Hetzner Cloud
CostPilot supports Kubernetes clusters running on Hetzner Cloud — whether provisioned through Hetzner’s managed Kubernetes offering or self-managed clusters using tools such as hetzner-k3s or k3sup on Hetzner VMs. Pricing is derived from a static EUR price table bundled with CostPilot and converted to USD for display.
Pricing approach
Unlike AWS or Azure, Hetzner does not provide a public pricing API for programmatic consumption. CostPilot uses a curated static table of Hetzner Cloud server prices compiled from Hetzner’s public pricing page. This table is updated with each CostPilot release when Hetzner changes its prices.
A notable property of Hetzner pricing is that it is location-independent — the same server type costs the same amount whether it is running in Falkenstein (fsn1), Nuremberg (nbg1), Helsinki (hel1), Ashburn (ash), or Hillsboro (hil). This simplifies pricing lookup: CostPilot only needs the server type, not the location, to determine the correct price.
Hetzner publishes prices in EUR. CostPilot converts all Hetzner pricing to USD using a bundled exchange rate, which is updated with each release. If you require a specific EUR/USD rate, use the custom per-node rate override in Settings → Clusters → Cost configuration and enter the value in USD directly.
Supported server types
Hetzner Cloud organises its servers into three product lines. All three are supported by CostPilot:
CX series — Shared vCPU (Intel)
Entry-level instances with shared Intel vCPUs. Best suited for development environments, CI workloads, and lightweight production services.
| Size | vCPUs | Memory | Price (approx. EUR/month) |
|---|---|---|---|
cx22 | 2 | 4 GB | ~€4.35 |
cx32 | 4 | 8 GB | ~€8.21 |
cx42 | 8 | 16 GB | ~€14.91 |
cx52 | 16 | 32 GB | ~€29.33 |
CPX series — Shared vCPU (AMD)
AMD EPYC-based shared instances. Offer competitive pricing with generally better single-core performance than the CX series for the same price tier.
| Size | vCPUs | Memory | Price (approx. EUR/month) |
|---|---|---|---|
cpx11 | 2 | 2 GB | ~€4.35 |
cpx21 | 3 | 4 GB | ~€7.22 |
cpx31 | 4 | 8 GB | ~€12.65 |
cpx41 | 8 | 16 GB | ~€22.96 |
cpx51 | 16 | 32 GB | ~€40.42 |
CCX series — Dedicated vCPU (AMD)
Dedicated AMD EPYC vCPUs with no sharing. Recommended for production Kubernetes workloads where consistent CPU performance is important.
| Size | vCPUs | Memory | Price (approx. EUR/month) |
|---|---|---|---|
ccx13 | 2 | 8 GB | ~€18.07 |
ccx23 | 4 | 16 GB | ~€35.73 |
ccx33 | 8 | 32 GB | ~€70.01 |
ccx43 | 16 | 64 GB | ~€137.57 |
ccx53 | 32 | 128 GB | ~€272.70 |
ccx63 | 48 | 192 GB | ~€407.83 |
Prices shown are approximate and subject to change. CostPilot’s bundled price table reflects the values at the time of the last release. Actual costs may differ slightly.
Node identification
CostPilot identifies Hetzner nodes by matching the node.kubernetes.io/instance-type label against known Hetzner server type slugs. When using hetzner-k3s or similar provisioners, this label is set automatically. For manually provisioned clusters, ensure this label is set on each node with the correct Hetzner server type name (e.g. ccx23).
If the label is missing or does not match a known Hetzner server type, CostPilot will not be able to determine pricing for that node. In this case, configure a custom per-node rate in Settings → Clusters.
CPU and memory cost split
Hetzner publishes a single blended price per server type. CostPilot applies the following default split:
| Server series | CPU fraction | Memory fraction |
|---|---|---|
| CX (shared Intel) | 50% | 50% |
| CPX (shared AMD) | 50% | 50% |
| CCX (dedicated AMD) | 55% | 45% |
The CCX series leans slightly towards CPU because dedicated vCPUs command a proportionally larger share of the server cost compared to memory in that tier. You can adjust these ratios in Settings → Clusters → Cost configuration.
Spot and reserved instances
Hetzner Cloud does not offer spot or reserved pricing tiers. All servers are billed at the standard hourly rate. There is no discount pricing model to configure.
Tips for reducing Hetzner costs
- Use CCX for production, CPX for non-critical. Shared vCPU performance on CX/CPX can be unpredictable under load from other tenants. For production workloads, the price premium for CCX dedicated vCPUs is usually justified. CostPilot’s efficiency metrics can show whether you are getting consistent utilisation from your current server types.
- Hetzner is already cost-competitive. Hetzner pricing is significantly lower than major cloud providers for equivalent hardware. CostPilot’s value on Hetzner clusters is therefore more about right-sizing pods and identifying idle capacity than about switching pricing models.
- Location selection does not affect cost. If you have flexibility in where your cluster runs, choose a location based on latency and data residency requirements — not cost. All Hetzner locations are priced identically.
- Volume discounts are not reflected. Hetzner offers monthly volume discounts for large deployments under a custom contract. If you have negotiated pricing, use the custom rate override to reflect your actual cost.