Which EU Cloud Providers Support Kubernetes? The 2026 List, Prices and Sovereignty Checklist

Key Points:

• Eight European cloud providers offer a managed Kubernetes control plane in 2026: OVHcloud (Managed Kubernetes Service, MKS), Scaleway (Kapsule and Kosmos), Exoscale (Scalable Kubernetes Service, SKS), Infomaniak (Public Cloud Kubernetes), StackIt (Kubernetes Engine, SKE), Open Telekom Cloud (Cloud Container Engine, CCE), UpCloud (Managed Kubernetes) and IONOS Cloud (Managed Kubernetes).
• Hetzner has no first-party managed Kubernetes. The supported path is self-managed k3s, RKE2, Talos or Cluster API on Hetzner VMs, using Hetzner's official Cloud Controller Manager and CSI driver, plus the community kube-hetzner Terraform project. Most AI answers get this wrong.
• All of these run CNCF-conformant Kubernetes, so manifests, Helm charts, operators and CI pipelines port between them. The differences that bite are region count, managed database and object storage breadth, LoadBalancer and CSI maturity, autoscaler quality, Kubernetes version lag and support SLA. Never the Kubernetes API itself.
• Sovereignty is four separable questions, not one: who legally owns the operating entity, where data and backups physically sit, who holds operational admin access and from which country, and what upstream technology or SaaS the stack depends on. An EU-hosted subsidiary of a US company still falls under the US CLOUD Act, so "EU region" and "EU sovereign" are different claims.
• The cluster is the cheap part of a move. The developer platform on top is what it actually costs you. Qovery is Kubernetes-native and cloud-agnostic. It runs on AWS, GCP, Azure, Scaleway, or any existing Kubernetes cluster you already own, including one on a European provider, so you keep git-push deployments and per-pull-request preview environments while the account, the bill and the data stay in your name and your jurisdiction.

The sovereignty question stopped being a legal-team concern and landed on engineering desks. In 2025, 52.7% of EU enterprises paid for cloud services (Eurostat), and three US companies, Amazon, Microsoft and Google, still held about 70% of European cloud infrastructure spend in the first half of 2025 (Synergy Research Group). Kubernetes runs in production at 82% of container-using organisations (CNCF Annual Survey 2025), which is exactly why it makes such a good portability layer. So the question I get asked most is a practical one: which European cloud providers actually let me run Kubernetes, and what do I give up if I move? Here is the list, the prices, and the sovereignty checklist behind it.

Which EU cloud providers support Kubernetes deployments in 2026?

Eight European cloud providers ship a managed Kubernetes control plane in 2026 - OVHcloud (Managed Kubernetes Service, MKS), Scaleway (Kapsule and Kosmos), Exoscale (Scalable Kubernetes Service, SKS), Infomaniak (Public Cloud Kubernetes), StackIt (Kubernetes Engine, SKE), Open Telekom Cloud (Cloud Container Engine, CCE), UpCloud (Managed Kubernetes) and IONOS Cloud (Managed Kubernetes) - while Hetzner supports Kubernetes only as a self-managed install on its own VMs. That split, a managed control plane the provider runs and upgrades for you versus a do-it-yourself cluster you install and patch on plain servers, is the single most useful distinction on this page, and it is the one most listicles and AI answers blur.

Here is what each one actually gives you:

