VPS, cloud instance or dedicated server: a decision guide
The three architectures differ in who owns the processor core, how they are billed and how quickly a system is provisioned. This article categorises them based on the questions that matter before placing an order.
Managed by the Platform department
Contents of the post+
At a glance
With a VPS, multiple instances share the cores of a host; with a dedicated cloud instance, a physical core belongs to your instance; with a dedicated server, you get the entire machine.
VPS and cloud instances are billed per hour and capped at the monthly price; a dedicated server is billed monthly.
Constant latency, a dedicated memory bus or licences tied to physical cores speak for a dedicated server; fluctuating loads and short terms for a VPS and cloud instance.
Who owns the processor core
A VPS is a guest system with a fixed share of vCPU, memory and local NVMe on a host shared by multiple customers – with full root access and its own IP address. In the shared families, the hypervisor schedules multiple instances on the same physical core; if you do not use your quota, you pass the share to your neighbours, but the price per vCPU is lower.
A cloud instance from the dedicated families runs on the same platform, but a physical core belongs exclusively to your instance.
A dedicated server is a physical machine that only you use: your own CPU, your own memory, your own drives, your own port – without a hypervisor and without neighbours on the same hardware.
A VPS plan includes root access via SSH, the KVM console in the customer area, one IPv4 and one IPv6 /64, basic DDoS protection, a private network and local NVMe, plus at least 20 TB of traffic per month. Every dedicated server includes basic DDoS protection, full IPMI and KVM access, basic hardware monitoring and basic support; traffic is unlimited on the included connection.
The boundary in one sentence
With a VPS, the core is shared; with a dedicated cloud instance, only the host is shared; with a dedicated server, nothing is shared.
Billing: hourly rate with a cap or monthly contract
VPS and cloud instances are billed the same way: tracked to the second, billed per hour and capped at the monthly price. The hourly rate is the monthly price divided by 730, the average month length, not the length of the specific calendar month.
If the sum of an instance's hours in a calendar month reaches the monthly price, the counting stops; even a month with 744 hours costs no more than the monthly price. There is no setup fee for any plan, and a commitment of 12 or 24 months is optional: it lowers the price by 5% and 8% respectively, and the term is paid in advance.
A test run on an hourly basis
Monthly price NP3, net
€35.99$41.75
Hourly rate, monthly price divided by 730
€0.0493$0.0572
Duration of the test
48 hours
Amount for the test, net
€2.37$2.75
A stopped instance continues to be billed because it occupies CPU quota, memory and local NVMe on the host. Only deleting it stops the billing; if you want to keep the state, create a snapshot first.
Beyond the included volume, VPS and instances with dedicated vCPU charge outgoing traffic to the public internet at €1.09$1.26 net per terabyte, without zone pricing; traffic between instances via the private network does not count. With a dedicated server, traffic is unlimited on the included connection.
A dedicated server is billed monthly. With a monthly term, the model's one-off setup fee is added, depending on the model €65.54$76.03 to €2,299.99$2,667.99 net per server; with a 12 or 24 month term, it is waived, and the monthly amount drops by the same term discount as in the cloud.
The entry-level server with a twelve month term
Monthly amount CORE-R7 without term, net
€65.54$76.03
Setup fee for a monthly term
€65.54$76.03
Setup fee for a 12 month term
€0.00$0.00
Monthly amount after 5% term discount
€62.26$72.23
Advance payment with the first invoice, net
€747.12$866.76
Cancellation also follows billing. You can delete an instance without a term commitment in the console without notice, and billing ends at the time of deletion. You can cancel a dedicated server with a one-month term with one day's notice to the end of the term, contracts for 12 or 24 months with 30 calendar days – otherwise they renew for twelve months.
All monthly and hourly amounts in this article are net catalogue prices for Frankfurt am Main. The same location factor applies to Düsseldorf. In Helsinki, the location factor is lower. The setup fee for a dedicated server is the same at every location.
Performance consistency and scaling
Wait time caused by neighbours is visible in the guest system as steal time, the time when the core currently belongs to a neighbour. On the shared families, it averages less than 3% per month, and up to 12% during peak loads on individual hosts; on the dedicated families, it is practically zero because no neighbour is computing on the core.
For the choice between shared and dedicated, a rule of thumb from operations applies: if the average utilisation of a vCPU is consistently above 20% over a week, it is worth comparing with the dedicated series, whose performance per vCPU remains the same even under load. Both variants are compared in the shared or dedicated vCPU section of the Public Cloud.
