I wanted to compare the prices to the current prices:
> That puts us up to €26,97 a month on our six machines so far.
The cost for this same calculation has risen to €41,94 for the exact same configuration (3x CX23, 3x CX33).
> These two additional load balancers (lb11) bump up the price with €5.39/mo.
lb11 is now €7,49 x 2 = €14,98
> A public IP address costs €1,70/mo which adds an additional €10,20
According to Hetzner's pricing page, this is now free on CX23, but costs an additional €0,50 on CX33.
> ...making our total €47,95/mo pre-VAT
€41,94 + €14,98 + €1,50 = €58,42, or roughly a 22% increase. Still over €10 cheaper than the prices the article quotes for a managed cluster on AWS/GCP, but as someone that does this for a living, €10-15 extra is a very negligible price to pay to avoid the amount of headaches you'll get by having to manage and upgrade the cluster yourself.
Not familiar enough about GCP internal, but AWS prices are way higher than what's quoted in the article. You'll get billed for bandwidth, should you want any logs of your control plane you'll pay them at a premium so high* it will dwarf the cluster price and machine price combined. And for ~$70/month, you get the cluster API, not the EC2 compute required for worker nodes.
I love Hetzner for some of the cheapest cloud components that can do anything.
One use case is a custom ACME cert broker and DNS server. Another is their storage boxes split up into sub accounts for my and friends backups and music.
I've moved all my web stuff over to exe.dev, but I also turn my Hetzer box into a similar "Agent Computer" by dropping in a binary with `claude` setup and a chat server:
My experience with Hetzner VM pricing has turned me off of considering them for anything else. Maybe the experience in other world regions is not the same but this is what happened to us:
I started to use Hetzner's VM options in the Virginia, US datacenter. They had a great performance to cost ratio when I compared them to Linode/Akamai and Digital Ocean about a year ago. At that time, 4vcpu / 8GB server cost $24.99 per month.
Now, to get a VM like it, it's quoted $73.49 a month. Almost a 3x change.
I understand data center economics are changing due to LLM usage. But the Hetzner jump has not been mirrored by their competitors. I'm disappointed I made the switch from Linode. We still have some older VMs there and we'll likely continue to use them instead of Hetzner moving forward.
You may have just caught the wrong end of it timing wise. They used to be super stable and cheap but being a low cost provider they have no real buffer to absorb even minor hikes let alone the shitshow we saw in ram/ssd.
This has been my experience too. I migrated a few dozen projects from DO to Hetzner, and my monthly bill went from about $80 to around $162. The only thing that changed was Hetzner's rate hikes.
Now I'm in a weird spot. I could downgrade some servers, but changing them would immediately move me onto the new pricing. So I either pay more for less hardware, or pay a lot, lot more if I ever need to scale back up.
They're the first, I strongly doubt they'll be the last to adjust prices that much. The big cloud providers might have enough margins to eat the cost increases, especially since bare VMs are only a small part of their offer. But I don't see how the smaller, more VM-focused providers will be able to absorb the high hardware prices.
Even with the new prices, this is almost twice as cheap as Scaleway for comparable hardware.
---
Only the NAT gateway and the LBs have public IPv4 addresses.
The CPX22 NAT gateway is a little overkill, but I was experimenting with making it both the NAT gateway and the load balancer, and I figured I needed dedicated CPU cores to handle inbound traffic. It was an attempt to cut the cost of Hetzner's own LBs, if it proved to scale. I haven't got around to that part, but today the economics wouldn't add up: LBs are cheaper and less work.
I've been running k3 on a single bigger hetzner dedicated machine to power https://demo.opensuite.online, a work-in-progress open source replacement for Google Workspace, a distribution of what the Dutch government is trying to switch to but they seemingly don't care about UX at all so I wanted to try build on their foundations (which in turn are basically just nextcloud, element, and some others).
Anyway can definitely recommend, it feels like a real cluster but it's all on a single piece of hardware which makes snapshots etc super simple.
Went a slightly different angle - Rasp 4s aren't impressing anyone these days, but 3 of them make for a fine HA controlplane. Once you've got a stable HA controlplane its much easier to connect whatever hardware you have lying around
(Just don't use SD cards...use M10 optanes). Also...Talos not k3s.