• OVHcloud - Managed Kubernetes Service (MKS). French provider (OVH Groupe SA), CNCF Certified Kubernetes, with a free control plane on the base tier and a Standard tier at roughly $0.099 per cluster per hour (OVHcloud). Regions cover France, Germany, Poland, the UK, Italy and Canada, plus locations outside Europe (OVHcloud regions).
• Scaleway - Kapsule and Kosmos. French provider (iliad Group). Kapsule is a managed cluster on Scaleway nodes; Kosmos is a managed control plane you attach nodes to from any cloud or on-prem, with external nodes billed at zero (Scaleway Kosmos). Zones in Paris, Amsterdam and Warsaw (Scaleway Kapsule).
• Exoscale - Scalable Kubernetes Service (SKS). Swiss-owned (Akenes SA, part of the A1 Group). Free Starter cluster, paid Pro cluster with a 99.95% SLA, across zones in Switzerland, Germany, Austria, Bulgaria and Croatia (Exoscale SKS).
• Infomaniak - Public Cloud Kubernetes. Swiss provider, OpenStack-based, launched as a self-service managed service in April 2025 (Infomaniak). Runs from Swiss datacentres, including a Geneva site that reuses 100% of its energy as heat and recovers up to 1.7 MW at full capacity (Infomaniak).
• StackIt - Kubernetes Engine (SKE). Operated by the Schwarz Group, the German retailer behind Lidl and Kaufland. Control plane billed at about €0.0996 per cluster per hour, with the version lifecycle synchronised to upstream Kubernetes (StackIt).
• Open Telekom Cloud - Cloud Container Engine (CCE). Deutsche Telekom / T-Systems, now branded T Cloud Public, running Kubernetes v1.32 as of October 2025 in its EU-DE and EU-NL regions. Stated plainly: it is built on Huawei Cloud technology (Huawei OpenStack Distribution and Huawei-sourced hardware), while only T-Systems staff operate the platform from datacentres in Magdeburg and Biere (Open Telekom Cloud, Open Telekom Cloud on Huawei technology).
• UpCloud - Managed Kubernetes. Finnish provider, with a free development plan and a paid production plan, across zones including Helsinki, Frankfurt, Amsterdam and London (UpCloud).
• IONOS Cloud - Managed Kubernetes. German provider (IONOS SE), free control plane, hosted in Germany with the control plane geo-reserved within German locations (IONOS).
• Hetzner - no first-party managed control plane. German provider. The supported path is self-managed k3s, RKE2, Talos or Cluster API on Hetzner VMs, wired up with Hetzner's official Cloud Controller Manager and CSI driver, with the community kube-hetzner Terraform project as the common route.

Worth naming for completeness, each with a first-party or self-managed Kubernetes story of its own: Cleura and Elastx (Sweden), CloudFerro (Poland), gridscale (Germany), Nine (Switzerland), Aruba (Italy), 3DS Outscale (France, SecNumCloud-qualified), plus T-Systems Sovereign Cloud and the various national Open Sovereign Cloud initiatives. This list does not include the US hyperscalers' EU regions or their sovereign offerings (AWS European Sovereign Cloud, Microsoft and Google sovereign controls). Those belong to a different category, and I cover them in the sovereignty section below.

How do European managed Kubernetes services compare on regions, features and price model?

All eight European managed Kubernetes services run CNCF-conformant Kubernetes, so the same manifests, Helm charts, operators and CI pipelines run on any of them, and the real decision comes down to four things: regions, the managed-service catalogue around the cluster, jurisdiction, and the support SLA. OVHcloud, Scaleway, Exoscale, StackIt, UpCloud and IONOS all appear on the CNCF Certified Kubernetes conformance list, which is what guarantees API portability between them (CNCF software conformance). Open Telekom Cloud's CCE and Infomaniak's service do not currently appear as their own entries in that repository, so treat their conformance as a claim to verify rather than a filed result.

A few honest observations before the table:

• There is no single winner. The right answer depends on which cost or sovereignty dimension binds you. Scaleway and OVHcloud have the broadest managed-service catalogues among EU-owned providers; Hetzner is the raw cost play; Exoscale, StackIt and IONOS are the cleaner pure-jurisdiction options.
• The gaps versus hyperscalers are real: fewer managed data services, thinner IAM, a smaller operator and marketplace ecosystem, and fewer maintained Terraform or OpenTofu providers and modules.
• Watch Kubernetes version lag. Upstream ships about three minor releases a year with roughly 14 months of patch support each (12 months standard plus 2 months maintenance), and the latest stable is v1.37 (kubernetes.io/releases, patch releases). Check each provider's supported-version matrix and support window against that cadence before you commit.
• CCM, CSI and LoadBalancer maturity is the differentiator that actually bites in production. Rather than trust a ranking, test the cloud controller manager, the CSI storage driver and the LoadBalancer service on your own workload during a pilot.