A dedicated server is worth it as soon as latency needs to be constant, a database needs the memory bus to itself, or an application's licence terms are tied to physical cores.
A VPS can scale up at any time: stop the instance, change the plan, start it, and the file system grows online. Scaling down is only possible within the same series and only if the occupied file system fits into the smaller volume. Beyond 16 vCPU and 32 GB of memory, you move on to a cloud instance or your own server.
A cloud instance can be resized later, and the hourly rate of the new size applies from the timestamp of the resize; changing the family requires a snapshot and a new instance. With a dedicated server, more computing power means a different model: you run both servers in parallel, switch over and cancel the previous one at the end of the term.
Until the system is ready
A VPS is ready to use in 35 seconds on average, guaranteed in under 60 seconds, measured up to the first SSH login.
A cloud instance is ready in 35 seconds on average, guaranteed in under 60 seconds. After payment, it transitions via BUILD and BOOT to the ACTIVE state; every transition is recorded with a timestamp in the project's activity log.
A dedicated server is provisioned at the selected location upon payment approval. If stock is low, handover depends on incoming supply; stock and lead time per model are shown in the server configurator, and access credentials and IPMI access are provided via ticket, without a callback or appointment.
The decision at a glance
The table compares the three deployment types side by side; the entry level is always the cheapest model of the deployment type in the catalogue.
VPS, cloud instance and dedicated server compared
Feature
VPS
Cloud instance with dedicated vCPU
Dedicated servers
Processor core
shared with neighbours
permanently assigned to your instance
entire machine, without hypervisor
Steal time
under 3% on monthly average
practically zero
not applicable, no hypervisor
Billing
per hour, capped at the monthly price
per hour, capped at the monthly price
monthly
Setup fee
none
none
only with a monthly term, €65.54$76.03 to €2,299.99$2,667.99
Entry level per month, net
NX2 with 2 vCPU: €5.99$6.95
ND2 with 2 vCPU: €38.80$45.01
CORE-R7 with 8 cores: €65.54$76.03
Delivery
35 seconds on average
35 seconds on average, every transition in the log
based on stock, lead time per model in the configurator
Typical use
Frontends, test environments, jump servers, small services
Production with continuous load
constant latency, databases with dedicated memory bus, licences on physical cores
If you are unsure, measure first: a week on an instance with shared cores will show the average load and steal time in the guest system. If the load remains above the rule of thumb threshold or the steal time increases significantly during peak loads, the dedicated series is the next step. From four vCPU, it costs more per month than the entry-level server CORE-R7, which has eight physical cores all to itself.
Within the deployment type, the load profile then decides:
A VPS is worthwhile where the load fluctuates and short delays go unnoticed; a test run over a weekend does not cost a full month.
Alongside the dedicated family, the Public Cloud offers memory-optimised instances with 8 GB memory per vCPU for in-memory databases, caches and analytics, as well as storage-optimised instances with large local NVMe directly attached to the instance.
The catalogue organises the dedicated servers into series, each with a specific focus: CORE for single-core clock speed per euro, FORGE as a scalable server platform with guaranteed uplink, TITAN for core density and VAULT for storage density.
Two limitations should be noted before ordering: automatic backups are not included with a VPS, and the local NVMe of a cloud instance survives reboots, but not the deletion of the instance. Data that must outlive an instance belongs on Block Storage or Object Storage.
The entry level does not have to be a final decision: VPS and dedicated servers are on the same private network, and if you start with a VPS, you can later migrate to your own hardware without changing your contract. Every model has a private network except CORE-R7. The entry-level server from the calculation example is therefore not included.
Published on 8 November 2026. Prices and product values in this post are taken from the price list and reflect their current status.
An AI builder builds a complete website from just a few sentences – how good it turns out depends on the description and the subsequent testing. Process, costs and limits using ENTRONYX STUDIO as an example.
How much data can an outage cost, and how long can it last? If you only answer both questions in an emergency, you will find that the technology has given a different answer.
Whether a language model fits on a graphics card can be calculated before ordering. The requirement consists of three items, only one of which is fixed with the model.
Platform department
Questions for the department
Technical questions about a post are answered by the person responsible for the department — in a conversation, not a presentation. For quotes and framework agreements, sales will lead the initial conversation.