Best Hetzner Cloud Alternatives After the 2026 Price Increase

For most of the last decade, Hetzner set the European price ceiling and everyone else had to explain why they cost more. On 15 June 2026 that stopped being true. The adjustment moved shared Intel and Arm plans (CX, CAX) up roughly 1.3x to 1.4x, which is irritating but liveable, and moved the AMD lines somewhere else entirely. CPX22 went from €7.99 to €19.49 a month. CPX32 went from €13.99 to €35.49. The dedicated CCX tiers rose between 2.1x and 2.7x, and US regions were hit harder still.

The hardware did not change. The tooling did not change. The Terraform provider is still excellent, the console is still fast, and the 20 TB of included traffic on EU instances is still generous. What changed is that "Hetzner is the cheapest good option" is now a statement about 2024, and every comparison built on it needs redoing.

This article ranks ten alternatives on price per core, API quality and what specifically you give up by moving. We are CloudBlast and we are first on this list, so read our entry with that in mind. There is a migration checklist at the end, and a section on who should simply stay, because for a good number of teams that is still the right answer.

What Actually Changed on 15 June 2026

The increase was not uniform, and which line you are on decides whether this is an annoyance or a reason to move.

US regions saw larger increases than EU ones, and US instances include 1 TB of traffic against the 20 TB EU instances get, so an American Hetzner deployment moved further than a German one.

The practical read: if you are on CX or CAX and your bill went up by a few euros, the tooling is probably worth it. If you are on CPX or CCX, your cost per core roughly doubled or worse, and the providers that were "slightly more expensive than Hetzner" last year are now cheaper than it.

How We Ranked These Alternatives

Price per core after the increase. Compared against a realistic small production box, roughly 2 to 3 cores, 4 GB of RAM and 30 GB or more of NVMe.

What replaces the tooling. Hetzner's API, Terraform provider and CLI are the real product for a lot of teams. An alternative that is cheaper and unautomatable is not cheaper.

Traffic terms. 20 TB included on EU instances is a high bar. Unmetered beats it, small metered allowances do not.

How hard the migration is. Snapshot portability, image format, whether floating IPs and private networking have equivalents, and how much Terraform you would rewrite.

Whether the price will hold. June happened because it could. We weighted providers that have not repriced under pressure, and we note that nobody can promise this, ourselves included.

Quick Comparison

Indicative entry pricing as of September 2026.

1. CloudBlast

Best for: teams leaving CPX or CCX who want EPYC and NVMe back at the price Hetzner used to charge for it.

Entry plan: VMA11, 1 AMD EPYC core, 3 GB RAM, 20 GB NVMe, €3.60/mo or €0.0049/h.

The direct comparison is the reason this article exists. A CPX22 is now €19.49 a month. Our VMA21, at 2 AMD EPYC cores, 4 GB of RAM and 30 GB of NVMe, is €4.80 a month or €0.0067/h. Both are shared AMD vCPUs on NVMe. That is roughly a quarter of the price for a comparable machine, and it is the specific arithmetic that has been driving migrations to us since June.

All eight plans run the same platform: AMD EPYC, NVMe, a 10 Gbps port, always-on DDoS mitigation with up to 2.5 Tbps of scrubbing capacity included. Traffic is unmetered in Amsterdam and Salt Lake City, so the 20 TB you are used to on an EU Hetzner instance becomes a number you stop tracking. Hong Kong is capped, from 500 GB up to 3 TB, and we publish that rather than pretending otherwise. Billing is hourly on every plan with a monthly cap, and add-ons keep the same model: backups €0.09/GB, block storage €0.15/GB, extra IPv4 €2.50/mo or €0.00347/h.

Where we are honest about the gap: Hetzner's Terraform provider is better than ours. We expose a public API and an MCP server, and if your infrastructure is defined in hcloud_server resources you will be rewriting that. If your deployment is Ansible, cloud-init, Docker Compose or a shell script, the move is close to trivial.

What we like: roughly a quarter of the post-increase CPX price for comparable specs, unmetered 10 Gbps, DDoS included rather than an add-on, hourly billing, no renewal increase.