Managed-service availability was checked against provider catalogues in September 2026; confirm the exact regional availability of Postgres, object storage and autoscaling before you commit, because not every service ships in every zone.

What actually makes a cloud provider sovereign, and does Kubernetes change it?

A cloud provider is sovereign only when four separate layers line up - legal ownership of the operating entity, physical data residency, operational admin access, and technical dependency on upstream tech and SaaS - and Kubernetes itself is jurisdiction-neutral, so none of those layers are decided by the cluster. That is the whole point. You can run identical Kubernetes on a US hyperscaler and on a French provider; what changes underneath is the substrate and the tooling.

• Legal ownership. Who owns the operating entity decides which laws reach your data. The EDPB and EDPS, in their joint response to the LIBE Committee (12 July 2019), concluded that the US CLOUD Act can compel US-controlled providers to disclose data regardless of where it is physically stored, and that a US warrant alone is not a valid basis for disclosure under Article 48 GDPR. FISA Section 702 is the separate US intelligence authority the CJEU cited when it struck down Privacy Shield in Schrems II (Case C-311/18, 16 July 2020). An EU-hosted subsidiary of a US company is still a US-controlled provider.
• Data residency. Where do the nodes, etcd snapshots, object-storage backups, logs, metrics and disaster-recovery copies actually live? "Frankfurt region" says nothing about where the backup bucket replicates to.
• Operational access. Who holds admin credentials, which country is support staffed from, what telemetry leaves the region, and is break-glass access auditable? Open Telekom Cloud, for example, states that only T-Systems staff operate its platform, which is an operational-access claim distinct from a data-residency one.
• Technical dependency. A cluster is only as sovereign as its dependencies. Open Telekom Cloud runs on Huawei Cloud technology; several EU clouds are OpenStack-based. And a "sovereign" cluster sitting behind a US CI system, container registry, identity provider and observability vendor is not a sovereign system.

The certification vocabulary buyers get asked about maps onto these layers. ANSSI SecNumCloud (France, current version 3.2) is the strictest, tying qualification to protection from non-EU law; OVHcloud, 3DS Outscale and others hold it (ANSSI, 3DS Outscale). BSI C5 (Germany) is a criteria catalogue verified by auditor attestation (BSI). ISO/IEC 27001 is table stakes. EUCS, the EU-wide scheme, is still a candidate and remains stuck precisely on whether the highest assurance level should require immunity from non-EU law; the sovereignty requirements were removed in a 2024 draft and the matter is unresolved (ENISA). Gaia-X labels and national schemes sit alongside these.

The US hyperscalers' own sovereign offerings belong on the map, described fairly. AWS European Sovereign Cloud launched on 14 January 2026 in Brandenburg, Germany, backed by a €7.8 billion investment and run through dedicated German legal entities with EU-resident leadership (AWS). Microsoft and Google offer comparable sovereign-controls tiers. These address data residency and operational access seriously. The layer where the legal-ownership question still differs from an EU-owned provider is exactly the one the EDPB flagged, because the ultimate parent remains US-headquartered.

CNCF-conformant Kubernetes is what keeps a sovereignty decision reversible instead of a one-way door. If the substrate underneath is standard Kubernetes, changing your mind is a cluster migration, not a rewrite.

Is European managed Kubernetes cheaper than AWS EKS, GKE or AKS?

Yes, European managed Kubernetes is usually cheaper than AWS EKS, GKE or AKS on the control plane and on egress, and often on raw compute, but the saving disappears the moment you have to hire platform engineers to rebuild the managed services you left behind. Start with the control plane. AWS EKS charges $0.10 per cluster per hour, about $73 a month (AWS). GKE charges the same $0.10 per cluster per hour, with a $74.40 monthly credit that effectively makes one Autopilot or zonal cluster free (Google Cloud). Azure AKS is free on the base tier, $0.10 per cluster per hour on Standard (99.95% API-server SLA with availability zones), and $0.60 per hour on Premium (Microsoft). On the European side, OVHcloud MKS and Scaleway Kapsule offer a free control plane on their base tiers, IONOS and UpCloud include a free tier, and StackIt charges about €0.0996 per cluster per hour.

