Grandfathered pricing, explained
Hetzner let you keep your old prices, but only if your servers are never deleted or rescaled. Normal Kubernetes deletes them constantly. paasbox does not. This is why the adopted-server feature exists, and it applies both to a cluster on your own landscape and to a cluster I run for you.
What happened on 15 June 2026
Section titled “What happened on 15 June 2026”Hetzner raised prices for new Cloud orders, effective 15 June 2026, 08:00 CEST. The increases are large: the AMD CPX line rose 144–173%, dedicated CCX ~122%, and the Intel CX / Ampere CAX lines roughly 30%.
The adjustment applies to new orders and to rescaling existing instances. A server you ordered before the cutoff keeps its old price, as long as that exact instance continues to exist.
Pets vs. cattle: why normal Kubernetes loses the rate
Section titled “Pets vs. cattle: why normal Kubernetes loses the rate”Grandfathering survives only while the server instance is never deleted or rescaled:
- Pets keep the price. Stop/start and rebuild (OS reinstall) all keep the same server, so the rate holds. Only deleting it, or resizing it, loses the rate.
- Cattle lose the price. Gardener and other Kubernetes managers replace worker nodes, that is delete and recreate them, on every Kubernetes or image upgrade, every autoscale event, and every node-health failure. Each replacement is a new order at the new price.
So an ordinary Kubernetes cluster re-orders its whole fleet at the new price within weeks. The grandfathered rate is gone without anyone deciding to give it up.
How paasbox keeps it
Section titled “How paasbox keeps it”paasbox adopts the servers you want to keep into a pooled node pool that is rebuilt in place and never deleted. On every cluster operation, upgrades included, the instance is reused, not replaced. The server continues to exist, so its locked-in price does too.
(In detail: adopted servers are managed as never-delete pool members, with Hetzner delete-protection as a second guard. The on-demand “managed” pools you can add for burst behave like normal Kubernetes nodes, at current prices.)
What it’s worth
Section titled “What it’s worth”The figures below are Hetzner list prices, gross (incl. 19% VAT, excl. IPv4). “You’d lose / mo” is the increase you avoid on each server by keeping it grandfathered.
| Type | vCPU / RAM | old gross | new gross | you’d lose / mo |
|---|---|---|---|---|
ccx43 | 16 ded / 64 | €148.74 | €328.43 | +€179.69 |
ccx33 | 8 ded / 32 | €74.36 | €164.80 | +€90.44 |
cpx42 | 8 / 16 | €30.33 | €82.69 | +€52.36 |
ccx13 | 2 ded / 8 | €19.03 | €51.16 | +€32.13 |
cpx32 | 4 / 8 | €16.65 | €42.23 | +€25.58 |
cpx22 | 3 / 4 | €9.51 | €23.19 | +€13.68 |
cax31 | 8 ARM / 16 | €19.03 | €24.98 | +€5.95 |
cx43 | 8 / 16 | €14.27 | €19.03 | +€4.76 |
A small fleet of 4× cpx42 loses ~€209/mo the moment Kubernetes rolls those nodes at the new
price. A cluster I run for you keeps them grandfathered from €49 a month (a test cluster), or
€99 for a production cluster with a highly available control plane, which is net positive from
the first month, before counting the self-hosted three-node control plane (~€127/mo) you no longer
run.
What would you save?
Tell us the pre-increase Hetzner servers you'd keep on their grandfathered rate.
Estimates in euro, gross (incl. 19% VAT), for one cluster. paasbox is €49 (test cluster) or €99 (production cluster) / cluster / month excl. VAT — VAT-registered businesses reclaim VAT, so net figures are roughly these ÷ 1.19. “Savings” is the price increase youavoid by keeping your servers on their grandfathered rate — not your total infrastructure cost. Your Hetzner bill for the nodes themselves is separate and paid to Hetzner, and an idle node still bills. Prices: Hetzner list, 2026-06-15.
An honest note
Section titled “An honest note”Grandfathering does not make cost elastic. You own the servers, so an idle pooled node still bills Hetzner; keeping a server powered off preserves its price but does not save money. The savings are real, but they come from three specific things:
- the grandfathered rate on servers you would be running anyway,
- not paying for a control plane (nodes, load balancers, and the operations around them), and
- sending spikes to on-demand nodes instead of over-provisioning a fixed fleet.
Your Hetzner bill for your own nodes is always separate, and paid directly to Hetzner.