What to watch: three regions (Amsterdam, Salt Lake City, Hong Kong) against Hetzner's larger footprint. No managed databases, no managed Kubernetes, no load balancer product, and no Terraform provider on par with theirs. If any of those four is load-bearing in your stack, budget for the replacement before you move.

2. netcup

Best for: staying in Germany with far more RAM and disk per euro, if you can accept contract terms.

netcup is the most direct German replacement for a CX or CPX box. A VPS 1000 G11 gives roughly 4 vCPU, 8 GB of RAM and 256 GB of storage for about €9.66/mo, with unmetered traffic on a 2.5 Gbps shared uplink. Against a post-increase CPX32 at €35.49, that is not a close comparison on resources.

The differences are structural. Plans are sold on monthly or annual contracts, sometimes with a setup fee, hourly billing is limited or absent depending on the line, and the practical price depends on which voucher code is live. Support is largely German-language. This is a hosting contract rather than a cloud subscription, and if your workflow involves creating and destroying servers programmatically, it is the wrong shape.

What we like: excellent resource allocation, German data centres, unmetered on a wide shared uplink, good uptime record, no exposure to cloud-style repricing.

What to watch: contract terms and setup fees, limited hourly billing, no first-party Terraform provider, confusing generation naming.

3. Contabo

Best for: maximum RAM and disk per euro, for workloads where the network is not the bottleneck.

If your Hetzner box was mostly memory, Contabo is the cheapest way to replace it: 3 vCPU, 8 GB of RAM and 75 GB of NVMe for $4.95/mo. Nothing else in this list matches the resource count, and Contabo has not raised prices at renewal, which after June is a point worth making.

What you lose relative to Hetzner is most of the network and all of the platform. The standard port is 200 Mbit against Hetzner's gigabit and up, disk I/O is shared aggressively, support is email-only with multi-day response times, and there is no hourly billing at all. As a replacement for a staging box, a self-hosted app server or a media library, it is excellent. As a replacement for a production API, think carefully.

What we like: unmatched RAM and storage per euro, stable renewal pricing, wide location list, free IPv4.

What to watch: 200 Mbit port, variable I/O under load, monthly commitment only, slow support, no automation story worth building on.

4. OVHcloud

Best for: European compliance and unmetered traffic from a company that owns its own backbone.

OVHcloud is the structural opposite of a provider that reprices under pressure: it owns its data centres and its network, and its VPS pricing has been stable while others moved. Entry plans are around $4.20/mo for 1 vCPU, 2 GB and 20 GB, stepping to roughly $6.30 for 2 vCPU and 4 GB, with unmetered traffic throughout.

The port is the compromise. Entry tiers run 100 Mbit to a few hundred, reaching 1 or 2 Gbps only on larger plans, so unmetered here is a lower ceiling than it sounds. The control panel is the other one: dense, slow, and a real step down from Hetzner's console. The compliance posture and EU sovereignty story, on the other hand, are stronger than almost anyone's.

What we like: unmetered traffic on every plan, owns its backbone, stable pricing, strong EU compliance credentials, official Terraform provider.

What to watch: narrow ports on entry tiers, clunky panel, no hourly billing on the VPS line, support quality varies by region.

5. Scaleway

Best for: teams whose real attachment to Hetzner is the infrastructure-as-code workflow.

Of everything here, Scaleway is the closest replacement for how Hetzner feels to operate. Real per-hour billing, a strong API, a well-maintained Terraform provider, a CLI, private networks, load balancers and managed services, all inside a French company with an EU sovereignty story.

The catch is that everything bills separately: instances per hour, flexible IPs at €0.004/h, snapshots per GB per hour. Your monthly forecast becomes an exercise, where Hetzner's was a multiplication. Stock in popular zones can be tight for the better instance families, and support is a paid tier. On raw price per core it does not match the value hosts here, so this is a like-for-like workflow replacement rather than a saving.

What we like: per-hour billing across the stack, strong Terraform provider and CLI, EU sovereignty, managed services and GPU in the same account.

What to watch: every resource meters separately, availability constraints, paid support, higher price per core than the value tier.

6. UpCloud

Best for: replacing a Hetzner box whose real bottleneck was disk rather than CPU.

