How to rent a server: VPS, dedicated or GPU, and what it costs

By the serverserv team · 8 min read · Updated 21 October 2026

Renting a server means paying a monthly fee for a computer in a data centre that you control remotely. For most websites, apps, databases and AI agents a rented VPS at about $5 to $50 a month is enough. A dedicated server costs about $40 to $600 a month, and GPU servers for AI training are usually rented by the hour. This guide explains the options, the real prices and how to choose.

What renting a server means

When you rent a server you do not buy hardware. A hosting company keeps the machine powered, cooled and connected in its data centre, and you get remote access to it, usually over SSH for Linux or Remote Desktop for Windows. You pay monthly or for a fixed term, and you can cancel or upgrade without selling anything.

• VPS (virtual private server): a slice of a powerful physical server with its own guaranteed CPU, RAM and disk. It behaves like your own machine, with full root access. This is what most people mean by "rent a server".
• Dedicated server: a whole physical machine for one customer. More power and no neighbours, but a higher price and slower setup.
• GPU server: a server with one or more graphics cards for training or running large AI models. Usually rented by the hour.
• Managed server: a VPS or dedicated server where the host also handles updates, security and monitoring for you.

How much it costs to rent a server

Type	Typical price	Good for
Entry VPS (1 to 2 vCPU, 2 to 4 GB RAM)	about $5 to $15 a month	Small sites, bots, AI agents, test servers
Mid VPS (4 vCPU, 8 to 16 GB RAM)	about $15 to $50 a month	Busy sites, apps, databases, n8n
Managed VPS	about $30 to $100 a month	Teams without a sysadmin
Entry dedicated server (4 to 8 cores, 32 to 64 GB RAM)	about $40 to $100 a month in Europe, $100 to $150 in the US	Heavy databases, many sites, steady high load
Mid dedicated server (12 to 32 cores, 64 to 256 GB RAM)	about $100 to $600 a month	Large workloads, virtualisation
GPU server (one RTX 4090 to one H100)	about $0.34 to $6.88 an hour	Training and serving large AI models

Ranges are on-demand list prices in USD from several providers in September and October 2026. Dedicated and GPU prices vary a lot by region and stock, so check the provider's own page before you decide.

serverserv prices for comparison: a KVM VPS starts at $8.99 a month on monthly billing, or $6.29 a month on a 3-year term, and every plan renews at the price you paid. See all sizes on the Linux KVM VPS page or the pricing page.

Which type of server do you need?

• A website or online store: start with web hosting. Move to a VPS when you need custom software or steady high traffic.
• An app, API or database: a VPS with 4 to 8 GB RAM covers most small and medium projects. Our guide to how much CPU and RAM you need helps you size it.
• An AI agent such as OpenClaw, or n8n automations: a 2 vCPU, 4 GB VPS is usually enough because the AI model runs at your AI provider. OpenClaw Hosting and n8n Hosting come set up.
• Windows software, trading platforms or remote desktops: a Windows VPS with Remote Desktop.
• You do not want to run the server yourself: a managed VPS.
• Training or self-hosting large AI models: a GPU server, rented by the hour for the time you need it.

Renting a server for AI and machine learning

Most AI projects need less hardware than people expect. An AI agent, chatbot or automation that calls OpenAI, Anthropic or another provider does the heavy thinking on that provider's GPUs, so your own server only runs the agent itself. A normal VPS without a GPU handles that well.

You need a GPU server only when you train models or run large open-source models yourself. Typical on-demand prices per GPU in September 2026 were about $0.34 to $0.74 an hour for an RTX 4090, $1.39 to $3.67 for an A100 and $2.89 to $6.88 for an H100. Running one H100 all month works out at roughly $1,500 to $2,850, so most teams rent GPUs only for the hours they train.

A common setup is a small VPS that runs the agent, database and website all the time, plus GPU time rented only when needed. Read more in what is AI agent hosting.

What to check before you rent

1. The renewal price, not only the first-term price. Many hosts double or triple the price after year one.
2. Root or administrator access, so you can install what you need.
3. Guaranteed resources: dedicated vCPU and RAM, and NVMe storage for fast disks.
4. Location: pick a data centre close to your users for lower latency.
5. Bandwidth: how many TB a month are included, and what happens if you go over.
6. Backups: included, paid, or your job. Keep a copy with a different company.
7. Contract: monthly with no lock-in, or a discount for a longer term.
8. Support: who answers, how fast, and whether they help with the operating system.

How to rent a server from serverserv

1. Choose a plan on the VPS, Windows RDP or Managed VPS page.
2. Pick monthly billing or a 1 to 3-year term. Longer terms cost less per month, and the price is locked in for the term.
3. Choose the operating system and pay by card, Apple Pay, Google Pay or PayPal.
4. You receive the login details by email and can connect over SSH or Remote Desktop.

Questions

Sources

• Cherry Servers: dedicated server prices in 2026
• Hostinger: cloud GPU pricing compared (September 2026)