n8n Hosting in 2026: What a Server Really Costs vs n8n Cloud, and the Size You Need
Where to host n8n and what it costs as of November 2026: n8n Cloud from EUR 20 a month for 2,500 executions, or a self-hosted server from $12 on DigitalOcean, $24 if you want the 4 GB, 2-vCPU size n8n's guides ask for with its AI code sandbox. The schedule setting that burns through executions, and when self-hosting pays.
Where should you host n8n, and what does it cost? The short answer
As of November 2026, you have two real choices. n8n Cloud, where n8n runs it for you, starts at €20 a month billed annually for 2,500 workflow executions. Self-hosting the free Community Edition costs whatever the server costs: on DigitalOcean, $12 a month for a 2 GB machine that covers n8n's own minimum, or $24 a month for the 4 GB, 2-vCPU machine n8n's install guides now ask for if you want its AI code sandbox. The server is the cheap part. The hour a month of updates and backups is what decides it.
Server sizes come from n8n's own hosting docs, server prices from DigitalOcean's droplet pricing page and Cloud prices from n8n's pricing page, all read November 12, 2026.
How big a server does n8n need?
Smaller than most people buy, unless you want the AI features. n8n's prerequisites page lists:
- Memory: 320 MB to 2 GB. n8n says memory matters more than CPU, and that n8n itself is not CPU intensive.
- Database: 512 MB to 4 GB of SSD, on SQLite or PostgreSQL. The Docker Compose guide calls SQLite fine for trying things out and PostgreSQL the choice for production.
- The warning: memory use depends on your workflows and data, not on n8n. Workflows that move large files can run a small server out of memory.
The part that catches people: n8n's current Docker Compose and Hetzner install guides both ask for at least 4 GB of RAM and 2 vCPUs, because the sandbox that runs AI-generated code uses Docker inside Docker. The Hetzner guide notes that the smaller CPX11 server is enough for most usage without that feature. So the AI sandbox roughly doubles the server you need.
Free playbook
What to Automate First
The SMB playbook for picking your first (and next) AI automation — the same prioritization we use with clients. PDF, emailed instantly.
Also subscribes you to The Service Business AI Brief. No spam, unsubscribe anytime.
What each hosting setup costs a month
| Setup | Monthly price | What you get | What you take on |
|---|---|---|---|
| n8n Cloud Starter | €20, billed annually | 2,500 executions, 5 running at once, hosted and updated by n8n | The execution cap; past it you move up a plan |
| Self-hosted, small (DigitalOcean 1 vCPU / 2 GiB) | $12 | 50 GiB SSD, 2,000 GiB transfer, unlimited executions | Updates, backups, security, uptime |
| Self-hosted, room to grow (DigitalOcean 2 vCPU / 2 GiB) | $18 | 60 GiB SSD, 3,000 GiB transfer | Same |
| Self-hosted with the AI code sandbox (DigitalOcean 2 vCPU / 4 GiB) | $24 | 80 GiB SSD, 4,000 GiB transfer, meets the 4 GB / 2 vCPU guide | Same, plus Docker-in-Docker to keep patched |
DigitalOcean bills per second with a 60-second minimum and caps each plan at its monthly price, so these are ceilings. Run PostgreSQL in the same Docker Compose stack on the same server and it costs nothing extra; a separate managed database is a second bill. Hetzner, the other host n8n writes a guide for, prices in euros before VAT and changed its cloud prices on June 15, 2026, so check its current card rather than an old blog's number. The plan-by-plan Cloud ladder above Starter is on our n8n pricing page.
The workflow that blows through 2,500 executions
n8n counts an execution each time a workflow runs start to finish, however many steps it has. That sounds generous until you add a schedule. A workflow with a Schedule Trigger every 5 minutes runs 12 times an hour, 288 times a day, about 8,640 times a month, whether or not it finds anything to do. That one workflow is more than three times Cloud Starter's allowance.
Two ways out. Trigger on events instead of the clock (a webhook from your form, CRM or phone system starts the workflow only when something happens), or check every 30 minutes instead of every 5, which cuts the same workflow to about 1,440 runs. If neither works, this is the case where self-hosting wins outright: the Community Edition doesn't meter executions, so the 8,640 runs cost the same $12 as 80. How that compares with Make's and Zapier's metering is in our n8n vs Make vs Zapier breakdown.
A worked choice: a 6-person service business
Say you run three workflows: new web leads into the CRM with a text back (about 300 runs a month), missed calls into a follow-up task (about 120), and a nightly report (30). That's about 450 executions a month.