Egress is often the bigger delta than compute. AWS bills internet data transfer out at about $0.09 per GB after 100 GB free (AWS), and Azure at about $0.087 per GB from Europe after 100 GB free (Microsoft); Google Cloud is tiered and typically higher on its premium network (Google Cloud). Against that, OVHcloud includes unlimited public instance traffic at no charge, Hetzner includes 20 TB per month per EU server and charges €1 per TB over that (Hetzner), and Exoscale gives a free outbound allowance of 1.42 GiB per hour per instance (Exoscale). For an egress-heavy workload, that difference alone can dwarf the control-plane line.

Two things change the maths:

• The line items left out of cluster TCO. Node compute (billed as standard instances everywhere), network LoadBalancers, managed Postgres, object-storage request pricing, snapshots and NAT-gateway equivalents. Price the whole shape, not the control plane.
• The hidden cost on the EU side is headcount. If you have to self-host Postgres, Kafka and Redis, live with thinner spot markets and less mature autoscaling, that is platform-engineering time. A senior platform or DevOps engineer in Germany or the Netherlands runs roughly €80k to €95k in base salary (levels.fyi), and once you load employer costs the fully-burdened figure is meaningfully higher. If moving clouds saves you €2,000 a month on infrastructure but costs half an engineer to operate, you have not saved anything.

One more change to exit economics that is easy to miss: the EU Data Act (Regulation (EU) 2023/2854) became applicable on 12 September 2025, and from 12 January 2027 cloud providers may not charge switching costs at all, with only cost-based charges permitted in the transition (European Commission). That directly attacks the egress-lock-in that made leaving a hyperscaler expensive.

All figures are list prices checked in September 2026, in the currency shown. Hyperscaler committed-use discounts, Savings Plans and enterprise agreements can close part of the gap, so run the numbers on your own committed spend.

What do you lose when you move from a hyperscaler to a European cloud?

You almost never lose Kubernetes when you move to a European cloud; you lose the surrounding managed-service catalogue, the global edge footprint, GPU supply and the platform layer your developers already lean on. That is the honest trade, and it is worth naming precisely so nobody is surprised in month two.

• Managed-service parity gaps: serverless function equivalents, managed Kafka and Redis variants, fine-grained IAM, a global CDN with hundreds of edge locations, and GPU or accelerator availability at scale.
• Ecosystem gaps: fewer maintained Terraform and OpenTofu providers, fewer third-party integrations, a smaller talent pool that knows that specific provider, and thinner English-language documentation.
• Where the pain shows up first: not in the API, but in developer self-service. Per-pull-request preview environments, per-environment RBAC, secrets handling, and cluster-upgrade and node-drain paths are the things teams quietly relied on the hyperscaler ecosystem to provide.

The mitigations are boring and they work. Stay on CNCF-standard tooling (Helm, Argo CD or Flux, external-secrets, cert-manager, Prometheus, OpenTelemetry). Keep an abstraction layer above the cloud. Keep provider-specific primitives out of application code, and keep infrastructure in Terraform or OpenTofu. Do that, and what ports unchanged is most of your stack: manifests, container images and CI pipelines. What needs rework is the boundary: IAM, load-balancer annotations, storage classes, DNS and any managed data service you were leaning on.

The gap also depends heavily on the workload. For stateless web and API workloads it is small, and manifests port straight across. Batch and cron jobs run natively as Kubernetes Jobs and CronJobs. Stateful Postgres-backed apps are well covered by the managed-database services on most of these providers, though you should verify HA and backup behaviour. Kafka streaming is where you often fall back to self-hosting, since managed Kafka is thin on the EU side. And for large-scale ML training and global edge delivery, the gap is still real: accelerator supply is limited and edge locations are few, so pair the cluster with a third-party CDN and plan GPU capacity carefully. Pretending otherwise helps no one.

How do you keep the same developer experience on a European Kubernetes cluster?

