n8n Self-Hosted vs. Cloud: The April 2026 Calculus

Published April 30, 2026 by Pondero Editorial

The short version

"Self-host n8n, it's free" is wrong more often than it's right, and the reason is not infrastructure cost. The license is free; the job is not. The decision turns on one question: does a named person own the operational tax week after week? If the answer is no, the hosting savings are a loan you repay in incident time. Under roughly 10,000 executions a month, once you price your own hours at a real rate, n8n Cloud is the cheaper option for the median ops team. Self-hosted wins clearly in exactly three cases, and you should be able to name which one applies before you reach for Docker: a hard data-residency rule, genuinely high volume on simple triggers, or workflows that reach into internal services a SaaS cannot.

The decision matrix

Dimension	Self-hosted n8n	n8n Cloud
Sticker price	Free (license)	From $20/mo
Real cost	Hosting + ops time + monitoring + upgrades	Subscription + usage tier
Data residency	Wherever you put it	Vendor-managed regions
Time to first workflow	Half-day (Docker) to a week (k8s)	Minutes
Best fit	Engineering-led, infra-fluent teams	Ops-led, automation-first teams
Hidden tax	Upgrade fatigue, on-call surface area	Less control over scaling cliffs
Where it bites	Forgotten upgrades cause drift and security holes	Per-execution pricing at high volume

For the broader category context, our best AI automation tools guide places n8n alongside Zapier and Make.

When self-hosted is unambiguously the right call

Data residency or compliance forces it (HIPAA workflows in a private VPC, EU public-sector clients, etc.).
Volume is high enough that Cloud's execution tiers stop being economical, typically when monthly executions push past tens of thousands and most workflows are simple, frequent triggers.
Workflows reach into internal services that would require punching new holes through your network to expose to a SaaS.
You already have a platform team running container workloads. The marginal cost of adding n8n is small.

When Cloud is the right call (and most teams underestimate this)

Ops-led teams without dedicated platform engineering. Self-hosting n8n is straightforward to start and surprisingly expensive to keep healthy. Upgrades, secrets rotation, backup/restore, and monitoring all cost real time.
Workflows are mostly outbound to SaaS APIs. If n8n is calling Slack, HubSpot, Stripe, and a few HTTP endpoints, the network argument for self-hosting is weak.
You're in the validation phase. Cloud lets you commit to n8n with hours, not days, of setup. Migrate later if the math shifts.

The candid cost equation

The self-host estimate that goes wrong always makes the same error: it prices the VM and stops. The VM is the cheapest line in the budget. The three that dominate it:

Upgrade tax. n8n ships fast. Staying current is a recurring half-day per quarter minimum, plus regression testing the workflows that the upgrade can silently change. Skip it for two quarters and you are now running a version with known credential-handling fixes you do not have. The cost is not the upgrade. It is that the upgrade is never urgent until it is.
Backup and DR. Workflow JSON, credentials, and execution logs need a restore you have actually tested. "It's in Postgres" is a backup of the database, not a recovery plan for the automation layer, and the difference shows up the one time you need it.
On-call surface. A self-hosted platform running your billing reconciliation is a Tier-1 system whether or not anyone labeled it one. The pager exists; the only question is whether you staffed it on purpose or discovered it at 2am.

Add those three at a real hourly rate and Cloud usually wins on total cost before vendor support even enters the math. The line items are not hard individually. They are a standing job collectively, and a standing job needs an owner, not good intentions.

The migration question

Already on Cloud and itching to move? Wait for a named reason. A residency clause landing in a signed contract. A workflow class that needs a private network. A volume cliff you have actually hit, not one you might. Don't migrate on vibes. Workflows export cleanly, so portability is real. The ops surface area you inherit is the real price, and it does not show up on the export.

Where Cloud's pricing model bites

n8n Cloud's tiers scale on executions. The cliff most teams hit is when a single high-frequency trigger (every-minute schedule, busy webhook) blows through the included executions on a mid-tier plan. Audit your workflows for "always-on" triggers before signing up; a single mis-configured cron is the most common surprise on the bill.

When this verdict flips

Default to n8n Cloud. The recommendation inverts the moment you can finish this sentence with a fact, not a feeling: "we must self-host because ___." Valid completions: a residency clause in a signed contract; a workflow that must call an internal service with no SaaS-reachable surface; a sustained execution volume on simple triggers where Cloud's per-execution tier costs more than a staffed ops slice. If you already run a platform team shipping container workloads, the marginal cost of adding n8n is small and the call flips on that alone. Absent one of those, "we'll just self-host" is the more expensive option wearing the cheaper price tag.

Verdict

Default to n8n Cloud. Self-host only when you can name the reason out loud and it is on the list above. The total-cost math tipped toward Cloud for the median ops team this year, and not because hosting is hard. It tipped because keeping a Tier-1 automation system healthy is a standing job, and most teams do not have a spare person to hand it to.

Try n8n Cloud: the fastest way to validate whether it fits your workflows.