n8n hosting in the Netherlands: self-hosting, managed or cloud โ what fits?
n8n has grown in two years from a niche project into the de-facto open-source alternative to Zapier and Make. Its appeal is clear: 400+ integrations, drag-and-drop editor, native AI nodes, and no need to spend hundreds of euros per month on per-task credits. But as always with open source: free isn't free.
At Virtual Computing we regularly get the question this year: "Can you host n8n for us?" And underneath that sits another: "How does that compare to n8n.io's Cloud version, and isn't self-hosting much cheaper?". The honest summary: yes and no. Below are the three options and which we recommend for which team type.
The three ways to run n8n
1. n8n Cloud (SaaS from n8n.io itself)
The company behind n8n sells a hosted variant. Plug-and-play, no server management. Starting at โฌ20/month for the Starter plan with execution limits. The Pro plan around โฌ50/month unlocks more integrations and serious volumes.
Pro: zero management, open a browser and start building.
Con: your workflows and all data passing through them run on n8n GmbH infrastructure in Germany. GDPR-fine, but for sectors with sensitive data (healthcare, finance, legal) you don't want your customer-API credentials in a third party's cloud. And the price scales fast: a team running 50,000 executions per month is over โฌ100/mo.
2. Self-hosting on your own VPS
n8n is open source under the Sustainable Use License. You pull the Docker image, run it on a Hetzner or TransIP VPS for โฌ10-15/month, and you're live. The n8n community has tutorials and even a self-hosted starter kit on GitHub that you can finish in an evening.
Pro: one of the cheapest options on infrastructure cost, and all data stays with you.
Con: you are now the DevOps team. That means:
- OS patches every month (Ubuntu LTS security updates, Docker engine, kernel)
- n8n updates yourself (n8n releases every 1-2 weeks, with regular breaking changes in major versions)
- Backups of both the PostgreSQL database AND the credentials encryption key โ lose the latter and your entire credential store is unusable
- SSL certificate management (Let's Encrypt auto-renewal usually works, until one time it doesn't and your workflows go offline)
- Reverse proxy (Traefik or Nginx) configuration and updates
- Monitoring โ without uptime checks you'll only learn n8n is down when workflows start failing
- Scaling moments โ if your workflow volume grows the database and worker config must scale with it
In practice we see that for an SMB team the "n8n VPS" consumes 2-4 hours per month of someone's time. At an average senior IT hourly rate that's โฌ200-400 per month, on top of the โฌ15 for the VPS itself.
And we haven't even touched security: n8n holds all your customer-system API keys, so a compromised n8n is a major leak. If you don't have password rotation, IP allow-listing and log monitoring in place, you're an attractive target.
3. Managed n8n hosting (containerized hosting with a Dutch partner)
The middle road: someone else runs the Docker container and the surrounding management, but your data stays on Dutch infrastructure and you keep full control of workflow content. At Virtual Computing we offer this from โฌ29/month via our managed Docker hosting for apps: own container, own subdomain, daily backup, SSL and monitoring included.
Pro: you get the benefits of self-hosting (data local, no execution limits, all integrations open) without the operational load.
Con: slightly more expensive than a bare VPS. But once you add the real hours, it's usually cheaper than self-hosting.
Honest cost comparison for an SMB team
Take a team that uses n8n seriously: 5 makers, 50 workflows, 30,000 executions/month, sensitive API keys (CRM, email, invoicing). 12-month view.
| Option | Monthly cost | Internal time | Total TCO |
|---|---|---|---|
| n8n Cloud Pro | โฌ50 | 0 hours | โฌ600 / year |
| Self-hosted on Hetzner VPS | โฌ15 (server) | 3 hours ร โฌ75 | โฌ2,880 / year |
| Managed at Virtual Computing | โฌ29 | 0 hours | โฌ348 / year |
Managed wins on both TCO and hours. And that's exactly why we offer it โ not because we're against self-hosting, but because in most real cases the operational hours are underestimated.
Also read our article on setting up your own AI infrastructure if you want not just n8n but a complete AI stack โ n8n is one of the five layers there.
When self-hosting IS smart
Self-hosting on your own VPS is a fine choice when:
- You have a team with full-time DevOps capacity (the hours are already budgeted)
- Your workflow volume is so large (>500,000 executions/month) that dedicated tuning pays off
- Your entire infrastructure stack (Kubernetes, own monitoring) is already in place
- You want to deeply customise n8n with community nodes and custom code
In all other cases โ so for the bulk of SMB teams โ managed is the more honest choice.
Hardware sizing: what does n8n need?
A common question: how much should I provision? Rules of thumb from our own environments:
- Small team, low frequency (5 workflows, a few hundred executions/day): 1 vCPU, 1 GB RAM, 10 GB storage. Fits on an entry-level container.
- SMB standard (50 workflows, 30k executions/mo): 2 vCPU, 2-4 GB RAM, 20-30 GB storage. Our standard โฌ29 container covers this.
- Scaling phase (100+ workflows, 100k+ executions, parallel queues): 4 vCPU, 8 GB RAM, dedicated PostgreSQL and Redis. Multiple worker containers.
The PostgreSQL database grows with your execution history. A retention policy (purge old executions after 30 days) keeps it manageable.
What's in our managed n8n?
Concretely what you get when you take n8n from Virtual Computing:
- Own Docker container on our Dutch infrastructure (two ISO 27001 data centres, GDPR-safe)
- Own subdomain by default, or attach your own domain with SSL
- Daily backup of database AND credentials encryption key, 14-day retention
- Automatic updates with your consent (we test the release first, then roll to you)
- 24/7 monitoring with automatic incident response during business hours
- PostgreSQL as database (no SQLite โ that's only fit for demos)
- One invoice per month and monthly cancellation
If you later want to move to self-hosting or another provider we deliver the full compose file + database dump. No lock-in.
Common questions
Can I connect n8n to our on-premise systems? Yes, via a VPN tunnel or an outbound webhook architecture. We help set it up.
Does n8n work with our Microsoft 365 tenant? Yes, n8n has native Microsoft Graph, SharePoint, Outlook and Teams nodes. OAuth authentication as you'd expect.
Can we extend the AI stack with LLM tools? Definitely โ see our AI infrastructure page for the full stack (n8n + LangFlow + Flowise + Dify + OpenWebUI). Teams often start with n8n and grow from there.
What about disaster recovery? We run backups to a second data centre within the Netherlands. Restore on a new container averages 1-2 hours of work. Also read our article on a solid disaster recovery plan in 5 steps.
Conclusion
Running n8n yourself on a VPS is technically simple, but operationally an underestimated burden for most SMB teams. n8n Cloud is plug-and-play but expensive at volume and carries the "data with third party" question. Managed hosting on Dutch infra is the more honest choice for most teams: open-source flexibility, NL data, fixed monthly price, no DevOps hours.
Want n8n containerized on our infrastructure? See our app catalogue or get in touch for a no-strings call โ call 085-013 4500.
Written by
Related articles
Work more efficiently with Microsoft Copilot
Microsoft Copilot makes your daily work routine more efficient. This smart AI assistant from Microsoft helps you with tasks in Word, Excel, Teams, and more.
AIHow AI transforms online working
The modern workplace feels logical today. Working wherever suits you. Collaborating in the cloud. Always having access to information. But anyone who thinks this happened by itself underestimates how much has been tried along the way.
AICopilot promo extended to June 2026: buy or wait?
Microsoft has extended the Copilot Business promo through 30 June 2026 โ buy now and lock in current pricing. We help customers decide when that's the right call, and when waiting or a different AI route is smarter.