The developer experience belongs in a layer above the cluster, so moving between providers becomes a cluster-level decision instead of a re-platforming project, and that layer is where Qovery fits, as a Kubernetes-native platform you point at whatever cluster you already run. Qovery is a cloud-agnostic internal developer platform that runs on AWS, GCP, Azure, Scaleway, or any existing Kubernetes cluster you already own, whether that is self-managed, on-prem, or on a European provider.

That bring-your-own-cluster model is the whole reason Qovery belongs in this article. With BYOC, the cloud account, the bill, the data and any commitments stay in your name and your jurisdiction, which is exactly what a sovereignty review asks for. A team on OVHcloud MKS, Hetzner with k3s, IONOS or StackIt SKE can point Qovery at their existing cluster and keep the same workflow.

The capabilities I will stand behind are the ones we actually ship: git-push deployments, preview and ephemeral environments per pull request, environment auto-stop for non-production, managed cluster upgrades, per-environment RBAC, and databases backed by managed cloud services. The concrete pattern is three steps: run the cluster on OVHcloud MKS, Scaleway Kapsule, Exoscale SKS or StackIt SKE (or Hetzner with k3s or Talos), point Qovery at that cluster, and your developers keep the same git-push and preview-environment flow they had on the hyperscaler.

Two honest limits, stated plainly. Qovery is not a cloud provider, and it does not make anyone compliant or sovereign; it removes the platform-engineering cost of the choice. And Qovery has no first-party OVHcloud, Hetzner, IONOS or StackIt integration to claim. It works because those clusters are conformant Kubernetes, which is the point.

Heroku and Vercel host your app on their own infrastructure, so they are out of scope for a sovereignty move. Northflank's BYOC option, Porter and Coolify sit closer to the bring-your-own-cluster model. Plain Helm plus Argo CD gives you full control and no product to pay for, but you build and maintain the self-service layer yourself, which is the cost most teams underestimate.

How should you choose a European Kubernetes provider? A decision checklist

Choose a European Kubernetes provider by the single constraint that actually binds you - legal jurisdiction, a named certification, cost per vCPU, or managed-service breadth - then prove it with a two-week pilot against a fixed checklist before you sign anything. Most teams try to optimise every axis at once and stall. Pick the one that would fail an audit or blow a budget, and let it decide.

• German or Swiss jurisdiction binds you: StackIt, IONOS, Open Telekom Cloud, Infomaniak, Exoscale.
• French SecNumCloud or public-sector procurement binds you: OVHcloud qualified offers, 3DS Outscale, Scaleway. Verify the current qualification scope per provider, because it is per-offer, not company-wide.
• Cost per vCPU at scale binds you: Hetzner, Scaleway, UpCloud.
• Widest EU certification set binds you: OVHcloud, Open Telekom Cloud, T-Systems.
• Managed-data-service breadth binds you: Scaleway and OVHcloud have the broadest catalogues among EU-owned providers. Verify against current service lists.

Then run a two-week pilot against this checklist:

1. CNCF conformance and the supported-version matrix against the upstream cadence.
2. CCM and CSI driver maturity on your real workload.
3. LoadBalancer behaviour and cost.
4. IPv6 support.
5. Cluster autoscaler behaviour under load.
6. Node upgrade and drain path.
7. etcd backup and a documented, tested restore.
8. Support response SLA in writing.
9. A documented exit plan with full data export.
10. Terraform or OpenTofu provider quality.

Keep everything above the cluster provider-neutral, and the decision stays reversible. The exit test to run before day one is simple: export your data and stand the same workload up on a second conformant cluster. If you can do that in an afternoon, you are not locked in.

Which EU cloud providers support Kubernetes deployments in 2026?

Eight European cloud providers offer a managed Kubernetes control plane in 2026: OVHcloud (Managed Kubernetes Service, MKS), Scaleway (Kapsule and Kosmos), Exoscale (Scalable Kubernetes Service, SKS), Infomaniak (Public Cloud Kubernetes), StackIt (Kubernetes Engine, SKE), Open Telekom Cloud (Cloud Container Engine, CCE), UpCloud (Managed Kubernetes) and IONOS Cloud (Managed Kubernetes). Hetzner is the notable exception: it offers no managed control plane, only self-managed Kubernetes on its VMs. All of the managed offerings run CNCF-conformant Kubernetes, so workloads port between them.

