of the risk meaningfully. If the provider never has the key material, a disclosure order served on the provider produces ciphertext. Some regulators and many customer security teams treat that as a substantive control, and it is one of the few technical measures that genuinely reduces exposure to foreign legal process.
It does not eliminate the question. Keys must live somewhere, and that somewhere has its own jurisdiction. Data in use is decrypted in memory on the provider’s hardware. Metadata, who talked to what and when, usually remains visible even when payloads are not. And operationally, holding your own keys means you own the consequences of losing them.
Done well it is a strong position. Done casually it creates a false sense of safety plus a new single point of failure. Build the key handling first, following cloud secrets management, and only then count it as a control.
Put the answers in a document, not a conversation
Whatever you conclude, write it down in a form an auditor or a customer can read. A short record per workload works: the data categories involved, the countries where the data rests, the countries where backups and logs rest, the operating company and its parent, the sub-processor list and its date, who can access plaintext, where keys are held, and the contractual commitments you have on each of those points.
Keep it current, and keep it with your access logging evidence, as set out in cloud audit logging from day one. Most sovereignty failures discovered in audits are not architectural. They are a backup target nobody revisited after a change eighteen months earlier.
When a European provider is not the answer
If your constraint is a managed service with no European equivalent, forcing the whole estate onto a regional provider to satisfy a requirement that applies to one dataset is a poor trade. The more usual resolution is to keep the regulated data with a provider that satisfies the constraint and run the rest where it suits you, accepting a split estate for a specific, documented reason.
On TrustCloudHub you can review verified listings for European providers including a Buy UpCloud Account and a Buy Contabo Account. We are an independent marketplace with no affiliation to any provider, and nothing here is legal advice: confirm region, backup and support arrangements inside the provider’s own console and contract before you rely on them. Where an account carries promotional credits, remember that credits are provider-controlled, expire on the provider’s schedule, exclude some services, and should be verified in the console after transfer.