UpCloud's MaxIOPS storage is genuinely faster than what most of this list runs, and if your Hetzner instance was a database that spent its life on small random writes, this is the upgrade that matters. Plans start around €4/mo, billing is hourly with a monthly cap, and there is a proper API and Terraform provider.

The location list is mid-sized rather than broad, covering the main European, US and APAC metros, and you pay more per core than the value hosts. It is a premium product priced in the middle, which after June makes it competitive with Hetzner on price for the first time in years.

What we like: best-in-class disk performance, hourly billing with a cap, official Terraform provider, reliable network.

What to watch: higher price per core than the value tier, mid-sized location list, no unmetered traffic option.

7. DigitalOcean

Best for: teams who would rather buy managed services than run more servers.

Moving from Hetzner to DigitalOcean is not a cost-saving migration and should not be sold as one. A 2 vCPU, 4 GB Droplet is $24/mo, well above even a post-increase CPX22. What you get is per-second billing since 1 January 2026, pooled transfer across the account, managed Postgres, MySQL and Redis, Spaces object storage, App Platform, and documentation that has no equal.

It is the right move if the reason you were self-managing a database on a cheap VPS was that the managed version cost more than the box. Once the box costs €19.49, that calculation changes, and consolidating a Postgres instance into a managed service can be cheaper overall than the sum of the servers it replaces.

What we like: per-second billing, pooled transfer, excellent managed services and documentation, strong Terraform provider.

What to watch: the most expensive raw compute on this list, egress overage at $0.01/GiB, easy to spend more than you did on Hetzner.

8. Vultr

Best for: anyone whose Hetzner constraint was the location list.

Hetzner runs a handful of regions. Vultr runs more than thirty, and if the reason you are shopping is that you needed a box in Sydney, São Paulo, Johannesburg or Mumbai, this is the shortest path. Hourly billing with a 672-hour monthly cap, bare metal and GPU in the same panel, and an official Terraform provider.

Skip the $2.50 Regular Performance tier, which is shared Intel with SATA SSDs. High Performance and High Frequency from around $6/mo are the tiers that compare to a CPX, on AMD EPYC or high-clock Xeon with NVMe. At that price it is meaningfully cheaper than the new CPX rates for similar hardware.

What we like: more than thirty locations, hourly billing with a cap, EPYC on the right tiers, bare metal and GPU alongside.

What to watch: the entry tier is a trap for price comparisons, modest bandwidth allowances on small plans, billed egress overage.

9. Akamai Cloud (Linode)

Best for: teams who want the thing Hetzner just stopped being, which is a provider whose prices do not move.

The $5 Nanode has been $5 for a very long time, and the 2 GB plan has been $12 through an acquisition by Akamai that could easily have gone the other way. After June, "nothing has happened here in years" is a legitimate selling point.

Transfer pools across the account at 1 TB on the entry plan, billing is hourly with a monthly cap, and the API, CLI and Terraform provider are all solid. The weakness is unchanged: shared tiers are underpowered for the money and dedicated tiers start around $36 to $45/mo, so you are buying stability rather than value.

What we like: genuinely stable pricing, pooled transfer, good API and CLI, extensive guides, wide region list.

What to watch: thin specs per dollar on shared tiers, expensive dedicated tiers, documentation harder to navigate since the rebrand.

10. Cherry Servers

Best for: CCX refugees who now find bare metal competitive.

This is the entry that only makes sense because of the increase. Dedicated CCX tiers rose between 2.1x and 2.7x, and once you are paying that for dedicated virtual cores, an entire physical EPYC machine from around $80/mo stops looking extravagant. You get the whole CPU, all the memory bandwidth, no steal time, and an API to provision it with.

Do the arithmetic on your specific tier before assuming it works, because the crossover depends on how much CCX you were buying. Provisioning is minutes rather than seconds, and the location list is smaller than any cloud here.

What we like: whole physical EPYC CPUs, API-driven provisioning, hourly options on some builds, no noisy neighbours by definition.

What to watch: high entry price, smaller location list, slower provisioning, overkill unless you genuinely needed dedicated cores.

