Managed vs unmanaged dedicated servers: the true monthly cost

The number on the page

A 48-core EPYC server with 256 GB of memory and two 3.84 TB NVMe drives lists for about €842 per month at a wholesale European provider in October 2026, plus a setup fee of roughly half a month. That buys hardware, power, a 1 Gbit port and a rescue system. It does not buy anyone to run it.

Whether that is a bargain depends entirely on what happens next. If an operations engineer with spare capacity picks it up, it is. If a developer picks it up between features, the cost moves off the hosting invoice and onto the payroll, where nobody tracks it.

What production actually requires

Any server that matters needs the same six things, regardless of who provides them. The table shows realistic monthly figures for one host from mainstream tools in 2026.

Add the rows and a single wholesale server costs between about €3,900 and €6,200 per month in practice. The hardware is 14 to 22 percent of it.

Where the hidden costs hide

• The restore test nobody ran. Backups that have never been restored are a hypothesis. The first restore under pressure takes a day and sometimes fails.
• The patch that waited. A critical CVE on a Friday afternoon is patched on Monday, or the next sprint, or after the incident.
• Context switching. A developer who carries the pager ships fewer features, and the cost of that never appears on an infrastructure line.
• Tool sprawl. Each SaaS tool has its own login, invoice, retention policy and alert fatigue.

When unmanaged is the right answer

Unmanaged servers make sense when infrastructure is your product, when you already employ operations engineers with capacity, or when you run dozens of identical machines so the tooling and process cost amortises. Development and batch workloads where an hour of downtime costs nothing also qualify.

When managed is the right answer

Managed servers make sense when the people who would run the server have a different job, when downtime has a price, or when you want a single number in the budget and a single party to call. For most product teams under about twenty servers, the managed price is lower than the true cost of unmanaged once engineer time is counted honestly.

Questions to ask any managed provider

1. What is the critical response time, in minutes, around the clock, and is it in the contract?
2. When was the last restore test on a customer system, and can I see the report?
3. What is the patch SLA for critical vulnerabilities?
4. Do I keep root access?
5. What happens to my data when I leave, and how long do I have to export it?