Servers and operations · July 18, 2026
VPS, VDS, or a dedicated server? Cheap infrastructure is a job
Hetzner, OVHcloud, Scaleway, and Netcup offer startling amounts of hardware for the price of a managed-platform add-on. The missing line item is the operator—and for a solopreneur, that operator is also the founder.
Method
What this comparison measures
- Compare current entry configurations without pretending advertised vCPU counts are performance benchmarks.
- Record VAT, setup fees, IPv4, included transfer, backups, snapshots, billing while powered off, and upgrade constraints.
- Model founder operations at $50/hour to expose the hidden labor term; readers can replace that rate.
- Use US and EU prices where sold, but keep European providers in their native ex-VAT currency when conversion would hide the source price.
VPS and VDS are labels, not performance guarantees
VPS means virtual private server. VDS usually means virtual dedicated server, implying more isolated or dedicated CPU resources. There is no universal industry boundary: one vendor’s VDS can resemble another vendor’s dedicated-vCPU VM. Hypervisor, CPU generation, allocation policy, storage latency, noisy-neighbor behavior, and network limits matter more than the acronym.
Hetzner makes the useful distinction explicit in its product families: shared cloud plans and dedicated-vCPU CCX plans have very different prices. The June 2026 price adjustment moved a small CAX11 Arm cloud server to €5.99/month before IPv4 and VAT, while the smallest dedicated-vCPU CCX13 moved to €42.99.[1]
OVHcloud’s new VPS range includes isolated virtual resources, a daily backup, anti-DDoS, and unlimited transfer at a stated port speed. That bundle is materially different from a provider that sells backup and transfer separately.[4]
| Product | Useful interpretation | Still verify |
|---|---|---|
| Shared VPS / cloud VM | Low-cost VM with shared host CPU | Steal time, CPU generation, disk latency, sustained network |
| VDS / dedicated vCPU | CPU allocation is intended to be more isolated | Exact scheduling guarantee and overcommit policy |
| Bare metal | Physical server assigned to one customer | Disk redundancy, remote hands, replacement SLA, setup fee |
| Managed PaaS | Provider operates host and deployment layer | Runtime quotas, egress, database, backups, observability |
The advertised monthly number is only the first row
Hetzner bills cloud servers until deletion, even when powered off. Outbound traffic beyond the included allowance is billed, and traffic alerts do not cap consumption. OVHcloud includes unlimited VPS traffic and a daily backup in the displayed range, but taxes and regional offers still differ.[2][3][4]
Scaleway Instances stop compute billing after deletion, while flexible IP addresses and storage volumes remain separate resources. Scaleway also sells Elastic Metal and Dedibox, which makes it useful for a gradual VM-to-bare-metal path—but every attached resource needs its own teardown check.[5][6]
Cheap hardware moves reliability into your calendar
The minimum production checklist is longer than installing Docker: automatic security updates with a reboot policy, SSH and firewall hardening, off-host backups, restore tests, filesystem and certificate monitoring, service supervision, capacity alerts, log rotation, vulnerability response, and a documented replacement procedure.
A single VPS is also one failure domain. RAID inside one server is not an off-host backup. A provider snapshot is not proof that PostgreSQL can restore. A second server improves availability only after replication, routing, monitoring, and failover are tested.
| Line item | Failure if omitted | Cost control |
|---|---|---|
| Off-host backup | Deletion, corruption, ransomware | Retention and restore schedule |
| Monitoring | Disk fills or service dies silently | Small metrics/log quota with hard retention |
| Public IPv4 | Unexpected monthly add-on | Prefer IPv6/proxy where product allows |
| Outbound traffic | Media or abuse creates overage | CDN, rate limit, provider traffic alert |
| Patch/reboot labor | Known vulnerability remains exposed | Automate and budget a maintenance window |
| Recovery capacity | No place to restore during failure | Document spare-region/provider path |
When a server is the right answer
A VPS wins when the workload is steady, fits on one machine, needs ordinary Linux primitives, and can tolerate a documented recovery window. It is especially compelling for queues, background jobs, self-hosted tools, and databases whose managed premium exceeds the value of automated operations.
Dedicated hardware becomes rational when sustained CPU, memory, local NVMe, or predictable transfer dominates the bill. It is a poor reaction to one expensive serverless invoice if utilisation is still bursty or no one can replace the machine at 03:00.
- Choose managed first when customer support and product work consume the same founder hours.
- Choose VPS when one-node recovery is acceptable and infrastructure is reproducible from code.
- Choose dedicated vCPU after measurement proves noisy neighbors or sustained CPU are the constraint.
- Choose bare metal when utilisation is high enough to amortize both the server and its operator.
Source register
Specifications and prices change. The links make this snapshot auditable.
Sources and commercial facts were checked on 2026-07-18. Prices exclude tax unless the source says otherwise.
- Price adjustment effective June 15, 2026 ↗Hetzner · pricing
- Cloud billing FAQ ↗Hetzner · billing
- Included server traffic ↗Hetzner · billing
- VPS plans and included services ↗OVHcloud · pricing
- Understanding Instance pricing ↗Scaleway · billing
- Elastic Metal quickstart ↗Scaleway · product
- VPS product range ↗Netcup · pricing