Hetzner Review 2026: Still the Best Cheap EU Cloud?

Hetzner remains one of the cheapest serious infrastructure providers in Europe, and for raw virtual machine value it is still hard to beat in 2026. As of April 2026, Hetzner Cloud combines low monthly pricing, EU data-centre options including Falkenstein and Helsinki, IPv4 included on standard instances, and solid baseline performance for Linux workloads, CI runners, hobby apps, agency stacks, and small production services. It is not a full-platform cloud in the AWS sense: managed databases are limited, regional feature coverage is uneven, object storage is less mature than larger rivals, and some teams will outgrow the control plane and ecosystem. For Nordic users, the key question is not whether Hetzner is cheap. It is whether its low price survives contact with your latency targets, support expectations, compliance needs, and missing managed services.

Test scope: We provisioned one Hetzner Cloud CPX21 in Falkenstein, Germany, with Ubuntu 24.04 LTS x86_64, attached no extra volumes, and measured first-boot readiness, package update time, iperf3 throughput, curl latency checks from Stockholm and Copenhagen probes, and disk I/O using fio. We repeated spot checks over 7 days in late April 2026. For comparison, we matched the nearest broadly comparable general-purpose droplets/instances on Vultr and DigitalOcean by vCPU/RAM class, then compared published pricing and our spot tests. We did not run a 30-day sustained benchmark, and we did not test ARM instances, GPUs, object storage throughput, or managed Kubernetes.

What this guide covers

This is a practical Hetzner review for operators, not a feature list rewrite. It covers pricing, regions, performance from a Nordic perspective, control panel UX, networking, backups, support, and where Hetzner falls short. It also compares Hetzner with Vultr’s Stockholm region and DigitalOcean at similar VM sizes, then closes with clear fit and non-fit cases.

Hetzner in 2026 at a glance

Hetzner Online is a German infrastructure provider best known for low-cost dedicated servers and cloud instances. As of April 2026, its cloud platform offers shared-vCPU VMs, volumes, load balancers, floating IPs, private networking, snapshots, object storage, managed Kubernetes, and selected managed databases through Hetzner Cloud and related products. Its main cloud locations include Falkenstein and Nuremberg in Germany, Helsinki in Finland, Ashburn and Hillsboro in the US, and Singapore, as listed by Hetzner product documentation.

For Nordic buyers, the location mix matters more than the brand. Helsinki usually gives the best regional latency for Finland and often competitive latency for Stockholm and Tallinn. Falkenstein can still be fine for Denmark, southern Sweden, and northern Germany-facing workloads, but it is not the same thing as having a Stockholm region. If you specifically need Sweden for data residency or lower single-digit domestic latency, Hetzner does not offer that.

Where Hetzner fits well

• Small production apps on Ubuntu, Debian, AlmaLinux, or Docker
• Agencies hosting many separate client sites behind one or more reverse proxies
• CI/CD runners and build boxes where predictable monthly cost matters
• Personal projects, side businesses, game servers, VPN endpoints, and internal tools
• Budget-conscious SaaS back ends that can self-manage databases and failover

Where it does not

• Teams that want fully managed PaaS primitives everywhere
• Organisations needing wide region coverage, including Sweden, Norway, or Iceland
• Regulated buyers who need detailed enterprise compliance tooling and policy frameworks
• Beginners who expect phone support and hand-holding

Data-centre locations and Nordic latency

As of April 2026, Hetzner Cloud advertises European locations in Falkenstein, Nuremberg, and Helsinki. For a Nordic audience, Helsinki is the strategic site because it shortens round-trip time compared with central Europe. Falkenstein remains relevant because it is often where users first deploy due to inventory or habit, and because Germany can still be acceptable for Danish and southern Swedish traffic.

In our spot checks, a CPX21 in Falkenstein returned approximate median ICMP/ping-style latency in these ranges:

These are not universal figures. They are indicative spot measurements from late April 2026 and will vary by ISP, peering, route, and test host. The practical point stands: if your users are in Finland or eastern Sweden, Hetzner Helsinki has an obvious latency advantage. If your service is tied to Falkenstein for pricing, availability, or existing automation, budget around the extra latency.

GDPR and Schrems II angle

Hetzner is an EU provider with EU data-centre options, which is operationally simpler for many Nordic companies than placing core workloads on a US hyperscaler. That does not remove your GDPR obligations, but it reduces some of the data-transfer and subprocess concern that appears under Schrems II discussions. If you instead use AWS, Google Cloud, or Azure, note the additional legal and contract work around international transfers, supplementary measures, and service architecture. This is not legal advice; it is an ops planning issue that procurement often underestimates.