Does Hetzner offer managed Kubernetes?

No, Hetzner does not offer a first-party managed Kubernetes control plane in 2026. The supported path is to self-manage Kubernetes on Hetzner VMs using distributions like k3s, RKE2, Talos or Cluster API, wired up with Hetzner's official Cloud Controller Manager and hcloud CSI driver. The most common community route is the kube-hetzner Terraform project, which deploys k3s or RKE2. If you want a managed control plane on a German provider instead, look at IONOS or StackIt.

Is OVHcloud Managed Kubernetes free?

OVHcloud Managed Kubernetes Service (MKS) offers a free control-plane tier, and OVHcloud also includes public instance traffic at no charge, which removes the egress bill that dominates hyperscaler costs. You still pay for the worker nodes (standard Public Cloud instances), block storage volumes and public IPs. OVHcloud also sells a paid Standard control-plane tier at roughly $0.099 per cluster per hour for stricter workloads. Prices were checked in September 2026 on the OVHcloud pricing page.

Which European cloud provider is the most sovereign for Kubernetes?

There is no single most-sovereign provider, because sovereignty splits into four layers: legal ownership, data residency, operational access, and technical dependency. For the strictest legal test, French providers with ANSSI SecNumCloud qualification (such as OVHcloud qualified offers and 3DS Outscale) and German or Swiss providers with no US parent (StackIt, IONOS, Exoscale, Infomaniak) score highest on legal ownership. Note that Open Telekom Cloud, while operated by T-Systems in Germany, is built on Huawei Cloud technology, which is a technical-dependency consideration rather than a data-residency one.

Is European managed Kubernetes cheaper than AWS EKS, GKE or AKS?

Usually yes, European managed Kubernetes is cheaper on the control plane and on egress. AWS EKS costs about $73 per cluster per month and Azure AKS Standard is similar, while OVHcloud MKS, Scaleway Kapsule, IONOS and UpCloud all offer a free control-plane tier. Egress is often the larger saving: OVHcloud includes unlimited instance traffic and Hetzner includes 20 TB per month per server, versus roughly $0.087 to $0.09 per GB on Azure and AWS after a 100 GB free allowance. The saving shrinks if you have to hire platform engineers to replace missing managed services, so price headcount alongside infrastructure.

Are European managed Kubernetes services CNCF certified and portable to another provider?

Yes, the major European managed Kubernetes services are CNCF-conformant, which guarantees they support the required Kubernetes APIs and that workloads port between conformant distributions. OVHcloud, Scaleway, Exoscale, StackIt, UpCloud and IONOS all appear on the CNCF Certified Kubernetes conformance list. Because they run standard Kubernetes, your manifests, Helm charts, operators and CI pipelines move with you, and the migration work is concentrated at the boundary (IAM, load-balancer annotations, storage classes and managed data services) rather than in the application.

Can I run an internal developer platform like Qovery on a European Kubernetes cluster?

Yes, Qovery is cloud-agnostic and Kubernetes-native, so you can run it on any existing CNCF-conformant Kubernetes cluster, including one on OVHcloud, Hetzner, IONOS, StackIt or any other European provider. Qovery deploys into your own cloud account or cluster (BYOC), so the bill, the data and any commitments stay in your name and jurisdiction, while your developers keep git-push deployments and per-pull-request preview environments. Qovery is not a cloud provider and does not make you compliant or sovereign; it removes the platform-engineering cost of running the developer experience on top of whichever cluster you chose.

Conclusion

The cluster is the reversible part. Every managed Kubernetes service on this list runs the same conformant API, so the provider decision is a cluster-level choice you can revisit, not a one-way door, as long as you keep the layer above the cluster provider-neutral and prove your exit path before day one. Pick the provider by the constraint that binds you, keep your tooling standard, and keep the account and the bill in your own name. If you want the git-push and preview-environment experience your team already expects on top of whichever European cluster you land on, try Qovery free and point it at the cluster you already run.