n8n server size calculator: how big a server do you need?
By the serverserv team · 4 min read · Updated
Most self-hosted n8n setups need far less server than people expect, until files, AI steps or high volume arrive. Answer a few questions below and the checker suggests a server size, with the reasons behind it.
The size checker
All workflows together, at the busiest time of day.
Workflows switched on, with triggers or schedules.
A rule of thumb based on n8n's own guidance that memory matters more than CPU, and on the sizes in our n8n system requirements guide. Prices are serverserv's monthly prices. You can move up a plan later without reinstalling.
What decides your n8n server size
n8n itself is light. An idle instance uses around 100 MB of memory, and n8n's own documentation puts the application at 320 MB to 2 GB of RAM. What decides the size of the server is the work: how many workflows run at the same time, how much data each run carries, and what else lives on the same machine.
RAM runs out long before CPU does. Most n8n steps wait on an API, a database or an AI model, which costs little CPU. But n8n keeps the data passing through a workflow in memory while it runs. Ten webhooks arriving together means ten executions in memory at once, and one workflow that downloads large PDFs can use more memory than fifty small ones.
Sizes by workload
| Workload | Typical server | serverserv plan |
|---|---|---|
| A handful to a few dozen workflows, webhooks, some AI steps | 2 vCPU, 4 GB RAM | n8n Starter |
| Hundreds of runs an hour, files, PostgreSQL on the same server | 4 vCPU, 8 GB RAM | n8n Personal |
| Thousands of runs an hour, heavy AI or Code steps on big data | 6 vCPU, 12 GB RAM | n8n Pro |
| Queue mode with several workers, very high volume | 12 vCPU, 48 GB RAM | n8n Team |
| Many parallel workers, local AI models alongside n8n | 18 vCPU, 96 GB RAM | n8n Enterprise |
Signs you need a bigger server
- Memory sits above 80% most of the day, or the server starts using swap.
- n8n restarts in the middle of workflows, or the logs show "JavaScript heap out of memory".
- Webhook replies slow down when several calls arrive together.
- You add PostgreSQL, a local AI model or another app to the same server.
Before you move up, try the cheap fixes: store files on disk with N8N_DEFAULT_BINARY_DATA_MODE=filesystem, process big lists in batches with the Loop Over Items node, and stop saving execution data for successful runs of your busiest workflows. Our system requirements guide walks through each one.
Questions
How much RAM does n8n need?
About 100 MB idle and up to 2 GB for the app under load. On a VPS, which also runs the operating system and a web proxy, 4 GB of RAM suits most small teams, and 8 GB or more suits files, heavy AI steps or PostgreSQL on the same server.
Is 2 GB of RAM enough for n8n?
For trying n8n and a few light workflows, yes. For anything in production, 4 GB gives you room for traffic spikes and avoids out-of-memory restarts.
Does n8n need a lot of CPU?
No. Most n8n work waits on other services, so 2 vCPU covers most teams. Add cores when you run queue mode with several workers or a local AI model.
When should I use queue mode?
When one n8n process cannot keep up, usually at thousands of runs an hour. Queue mode splits work across workers using Redis and PostgreSQL, so it needs a larger server or several servers.
Can I upgrade my serverserv n8n plan later?
Yes. You can move up a plan as your workflows grow without reinstalling n8n.