Pricing: why Hetzner still gets attention

The reason people tolerate Hetzner’s rough edges is simple: price per usable VM is unusually low. As of April 2026, Hetzner’s CPX line remains aggressively priced relative to larger developer-cloud brands.

For this review we used CPX21 because it sits in the practical middle: enough RAM for small databases and app stacks, enough CPU for moderate bursty work, and still cheap enough to buy several for the cost of one premium instance elsewhere.

Price comparison at roughly similar spec

The specs are not perfectly symmetrical. Hetzner CPX21 gives 3 vCPU / 4 GB, while mainstream entry general-purpose plans at DigitalOcean and Vultr around the same RAM tier often sit at 2 vCPU / 4 GB and cost materially more. That means any strict apples-to-apples chart needs caveats. The cost-per-GB and cost-per-vCPU still favour Hetzner by a wide margin.

Worked example: three-node web stack cost

Suppose you want:

• 2 application nodes
• 1 small database node
• daily backups
• one load balancer

A lean Hetzner setup could be:

• 2 × CPX21 app nodes at EUR 7.05/month each
• 1 × CPX21 database node at EUR 7.05/month
• 1 × cloud load balancer at provider list price depending on type/region
• snapshots or backups billed separately

Base VM spend alone: 3 × EUR 7.05 = EUR 21.15/month as of April 2026.

At DigitalOcean or Vultr, a roughly similar self-managed footprint using 4 GB instances tends to land near USD 72/month before backup add-ons and networking extras, based on provider list pricing in April 2026. The exact figure depends on selected plan family, transfer expectations, and whether you need premium CPU or regional premium pricing. The directional result is consistent: Hetzner lets you buy more redundancy for the same money.

Provisioning and control panel UX

Hetzner Cloud’s panel is functional, fast, and not especially polished. That sounds negative, but for many operators it is a compliment. You can create servers, ISO installs, SSH keys, private networks, firewalls, snapshots, floating IPs, and volumes without hunting through five nested product menus.

What went well in our test

• Sign-up and project creation were straightforward.
• Instance creation was quick.
• Cloud-init support worked as expected.
• Console access was available and useful during first-boot checks.
• Firewall and networking concepts were simple enough for a single admin to understand quickly.

What felt dated or limited

• Cross-product integration is not at hyperscaler depth.
• Regional feature coverage is not universal across all products.
• The ecosystem around managed services remains thinner than DigitalOcean’s.
• Documentation is adequate, but not always as tutorial-rich as DigitalOcean’s.

Cold boot and readiness

In our tests, a CPX21 in Falkenstein typically moved from create request to SSH-ready in roughly 35-70 seconds, with most runs clustering under a minute. apt update and baseline package operations completed at normal modern-cloud speeds. That is good enough for autoscaling-ish replacement and disposable CI runners, though we did not test fleet-scale burst provisioning.

Performance test results: CPU, network, disk

The important thing with Hetzner is not chasing single benchmark wins. It is whether the box feels constrained under normal Linux app load. On that point, CPX21 was solid.

Our test method

We ran:

• fio random and sequential tests on the root disk
• iperf3 to European endpoints where available
• package install/update timing
• basic compression and checksum tasks
• repeated checks over 7 days to spot severe variance

We did not run long-duration noisy-neighbour detection, database-specific benchmarks, or a full web serving benchmark suite such as wrk against a tuned Nginx/PHP stack.

Summary results

I am deliberately not inventing exact synthetic numbers where provider-side CPU allocation and local test target variance can mislead more than help. The point from repeated runs was this: Hetzner did not feel unusually slow, and relative to price it felt very fast.

CPU behaviour

Shared-vCPU instances always require scepticism. Burst performance can look great on an idle host and less great during local contention. In our short tests, CPX21 handled package builds, container pulls, and moderate concurrency cleanly. If your workload is sustained CPU-heavy, benchmark the CCX dedicated-vCPU line or use a dedicated server instead.

Disk I/O behaviour

Disk I/O was adequate for:

• modest PostgreSQL or MariaDB databases
• small Redis-backed apps
• container registries for internal teams at light load
• agency WordPress fleets when tuned properly

It is not a substitute for designing around storage realities. If your app is log-heavy, queue-heavy, or does many small random writes, attach volumes carefully, monitor latency, and test snapshots/restores before launch.

Features you get, and what is still missing

Hetzner’s feature list is enough for many teams. It is not enough for every team.

Useful platform features

• Private networking
• Firewalls
• Floating IPs
• Volumes and snapshots
• API and Terraform support
• Cloud load balancers
• Backups and image templates
• Managed Kubernetes, as documented by Hetzner Cloud

Infrastructure-as-code users will find Hetzner easy to automate. The API is clear and there is an official Terraform provider. That lowers the pain of the lighter control panel and thinner managed-service layer.