Instead of Rasp 4s, just get a dell optiplex micro, hp elitedesk, lenovo m920q etc - much more stable, quite low power, and can also use it as a true node, and comes with quick sync for video related tasks.
Install proxmox onto each, create talos VM(s) on each for kubernetes nodes, and you are off to races.
"Cheap" in terms of money, ~30K in terms of lines of code that you eventually gonna need to bite the apple and read through and understand, 10K of which is HCL.
End cost seems to be ~60 EUR/month for a cluster of machines, which isn't too bad. Still, I suppose once you're even dealing with such small amounts, even thinking about distributed architecture or similar things feels a bit too early.
Given 6 months have passed since this article has been published, would you also pick MicroOS? I'm currently setting up a single-node k8s cluster on Hetzner but I went with Talos. Curious about your takes, IaC and k8s is new terrain for me.
For those interested in this I also recommend autobase, a set of ansible roles for a highly available postgres cluster on hetzner or other clouds: https://github.com/autobase-tech/autobase
Scaleway has a great managed kubernetes offering. They host the control plane and all you pay are the instances starting at 30-something € per instance-month.
I wanted to compare the prices to the current prices:
> That puts us up to €26,97 a month on our six machines so far.
The cost for this same calculation has risen to €41,94 for the exact same configuration (3x CX23, 3x CX33).
> These two additional load balancers (lb11) bump up the price with €5.39/mo.
lb11 is now €7,49 x 2 = €14,98
> A public IP address costs €1,70/mo which adds an additional €10,20
According to Hetzner's pricing page, this is now free on CX23, but costs an additional €0,50 on CX33.
> ...making our total €47,95/mo pre-VAT
€41,94 + €14,98 + €1,50 = €58,42, or roughly a 22% increase. Still over €10 cheaper than the prices the article quotes for a managed cluster on AWS/GCP, but as someone that does this for a living, €10-15 extra is a very negligible price to pay to avoid the amount of headaches you'll get by having to manage and upgrade the cluster yourself.
Not familiar enough about GCP internal, but AWS prices are way higher than what's quoted in the article. You'll get billed for bandwidth, should you want any logs of your control plane you'll pay them at a premium so high* it will dwarf the cluster price and machine price combined. And for ~$70/month, you get the cluster API, not the EC2 compute required for worker nodes.
* Because vended logs to s3 are not supported (https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/AWS..., https://github.com/aws/containers-roadmap/issues/1141)
I love Hetzner for some of the cheapest cloud components that can do anything.
One use case is a custom ACME cert broker and DNS server. Another is their storage boxes split up into sub accounts for my and friends backups and music.
I've moved all my web stuff over to exe.dev, but I also turn my Hetzer box into a similar "Agent Computer" by dropping in a binary with `claude` setup and a chat server:
https://github.com/housecat-inc/scratch
My experience with Hetzner VM pricing has turned me off of considering them for anything else. Maybe the experience in other world regions is not the same but this is what happened to us:
I started to use Hetzner's VM options in the Virginia, US datacenter. They had a great performance to cost ratio when I compared them to Linode/Akamai and Digital Ocean about a year ago. At that time, 4vcpu / 8GB server cost $24.99 per month.
Now, to get a VM like it, it's quoted $73.49 a month. Almost a 3x change.
I understand data center economics are changing due to LLM usage. But the Hetzner jump has not been mirrored by their competitors. I'm disappointed I made the switch from Linode. We still have some older VMs there and we'll likely continue to use them instead of Hetzner moving forward.
You may have just caught the wrong end of it timing wise. They used to be super stable and cheap but being a low cost provider they have no real buffer to absorb even minor hikes let alone the shitshow we saw in ram/ssd.
This has been my experience too. I migrated a few dozen projects from DO to Hetzner, and my monthly bill went from about $80 to around $162. The only thing that changed was Hetzner's rate hikes.
Now I'm in a weird spot. I could downgrade some servers, but changing them would immediately move me onto the new pricing. So I either pay more for less hardware, or pay a lot, lot more if I ever need to scale back up.
