Self-Hosted n8n vs n8n Cloud: Where the Break-Even Sits
The break-even between n8n Cloud and a self-hosted instance isn’t a number — it’s a curve, because Cloud prices per execution and a VPS prices per month.
Plot four points on that curve and the shape is unmistakable.
The four-point model
| Executions/month | n8n Cloud | Self-hosted | Difference |
|---|---|---|---|
| 1,000 | $21.60 (Starter) | $11.99 | $9.61 |
| 10,000 | $54.00 (Pro) | $11.99 | $42.01 |
| 50,000 | $864.36 | $14.99 | $849.37 |
| 100,000 | $1,584.36 | $29.98 | $1,554.38 |
The self-hosted column already includes the second server you need at 100,000 executions, which most comparisons omit. More on that below.
Why the gap explodes at 10,000
n8n Cloud’s tiers are €20/month for 2,500 executions (Starter), €50 for 10,000 (Pro), and €667 for 40,000 (Business), all on annual billing.
Look at that last step. From Pro to Business is 13× the price for 4× the executions.
Normalised per thousand executions:
| Plan | Per 1,000 executions |
|---|---|
| Starter | $8.64 |
| Pro | $5.40 |
| Business | $18.01 |
| Business overage | $14.40 |
Business is more than three times worse per execution than Pro, and worse than Starter. It isn’t a volume discount; it’s a different product with governance features attached, priced accordingly.
That cliff is the entire story of this comparison. Below 10,000 executions you’re on sanely-priced tiers. The moment you cross it, your options are a $720/month plan or a $15 server.
The ops time, honestly
The saving figures above mean nothing until you price your own time, which is the part self-hosting advocates skip.
Setting up n8n properly — Docker Compose, Postgres rather than SQLite, a reverse proxy with TLS, backups — takes two to four hours the first time. Keeping it running is perhaps 30 minutes a month: updates, checking executions haven’t filled the disk, occasional restarts.
So how much ops time does each volume justify?
| Executions/month | Monthly saving | Breaks even at $50/h | At $100/h |
|---|---|---|---|
| 1,000 | $9.61 | 0.2 hours | 0.1 hours |
| 10,000 | $42.01 | 0.8 hours | 0.4 hours |
| 50,000 | $849.37 | 17 hours | 8.5 hours |
| 100,000 | $1,554.38 | 31 hours | 15.5 hours |
Read those bottom rows. At 50,000 executions, self-hosting would have to consume seventeen hours of your month at $50/hour before Cloud became the better deal. Nobody spends seventeen hours a month maintaining an n8n instance.
At 1,000 executions the opposite holds: the saving is $9.61, which buys about twelve minutes of a $50/hour professional. You will spend more than twelve minutes a month on it. Cloud wins, clearly and unembarrassingly.
Fill in your own rate: divide the monthly saving by your hourly rate. If the result is more hours than you’d actually spend on maintenance, self-host. That’s the whole calculation, and it’s why the answer differs so much between a hobbyist and an agency.
The second server nobody counts
Here’s where most “self-hosting is cheaper” posts quietly cheat.
n8n’s default execution mode is regular — one process runs everything. That’s fine into the tens of thousands of executions. Push further, or run long workflows that block others, and you need queue mode: Redis plus one or more separate worker containers.
The licensing is better than commonly claimed — single-main queue mode is free on self-hosted, and it’s multi-main setup that needs an Enterprise licence. We set that out precisely in self-hosting n8n with Docker Compose.
But it’s not free in infrastructure. You’re adding Redis and a worker, which realistically means a second server. The 100,000-execution row above uses $29.98 rather than $14.99 for exactly that reason.
Note what it does to the conclusion: doubling the infrastructure changes the saving from $1,569 to $1,554. The second server is a rounding error against a $1,584 Cloud bill. It’s honest to include it, and it doesn’t rescue the Cloud case.
The thing to check before you commit
One detail I could not resolve, and it matters if you’re near a limit.
n8n’s own pricing page states that if you exceed your quota, “your workflows will continue running without interruption, but overage charges may apply if you do not upgrade to the next usage tier.” Several third-party guides claim the opposite — that workflows stop at the cap on Starter and Pro, with overage buckets available only on Business.
Those cannot both be true, and the difference is the difference between a surprise bill and a silently broken automation. n8n’s own wording is the more authoritative, and it’s what we’d rely on — but if your workflows are business-critical and you’re running near your quota, get this confirmed in writing before you find out empirically.
What’s documented clearly: on Business, additional executions come in buckets of 300,000 for €4,000.
Memory is the other ceiling
Price isn’t the only axis where Cloud constrains you. n8n publishes the memory allocated to each Cloud tier, and it’s low — the Starter trial tier is 320 MiB, and even Enterprise is 4,096 MiB.
A commodity 4GB VPS matches n8n Cloud’s Enterprise memory allocation. That matters because memory is what kills n8n workflows — large payloads, big loops, anything holding a dataset — and on Cloud you cannot buy past the ceiling without changing tier. We covered the full comparison in migrating n8n Cloud to self-hosted.
If your workflows die on Cloud with large datasets, that’s the constraint you’ve hit, and no amount of execution budget fixes it.
The verdict
Under 10,000 executions a month: use n8n Cloud. The saving doesn’t cover your time, you get updates and backups handled, and the tiers are fairly priced at $5.40–8.64 per thousand executions.
From roughly 25,000 upward: self-host. You’re on the wrong side of the Pro-to-Business cliff, and the saving buys more hours of maintenance than the instance will ever need.