What You Actually Lose by Leaving Hetzner

Be honest about this before you move, because a migration that costs you a week is not a saving on a €15 monthly difference.

The Terraform provider. hcloud is one of the best in the industry. Most alternatives here have something, and only Scaleway, DigitalOcean and Vultr have something comparable. If your infrastructure is fully declared in Terraform, price the rewrite in hours and compare it against the monthly delta.

20 TB of included traffic on EU instances. Only unmetered providers clearly beat that. Metered alternatives with 1 TB allowances are a step down that shows up on the invoice, not the spec sheet.

Private networking and load balancers. Hetzner's are mature, cheap and well integrated. Several providers here have no equivalent, which turns a managed load balancer into an HAProxy box you now maintain.

Snapshots, floating IPs and firewalls that all work together. The cloud firewall in particular is easy to take for granted until you are writing nftables rules again.

Volume pricing and the console. Both are good. Nothing on this list is dramatically better.

If you use none of that beyond compute plus a snapshot, your migration is an afternoon. If you use all of it, the cheaper provider is not necessarily cheaper.

Migration Checklist

A sequence that works for a typical single-server or small-fleet move.

1. Inventory what you actually use. Servers, volumes, floating IPs, firewall rules, load balancers, private networks, snapshots. Anything on that list without an equivalent at the new provider is work, not configuration.
2. Rebuild rather than lift. Snapshots are not portable between providers in practice. Provision a clean instance and reapply your configuration management. If you have none, this is the moment to write a short Ansible playbook or a docker compose file, because you will need it again.
3. Move data with rsync over SSH, or restore from your existing backups. Time a test run first: 100 GB over a 200 Mbit link is roughly an hour and a half, over 10 Gbps it is minutes. This is where port speed stops being theoretical.
4. Drop DNS TTL to 300 seconds at least 24 hours before the cutover, so the switch is fast when you make it and reversible if it goes wrong.
5. Run both in parallel for a few days. Hourly billing makes this nearly free, and it is the only way to compare real load rather than benchmarks. A week of overlap on a €4.80 instance is about €1.10.
6. Recreate firewall rules explicitly. This is the step people skip, and it is how a database ends up on a public IP. Verify with nmap from outside before you send traffic.
7. Verify backups on the new provider before destroying the old one. Restore one, actually boot it. An untested backup is a hope.
8. Keep the old server for a week after cutover. Then destroy it, because on hourly billing an instance you forgot about is the only way this migration costs more than it saves.

Who Should Stay on Hetzner

Not everyone should move, and pretending otherwise would be a sales pitch rather than a recommendation.

Stay if you are on CX or CAX. The increase was 1.3x to 1.4x on lines that were already inexpensive. Arm CAX instances in particular are now Hetzner's value line and remain competitively priced if your software runs on Arm.

Stay if your infrastructure is fully Terraformed. The provider quality is a real asset. A rewrite plus the risk of a migration is worth more than a few euros a month unless the fleet is large.

Stay if you depend on the platform. Load balancers, private networks, cloud firewalls and volumes that all integrate cleanly are not free to replace, and the replacement is usually a server you now operate.

Stay if you are in a Hetzner-only region and latency matters. Being close to your users beats being cheap.

Move if you are on CPX or CCX and buying compute. That is the case where your cost per core roughly doubled or worse for hardware you can rent elsewhere at the old price. It is also, judging by our own signups since June, most of the people asking the question.

Frequently Asked Questions

The Verdict

CloudBlast if you were on CPX or CCX and want EPYC with NVMe and unmetered 10 Gbps back near the old price, from €3.60/mo. netcup for the most German RAM per euro, if contract terms are acceptable. Contabo for capacity when the network does not matter. OVHcloud for unmetered traffic and EU compliance from a network owner. Scaleway if the infrastructure-as-code workflow is what you are really replacing. UpCloud if disk was your bottleneck. Akamai for prices that have not moved in years. Cherry Servers if a CCX bill now buys you bare metal.

And stay on Hetzner if you are on CX or CAX with a Terraformed stack. The increase was real, and it did not make every Hetzner deployment a mistake. It made the AMD lines a decision.