
Self-hosted Kubernetes: when running a cluster outside AWS, Azure, and Google pays off
Predictable costs, data that stays in Central Europe, and no egress fees. Plus an honest look at when your own Kubernetes doesn't make sense.
Written by DENEVERSE team, Managed infrastructure specialists.
Self-hosted Kubernetes pays off when your load is steady, you want your data under control in the EU, and you don't want to pay for every gigabyte that leaves the cloud. It doesn't pay off for a single small app, or when nobody on your team can run a cluster. For those teams, managed Kubernetes from a local provider combines the best of both worlds.
Kubernetes for companies
Kubernetes for companies makes sense once one application grows into several services and an hour of downtime costs more than running the cluster. A company usually needs more than the orchestrator itself. It needs to know who upgrades the cluster, where the data sits, and how fast support answers.
Kubernetes is no longer an experiment
According to the CNCF survey published in January 2026, 82% of organizations that use containers run Kubernetes in production, up from 66% in 2023. Kubernetes has become the standard way to run modern applications. The question is no longer whether to use Kubernetes, but where it runs and who looks after it.
What self-hosted Kubernetes actually means
The term covers three different models:
- Your own hardware and your own team: you buy servers, place them in a data center, and build and operate the cluster yourself.
- Unmanaged Kubernetes from a local provider: the provider handles hardware, network, and the data center; you get a cluster with full admin access and run it yourself.
- Managed Kubernetes from a local provider: the provider handles upgrades, security, backups, and monitoring; you deploy applications.
What they share: the cluster doesn't run at a US hyperscaler (Amazon EKS, Azure AKS, Google GKE), but on infrastructure where you know exactly where it sits and who can access it.
Reason 1: Predictable costs
At a hyperscaler you pay separately for compute, disks, load balancers, NAT gateways, logs, and above all for data leaving the cloud. The monthly bill is hard to predict, and as the app grows it grows faster than the traffic itself.
The best-known example is 37signals, the company behind Basecamp. After moving its applications from AWS to its own servers, its annual cloud bill fell from $3.2 million to $1.3 million, with the remainder being S3 storage that it is moving as well. OneUptime moved a 28-node Kubernetes cluster from Amazon EKS to its own hardware and cut monthly costs from over $38,000 to roughly $8,000.
These savings aren't automatic. They work for steady, predictable load. If your app needs ten times the capacity for a few hours once a month, public cloud elasticity is real value.
Reason 2: Data under your control
The EU NIS2 directive has greatly expanded the number of companies with cybersecurity obligations. Thousands of businesses that previously had none must now, among other things, assess supplier risk, including whoever runs their infrastructure. In Czechia it applies through the new Cybersecurity Act, in force since 1 November 2025.
With US-headquartered providers, the US CLOUD Act also applies. It lets US authorities request data even when it physically sits in a European data center. The law doesn't ban US cloud, but in an audit it is much simpler to say "the data is in Prague and a European company operates it."
Reason 3: No lock-in to one vendor
Kubernetes is an open standard. If you build on the plain Kubernetes API and open tools such as Argo CD, cert-manager, or Gateway API, you can move your app to another provider without rewriting it. The more proprietary hyperscaler services you use, the more expensive leaving becomes. OneUptime could leave AWS precisely because its whole product ran on Kubernetes and Helm.
Reason 4: Close to your users and your support
For customers in Czechia and Slovakia, a data center in Prague or Bratislava is closer than Frankfurt or Amsterdam, which means lower latency. And with a local provider you talk to people who know your cluster, not an anonymous ticketing system.
When self-hosted Kubernetes doesn't make sense
- You have one or two small apps. Docker on a virtual server is simpler and cheaper.
- Nobody on your team understands Kubernetes, and you don't want to pay someone to manage it.
- Your load is extremely spiky, or you need to run on several continents at once.
- Your whole product is built on one cloud's proprietary services, and replacing them would cost more than you'd save.
The hidden costs of your own cluster
Running Kubernetes isn't a one-off installation. A new version ships roughly three times a year, and each is supported for about a year. If you run the cluster yourself, you have to handle:
- regular cluster upgrades without application downtime
- backups of cluster configuration and application data, and test restores
- security policies, network separation, and access rights
- certificates, ingress, and DNS
- monitoring, alerting, and on-call outside business hours
For a small team that can mean one person who no longer works on the product. Count the total cost, not just the price of servers.
Managed or unmanaged Kubernetes?
Unmanaged Kubernetes with admin access suits experienced teams that want full control and have operations know-how in-house. Managed Kubernetes suits teams that want to focus on development: they get their own namespaces and deploy applications, while the provider handles upgrades, security, backups, and monitoring.
How we do it at DENEVERSE
We run Kubernetes as a Service on our own infrastructure in Prague and Bratislava. A new cluster is typically ready within an hour of ordering, runs the current stable release, and we support the last two major versions. You can choose Admin access (unmanaged) or User access (managed).
- RKE2 cluster management through Rancher
- GitOps deployments with Argo CD
- security policies with Kyverno and network separation at the namespace level
- cluster and workload backups with Velero
- automatic certificates with cert-manager and Let's Encrypt
- SLA to match your needs: 5×8 or 24/7 support
Summary
Self-hosted Kubernetes isn't an ideology; it's a calculation. It makes sense with steady load, sensitive data, and growth that makes the hyperscaler bill hurt. It doesn't make sense for one small app. And if you don't want to build your own operations team, managed Kubernetes from a local provider gives you control over data and costs without the work of running the cluster.
Considering a move from AWS, Azure, or Google Cloud, or building your first cluster? Get in touch. In a free 30-minute call we'll work out which model and cluster size make sense for you.
Need a stable platform for your applications?
Contact usMore articles

Shared hosting, VPS, Docker, or Kubernetes: how to choose the right hosting
Don't pick a plan; pick an operating model. Five questions that decide whether shared hosting is enough or you need a virtual server or Kubernetes.
Read article →
How to host a vibe-coded app: from prototype to production
Your Lovable or Cursor app works on the first try. Production needs more: safe keys, a backed-up database, and a server someone actually looks after.
Read article →
