Over the past year we moved 11 clusters off AWS and GCP onto Hetzner. Savings range from 50% to 78% depending on scenario. But numbers without context lie. Here's where the move pays off — and where it backfires.
The real economics
A 6-node cluster (32 vCPU, 64 GB RAM each) on AWS EKS runs about $4,200/month. The same cluster on Hetzner Cloud with self-managed Kubernetes is around $850. But to the latter you must add a DevOps engineer at 0.3–0.5 FTE — about $2,500/month. The bottom line: there's still savings, but less than the headline number.
When Hetzner is the right choice
- Steady-state workloads without big bursts (auto-scaling on Hetzner is its own story)
- European customers where you don't need a global CDN with 30+ regions
- A team that's willing to own infrastructure as code and monitoring
- Budget for an in-house DevOps engineer — Hetzner is not self-driving
When to stay on AWS/GCP
- 01Regular need for managed services: SageMaker, BigQuery, Bedrock
- 02Global delivery with P95 < 100 ms across 5+ continents
- 03Teams without DevOps — managed AWS is cheaper than poor self-managed
- 04Enterprise compliance requirements: SOC 2, HIPAA, FedRAMP
How we build clusters on Hetzner
- Talos Linux as the OS — minimum surprises, safe defaults
- Cilium instead of Calico — eBPF gives better observability
- Hetzner Cloud Controller Manager + CSI — native volumes and load balancers
- ArgoCD for GitOps — cluster restored from Git in 30 minutes
- Backup to Hetzner Storage Box + offsite to Wasabi
- Monitoring — VictoriaMetrics + Grafana + Loki on separate nodes
Hetzner is not “the cheap AWS”. It's a different philosophy: you pay less in money and more in engineering time. Choose deliberately.