Beyond the Hyperscalers: An Overview of EU Cloud Providers and the Sovereign Imperative

For the better part of a decade, the cloud story in Europe was written in three names from the other side of the Atlantic. Most of us defaulted to them because the convenience was real, the catalogue of services was deep, and the path of least resistance was easy to justify. That period is ending.

The growth of the data economy and a more volatile geopolitical landscape have made the cost of that default visible. What used to read as pragmatism now reads, increasingly, as concentration risk. Boards, auditors, and customers are asking where data lives and who can be compelled to hand it over - and they are not satisfied with answers that point at a data centre on a map.

This article looks at the European cloud ecosystem through five of the providers building it: OVHcloud, Scaleway, Hetzner, STACKIT, and UpCloud. The aim is practical: what each one is good for, where they fit in a sovereign strategy, and how to think about the trade-offs.

Why this conversation has changed

Sovereignty in the cloud is not the same as residency. A US hyperscaler can run servers in Frankfurt and still be subject to the laws of its headquarters. The US CLOUD Act lets American authorities compel US-headquartered companies to hand over data regardless of where it sits, and US courts have repeatedly held that the parent company's legal obligations extend to subsidiaries. For organisations handling sensitive citizen records, regulated industrial data, or healthcare information, that residual exposure is hard to defend in a procurement review.

The second issue is vendor lock-in. Hyperscaler ecosystems are built on dense, proprietary managed services that make it easy to start and expensive to leave. A sovereign strategy needs reversibility: a contractual and architectural ability to move workloads without paying punitive egress fees or rewriting half the stack. That is partly a procurement question and partly a technology choice - portable interfaces (Kubernetes, OpenStack, S3-compatible storage) keep options open in ways that proprietary services do not.

The third is structural. Spend on cloud is an investment decision: every euro routed through a non-EU vendor funds engineering capacity outside the EU. A working European cloud ecosystem is also a precondition for European AI - the compute that trains a foundation model shapes the values baked into it. If the infrastructure layer is not European, the layer above will not be either.

The five providers

OVHcloud is the largest European cloud provider, headquartered in Roubaix. Its operational distinctive is vertical integration: it designs its own servers, builds its own data centres, and runs its own water-cooling technology across factories in France and Canada. For organisations that need the strongest sovereignty posture available today, OVHcloud's Hosted Private Cloud is SecNumCloud-qualified by ANSSI, the French national cybersecurity agency - the qualification was extended to Bare Metal Pod in early 2025. On the enterprise availability side, OVHcloud now runs continuous 3-Availability-Zone regions in Paris, Italy, and Berlin. It is a founding member of Gaia-X and offers a managed Kubernetes service on its Public Cloud. If you need to compete with hyperscalers on scale and on certified sovereignty, this is where the conversation starts.

Scaleway is a Paris-headquartered provider owned by the Iliad Group, with a developer-first product surface and an explicit positioning as the sovereign alternative to the US giants. Its cloud regions today span Paris, Amsterdam, Warsaw, and Milan, with new regions announced for Sweden and Germany. Its current centre of gravity is AI infrastructure: in 2024 it launched Nabuchodonosor, a supercomputer built on 127 NVIDIA DGX H100 systems, alongside L40S GPU instances for inference workloads. The point is to give European startups and enterprises somewhere to train models without sending their data through US infrastructure. Sustainability is unusually transparent for the sector - Scaleway runs on 100% renewable energy and publishes its PUE and water usage metrics. On pricing, outbound traffic is included with each instance rather than billed per gigabyte, which removes the egress trap that makes leaving hyperscalers expensive. Managed Kubernetes, serverless, and a broadly open-source-compatible stack make Scaleway a natural home for cloud-native teams.

Hetzner is the price-performance choice. Based in Gunzenhausen, it runs large, efficient data centres in Germany and Finland and has built a loyal developer following by doing fewer things, well. The catalogue is deliberately narrow - compute, storage, networking, dedicated servers - and the price-to-performance ratio is consistently better than the hyperscalers'. Hetzner does not try to be a walled garden of managed services. That means more architectural work for the customer, but it also means a stack you understand and can move. For European startups that need to bootstrap without spending the first year of runway on cloud bills, or for DevOps-mature teams that already run their own platform layer, Hetzner is hard to beat.

STACKIT is the cloud arm of the Schwarz Group, the German retail group behind Lidl and Kaufland. The origin story matters: the Schwarz Group needed enterprise-scale cloud, refused to put its retail data on US infrastructure, built its own, and then offered it to the wider market. STACKIT is built on OpenStack, which removes one of the biggest sources of lock-in by design. It runs colocation sites in Neckarsulm and Ellhofen (Germany) and Ostermiething (Austria), exposed as the eu01 and eu02 cloud regions, with a 200 MW campus under construction in Lübbenau - the entire footprint is DACH-aligned for compliance. Where STACKIT stands out is Confidential Computing: through its partnership with Edgeless Systems, it offers Confidential Kubernetes, which keeps data encrypted not only at rest and in transit but in use at the CPU level. For finance, healthcare, and public-sector workloads, that is a genuinely differentiated security posture.

UpCloud is the performance specialist of the group. Headquartered in Helsinki, it built its reputation on MaxIOPS, a proprietary block storage technology that consistently benchmarks ahead of equivalent SSD storage at the larger providers - which makes it a natural fit for I/O-bound workloads, large transactional databases, and high-traffic e-commerce. Its European footprint covers Helsinki, Amsterdam, Frankfurt, Madrid, London, Warsaw, Stockholm, Copenhagen, and - as of January 2026 - Stavanger, with additional regions in the US, Singapore, and Australia for global customers. As of May 2025, UpCloud advertises a 99.999% SLA across its core services. The developer experience is clean - well-maintained APIs, a Terraform provider, and predictable pricing that protects against the unexpected bill at the end of the month. UpCloud's legal footing and commitments remain European regardless of where the workload runs.

How to use this

The transition to European cloud is not a one-step migration, and any honest framing has to admit that. The starting point is a clear-eyed read of which data carries which risk. Public-facing marketing assets and ephemeral build artefacts do not carry the same exposure as customer databases, financial records, or AI training data. Reserve the strongest sovereignty posture for the workloads that justify it; spend the engineering budget where it changes the risk profile.

The second move is to treat multi-cloud as a design principle rather than an accident. There are workloads where a hyperscaler service is the right answer for now, and there are workloads where the answer is clearly a European provider. A deliberate architecture lets each one sit where it belongs and lets you change your mind later without rewriting the application. STACKIT and OVHcloud are natural anchors for regulated data; Scaleway and UpCloud cover most cloud-native and performance-sensitive use cases; Hetzner is the price-performance baseline.

The third is contractual. Reversibility belongs in the contract, not in the architecture brief. An exit strategy that has been costed, written down, and tested is what separates a sovereign strategy from a marketing claim. The technical side of reversibility is more boring than it sounds - portable orchestration (Kubernetes), portable infrastructure layers (OpenStack), and open data formats - but it is the part that lets you move workloads when the procurement decision changes.

And finally, procurement spend itself is a lever. Cloud budgets are large, and the cumulative effect of European organisations routing those budgets through European providers is what funds the next generation of EU infrastructure. The infrastructure is ready, the legal case for using it is clearer than it has been in a decade, and the providers are no longer making customers choose between sovereignty and capability.

Thinking about migration? Book a free consultation to discuss your situation.