Hook
Hetzner is worth evaluating when your first question is not "who is cheapest?" but "where does the CPU line actually move?" The useful operational insight is that Hetzner splits its cloud range into low-cost CX, CPX, and CAX instances for general-purpose use, higher-cost CCX dedicated-vCPU plans for steadier CPU behavior, and then separate dedicated servers managed through Hetzner Robot. That matters because buyers can stay inside one provider while changing from bargain shared compute to more predictable isolation instead of discovering the limits only after sustained load.
Positioning
Hetzner Online GmbH, founded in Germany, operates its own hardware footprint across six listed locations: Germany (Nuremberg), Germany (Falkenstein), Finland (Helsinki), USA (Ashburn), USA (Hillsboro), and Singapore. In our local data, Hetzner currently has 69 plans with a minimum visible monthly price of EUR 3.49, CPU families ranging from Intel/AMD cloud instances to AMD EPYC 9454P, Intel Xeon Gold, and newer Ryzen dedicated servers, and NVMe SSD storage across the plan set. The support model is standard with ticket, email, and phone channels plus a stated 24h response target. Billing is hourly and monthly, payment methods include credit card, PayPal, bank transfer, and SEPA, and the reliability posture includes a 99.9% uptime guarantee, announced maintenance windows, and published SLA credits. Platform capability is stronger than many budget hosts: API access, custom ISO support, rescue tooling, private networking, and the Hetzner Robot API for dedicated-server operations are all material for self-managed teams. The trade-off is equally clear. There are no managed services in the local provider metadata, and Hetzner is not positioned as a hand-holding provider.
Pricing Transparency
The visible Hetzner entry point in the local plan data is EUR 3.49 per month for a cloud plan, and the provider metadata confirms hourly plus monthly billing rather than long annual commitments. That makes Hetzner easier to test than many budget hosts, but it does not prove long-term price stability. Renewal pricing was not verifiable from local data, so the safe reading is this: treat the current listed monthly rate as the present price, not as a guaranteed forever-renewal promise. The useful comparison is not only the entry number. Check whether your workload actually belongs on a shared CX-class instance, a dedicated-vCPU CCX plan, or a Robot-managed dedicated server, because that product-line jump changes the real monthly cost much faster than the teaser price suggests.