Between 10,000 and 25,000: it depends on your rate and your appetite. If you already run servers, self-host. If you don’t, staying on Pro and keeping your executions under 10,000 — by batching, filtering triggers earlier, or turning off success logging — is often the better engineering decision anyway.
Above 100,000: self-host, and budget for queue mode — Redis, a worker node and a second server. It’s still a fraction of the Cloud bill.
What the self-hosted column actually contains
The $11.99 and $14.99 figures above are the VPS alone. A complete instance adds a domain, backups, sometimes object storage, and at agency scale a worker node and possibly managed Postgres and Redis. We itemised all three builds — they come to $12.99, $26.99 and $86.13 a month, and the managed database line costs more than the server it replaces.
Where to run it
Hostinger is the strongest fit here because it has a one-click n8n VPS template, which removes most of the two-to-four hours the ops calculation above is built on. Its 8GB KVM 2 at $8.79 promotional and $14.99 at renewal is the tier most real instances land in, and the $11.99 4GB KVM 1 covers lighter use.
DigitalOcean is the predictable choice with excellent documentation, at $24/month for 4GB — double Hostinger’s renewal rate for the same memory. Its tracked link isn’t working on our end, so we’ve named it without linking.
Hetzner is a natural Docker host and normally the price leader, but check stock — its cheap CX tier has been showing as unavailable, and the CPX line you’d fall back to is €19.99 for 4GB after its June 2026 repricing.
Cloudways is the managed middle path: you get the resources without administering the box, billed monthly with no prepaid term. That directly reduces the ops-time line in the table above, which is the variable that decides this whole comparison.
Railway removes the server entirely, but its metered model bills for residency — and n8n has to stay resident to catch webhooks. We priced that out in Railway vs a VPS for n8n and it works out dearer than a flat VPS quickly.
For sizing rather than pricing, best VPS for n8n covers the RAM brackets and the pruning settings new self-hosters miss.
How we checked this
n8n Cloud plan prices, execution allowances and concurrency limits are from n8n’s own pricing page, read in August 2026: Starter at €20/month annually for 2,500 executions and 5 concurrent, Pro at €50 for 10,000 and 20 concurrent, Business at €667 for 40,000, with Business overages at €4,000 per 300,000 executions. Euro figures are converted at approximately 1.08. We were unable to capture that page — it wouldn’t render its plan cards for us — so unlike most articles in this series there’s no screenshot of the source.
Self-hosted costs are Hostinger’s published KVM renewal rates of $11.99 for 4GB and $14.99 for 8GB, with the 100,000-execution row doubling the 8GB figure to account for a queue-mode worker.
The n8n Cloud memory allocations, the regular versus queue execution modes and the licensing distinction between single-main and multi-main are from n8n’s own documentation, as covered in our Docker Compose guide.
The ops-time figures are estimates, not measurements — two to four hours to set up, about 30 minutes a month to maintain. They’re conventional rather than observed, and they’re the load-bearing assumption in the break-even table. That’s why we’ve given the arithmetic as a formula you can substitute your own numbers into rather than a single answer: at 1,000 executions the conclusion flips on whether maintenance takes more or less than twelve minutes a month, which is well inside the error bar. At 50,000 it would take seventeen hours to flip, which is not.
On the overage contradiction: we’ve quoted n8n’s own page and flagged the disagreement rather than picking a side. We have not tested what happens at the cap, and we’d treat any confident claim either way — including a third party’s — as unverified.
What we did not do: we haven’t run an n8n instance at these volumes, measured what execution counts a given VPS actually sustains, or tested queue mode. The claim that a 4GB box handles 10,000 executions comfortably is conventional guidance, and your payload sizes move it more than your provider does.
The hosting links above are affiliate links. This article recommends n8n Cloud — which pays us nothing — for anyone under 10,000 executions a month.
FAQ
Is self-hosting n8n cheaper than n8n Cloud?
Above roughly 10,000 executions a month, decisively. At 50,000 executions Cloud costs about $864 against $15 for a VPS. Below 10,000 the saving is too small to cover the time you’ll spend maintaining it.
How much does n8n Cloud cost?
€20/month for 2,500 executions on Starter, €50 for 10,000 on Pro and €667 for 40,000 on Business, billed annually. That works out at $8.64, $5.40 and $18.01 per thousand executions respectively.
Why is n8n Cloud’s Business plan so expensive?
It’s 13× the Pro price for 4× the executions, because it’s a governance product rather than a volume tier. Per execution it’s worse value than both Pro and Starter, which is why the self-hosting case becomes overwhelming just above 10,000 executions.
What happens if I exceed my n8n Cloud execution limit?
n8n’s pricing page says workflows continue running and overage charges may apply if you don’t upgrade. Some third-party guides claim workflows stop instead. Confirm this with n8n directly if you run near your quota — the two outcomes are very different.
How much does it cost to self-host n8n?
About $11.99/month for a 4GB VPS or $14.99 for 8GB, plus your own time — two to four hours of setup and roughly 30 minutes a month. Above 100,000 executions add a second server for queue mode.
Do I need Redis and queue mode for n8n?
Only when a long-running workflow blocks others or concurrent webhooks get dropped, typically well above 50,000 executions a month. Single-main queue mode is free on self-hosted; it’s multi-main setup that needs a licence.
At what point should I move off n8n Cloud?
When your executions approach 10,000 a month, because the next tier up is a 13× price jump. That’s the moment to either optimise your executions downward or move to a VPS.