- n8n Cloud Starter: €20 a month, uses under a fifth of the allowance, nothing to maintain.
- Self-hosted on a $12 server: $12 a month plus your time. If updates, a backup check and the occasional restart take one hour a month, and your hour is worth $50, self-hosting costs about $62.
Cloud wins at this size. Self-hosting starts to pay when one of three things is true: you run schedule-heavy workflows that would push you up Cloud's ladder, you have someone on staff who already runs servers, or your data has to stay on infrastructure you control.
What self-hosting actually asks of you
- Updates. n8n ships often, and each update is yours to apply and test against your workflows.
- Backups. Your workflows, credentials and execution history live in the database. No backup, no recovery after a bad update or a dead disk.
- Security. The instance holds the keys to your CRM, email and payment tools. It needs HTTPS, a locked-down login and a server that gets patched.
- Uptime. If the server stops at 2am, every webhook sent to it until morning is lost unless the sender retries.
If that list is the reason you're reconsidering n8n, the hosted options are compared in our n8n alternatives guide.
Who can host and build n8n workflows for my business?
A freelancer can stand up a server in an afternoon; the longer work is the workflows and keeping them running. Our guide to who can build Zapier, Make and n8n workflows compares freelancers, agencies and studios. SuperDupr, the AI ops shop Justin McKelvey runs, is one of the studios: our published price for AI workflow automation is $10,000 to $25,000 for a connected build, with $200 to $2,000 a month ongoing depending on volume, and most builds go live in 2 to 4 weeks. We pick Cloud or self-hosted by what it costs at your volume, and the accounts stay in your business's name. The first step is a free 30-minute strategy session, booked from that page.
Sources
- n8n docs, hosting prerequisites (docs.n8n.io), read November 12, 2026: memory 320 MB to 2 GB, database 512 MB to 4 GB SSD on SQLite or PostgreSQL, memory matters more than CPU, large binary files can exhaust resources.
- n8n docs, "Install using Docker Compose" and "Deploy to Hetzner," read November 12, 2026: SQLite for trying out, PostgreSQL for production; at least 4 GB RAM and 2 vCPUs for the AI code sandbox (Docker-in-Docker); CPX11 enough for most usage without it.
- n8n pricing, n8n.io/pricing, read November 12, 2026: Starter €20 a month billed annually, 2,500 workflow executions, 5 concurrent executions, hosted by n8n; executions counted per full workflow run; Community Edition free to self-host.
- DigitalOcean droplet pricing, digitalocean.com/pricing/droplets, read November 12, 2026: Basic 1 vCPU / 2 GiB $12, 2 vCPU / 2 GiB $18, 2 vCPU / 4 GiB $24 a month; per-second billing, 60-second minimum, monthly cap.
- Hetzner press notice on server price adjustments effective June 15, 2026, read November 12, 2026 (no current cloud price quoted here).
- Execution counts, the 6-person example and the $50 hour: SuperDupr's arithmetic and stated assumptions.
Frequently Asked Questions
-
As of November 2026, n8n Cloud Starter is EUR 20 a month billed annually for 2,500 workflow executions. Self-hosting the free Community Edition costs the server: DigitalOcean lists a 1 vCPU, 2 GiB Basic droplet at $12 a month and a 2 vCPU, 4 GiB droplet at $24, plus the time to keep it updated and backed up.
-
n8n's prerequisites page lists 320 MB to 2 GB of memory, and says memory matters more than CPU. Its current Docker Compose and Hetzner install guides ask for at least 4 GB of RAM and 2 vCPUs if you want the sandbox that runs AI-generated code, because it uses Docker inside Docker.
-
PostgreSQL for anything in production. n8n's Docker Compose guide calls SQLite fine for trying n8n out and PostgreSQL the choice for production. Running Postgres in the same Docker Compose stack on the same server adds no hosting cost.
-
Not for a small team with a few workflows. At 450 executions a month, Cloud Starter at EUR 20 beats a $12 server once you count an hour a month of upkeep. Self-hosting pays when schedule-heavy workflows would push you up Cloud's plans, when someone on staff already runs servers, or when your data has to stay on your own infrastructure.
-
Usually a schedule. n8n counts one execution per full workflow run, so a Schedule Trigger every 5 minutes runs about 8,640 times a month even when it finds nothing to do. Triggering on webhooks instead, or checking every 30 minutes, cuts that to a fraction.
Written by
Justin McKelveyFormer CTO at an AI startup and former Head of Product, with 50+ products shipped. I ran snowboard and bike shops before I taught myself to code, so I build AI systems and websites with the owner in mind.