Yeah, RAM shortage has caught up with them. Send a thank you note to Anthropic and OpenAI.
They're the first, I strongly doubt they'll be the last to adjust prices that much. The big cloud providers might have enough margins to eat the cost increases, especially since bare VMs are only a small part of their offer. But I don't see how the smaller, more VM-focused providers will be able to absorb the high hardware prices.
I also run Kubernetes on Hetzner and created my nodes prior to two price hikes.
Here's how expenses look and would look if I recreated it today:
Even with the new prices, this is almost twice as cheap as Scaleway for comparable hardware.---
Only the NAT gateway and the LBs have public IPv4 addresses.
The CPX22 NAT gateway is a little overkill, but I was experimenting with making it both the NAT gateway and the load balancer, and I figured I needed dedicated CPU cores to handle inbound traffic. It was an attempt to cut the cost of Hetzner's own LBs, if it proved to scale. I haven't got around to that part, but today the economics wouldn't add up: LBs are cheaper and less work.
This is Talos Linux: https://talos.dev
Talos can be upgraded via talosctl, so even less maintenance than k3s.
The NAT gateway effectively is a bastion host and would serve fine as a VPN gateway.
The nodes themselves can join a tailnet,
Is it per hour/day/month?
I've clarified; it's per month.
I've been running k3 on a single bigger hetzner dedicated machine to power https://demo.opensuite.online, a work-in-progress open source replacement for Google Workspace, a distribution of what the Dutch government is trying to switch to but they seemingly don't care about UX at all so I wanted to try build on their foundations (which in turn are basically just nextcloud, element, and some others).
Anyway can definitely recommend, it feels like a real cluster but it's all on a single piece of hardware which makes snapshots etc super simple.
I've been doing the same for over a year. I think single node kubernetes is very underrated, both for running a single project or for multiple.
> seemingly don't care about UX
That seems to be the case on every project with the folks that sign the checks or contracts for integrating other vendors.
I think Jobs was the rare exception there.
> Cheap
> Hetzner cloud
> published in December 2025
which means that it's twice as expensive now as it was when the article was published
So are computers
Went a slightly different angle - Rasp 4s aren't impressing anyone these days, but 3 of them make for a fine HA controlplane. Once you've got a stable HA controlplane its much easier to connect whatever hardware you have lying around
(Just don't use SD cards...use M10 optanes). Also...Talos not k3s.
Instead of Rasp 4s, just get a dell optiplex micro, hp elitedesk, lenovo m920q etc - much more stable, quite low power, and can also use it as a true node, and comes with quick sync for video related tasks.
Install proxmox onto each, create talos VM(s) on each for kubernetes nodes, and you are off to races.
I've got worker nodes on m920qs and proxmox VMs too for workers, but like the clean segregation of a HA control node group.
I doubt it matter much either way tbh. K8S is inherently resilient.
I did have the rasps on hand though...not sure I'd go and specifically buy them just to achieve said segregation.
Also have my website running on a Hetzner k8s cluster https://simon-frey.com/blog/kubernetes-on-hetzner-cloud/
I developed a Karpenter provider for Hetzner recently. Might interesting for you to integrate.
https://github.com/paperclipinc/karpenter-provider-hetzner
"Cheap" in terms of money, ~30K in terms of lines of code that you eventually gonna need to bite the apple and read through and understand, 10K of which is HCL.
End cost seems to be ~60 EUR/month for a cluster of machines, which isn't too bad. Still, I suppose once you're even dealing with such small amounts, even thinking about distributed architecture or similar things feels a bit too early.
Terraform has been such a delight in the age of LLMs.
Given 6 months have passed since this article has been published, would you also pick MicroOS? I'm currently setting up a single-node k8s cluster on Hetzner but I went with Talos. Curious about your takes, IaC and k8s is new terrain for me.
For those interested in this I also recommend autobase, a set of ansible roles for a highly available postgres cluster on hetzner or other clouds: https://github.com/autobase-tech/autobase
You really put your admin credentials in a blog post. Lucky I can't seem to hit that IP.
Scaleway has a great managed kubernetes offering. They host the control plane and all you pay are the instances starting at 30-something € per instance-month.