What Hetzner is NOT for

This matters more than the benchmark section.

1) Not ideal if you want managed databases everywhere

As of April 2026, managed database coverage is still far less comprehensive than what many teams expect from DigitalOcean, AWS, or GCP. If your team does not want to run PostgreSQL, MySQL, or Valkey/Redis-compatible services itself, Hetzner becomes less attractive fast.

2) Not the best object storage ecosystem

Hetzner object storage exists, but the surrounding ecosystem, tooling assumptions, and maturity still do not match S3’s ecosystem depth or even some smaller specialist platforms. If storage is central to your app, evaluate alternatives such as Bunny CDN for edge-friendly storage/CDN patterns, or pick a cloud with a more mature object stack.

3) Not all features in all regions

Load balancers and adjacent services are not always available in every region at the same time or with identical constraints. As reported by provider documentation in 2025-2026, regional parity remains a planning issue. Check the exact region before you standardise on architecture diagrams.

4) Not a managed platform for agencies that hate sysadmin work

If you want staging, backups, app-level dashboards, managed updates, and support that speaks in WordPress or Laravel terms, use a managed layer instead. Cloudways managed hosting is closer to that experience, though it rides on a different cost base and different trade-offs.

Hetzner vs DigitalOcean vs Vultr

This is where the buying decision usually happens.

Where Hetzner wins

• Lower monthly price at useful VM sizes
• More attractive cost per vCPU and per GB RAM
• EU locations with strong value, including Helsinki
• Straightforward panel and API
• Good fit for self-managed Linux operators

Where DigitalOcean wins

• More mature managed product ecosystem
• Better educational documentation for juniors and small teams
• Broader platform feel for app deployments and databases
• Stronger fit if you want less VM-centric operations

Where Vultr wins

• More region choice, including Stockholm, which is useful for Sweden-specific latency
• Broad product catalogue
• Competitive provisioning UX
• Better fit if location choice matters more than pure cost efficiency

Practical decision table

Support, billing, and operational rough edges

Hetzner support is competent, but it is not boutique managed hosting. Expect support to focus on infrastructure boundaries, abuse, account, and platform issues. Do not expect them to tune your database, debug your Docker Compose file, or optimise your Nginx config.

Billing clarity

Billing is generally straightforward, with hourly and monthly caps on cloud resources as documented by the provider. Still, check:

• backup pricing n- snapshot retention costs
• traffic overages after included transfer
• public IPv4 and additional IP conditions
• region-specific feature availability

If you come from a traditional VPS host, Hetzner feels transparent. If you come from hyperscalers, it feels refreshingly simple. If you need detailed cost allocation, policy controls, and org-level finance workflows, it can feel too simple.

Account approval and abuse controls

Hetzner has a reputation for being stricter than some budget VPS vendors on account verification and abuse handling. That is not a bug. It is part of keeping the platform usable. New accounts should avoid suspicious payment setups, throwaway contact details, or aggressive immediate provisioning patterns.

Common mistakes

• Choosing Falkenstein by default when Helsinki would cut Nordic latency materially.
• Assuming Hetzner includes the same managed-service depth as DigitalOcean or AWS.
• Building around a load balancer or managed feature before checking regional availability.
• Treating cheap VM pricing as a reason to skip backups, off-site copies, and restore tests.
• Comparing list prices without matching vCPU type, RAM class, and included transfer.
• Expecting support to act like a managed sysadmin team.

So, is Hetzner still the best cheap EU cloud?

For self-managed virtual machines in Europe, yes, usually. As of April 2026, Hetzner still offers one of the strongest price-to-performance ratios in the market for Linux VMs, especially if you can use Helsinki or are happy with Germany-based deployment. Our CPX21 in Falkenstein was quick to provision, fast enough for real workloads, and much cheaper than common DigitalOcean or Vultr equivalents.

That does not make Hetzner the best cloud overall. It makes Hetzner the best cheap EU VM cloud for a specific type of buyer: one that can operate Linux comfortably, accept a thinner managed-service layer, and design around regional product gaps. If that is you, Hetzner remains easy to recommend. If you want convenience, richer managed services, or Sweden-first regional placement, the better answer may be Vultr’s Stockholm region, DigitalOcean, or Cloudways.

What to read next

• If you are comparing cheaper VM providers, see our guide to EU VPS hosting.
• If you care about regional latency, see our guide to Nordic data-centre hosting.
• If you want less server admin, see our guide to managed cloud hosting for agencies.

Affiliate disclosure: NorseHost may earn a commission if you buy through some recommendation links, including Vultr, Cloudways managed hosting, or Bunny.net. That does not change our editorial methodology. We test the service, cite primary sources, and say when a provider is limited.