Railway vs a VPS for n8n: You're Billed for Residency, Not Executions

Railway’s n8n template is genuinely excellent: click, wait about ninety seconds, and you have a working instance with none of the Docker setup. For getting started it is hard to beat.

The question is what it costs once it’s real — and the answer turns on a detail that most comparisons miss.

Railway bills resource-time, not executions. You pay for the vCPU and RAM your container holds, per second. And n8n cannot idle: it has to stay resident to receive webhooks and fire scheduled workflows. So the thing Railway’s pricing model rewards — scaling to zero between requests — is precisely the thing n8n can’t do. The same arithmetic decides any always-on app: we worked the crossover for a Node service and it lands near 1GB resident, not at any particular request count.

Arriving from Heroku? The same residency question decides it — Heroku alternatives maps every Heroku primitive to its replacement and prices a real app on each platform.

The rates

Plans: Hobby at $5/month including $5 of usage, Pro at $20/month per workspace, seats included, with $20 of usage. Both bill resources above the included credit at the same per-second rates.

Because n8n stays resident, the RAM line accrues around the clock:

Published estimates put a real n8n instance on Railway at roughly $15–25/month once CPU for executions and egress for API calls and webhooks are added to a ~1 GB footprint.

The crossover

Set that against a flat VPS with unlimited executions:

Railway at roughly 1 GB already costs more than an 8 GB VPS at renewal, while giving you an eighth of the memory.

The brief for this article assumed there’d be an execution volume where Railway’s metered pricing crosses a flat VPS. There mostly isn’t one, because execution count isn’t what you’re billed for. You’re billed for staying switched on, and n8n is switched on continuously by design. Whether you run 100 executions a month or 100,000, the RAM-time bill is broadly the same — it’s the memory footprint and uptime that set it.

That’s not Railway being expensive. It’s a pricing model built for bursty, scale-to-zero workloads meeting a workload that is neither. The same model looks very different against another metered platform — we compare it directly in Railway vs Fly.io, where Railway’s habit of billing actual consumption rather than provisioned capacity is usually the cheaper side.

The three usage patterns

Bursty — a daily digest that runs for two minutes. You might expect this to be Railway’s best case. It isn’t, because the container has to stay up to fire the schedule. Sleeping it would mean missing the trigger. So you pay for 24 hours of residency to use two minutes of it.

Always-on webhook receiver. The honest baseline. Railway bills continuously; the VPS bills a flat rate. The VPS wins on cost and the gap widens with memory.

Heavy scraping. The worst case for Railway, because you add sustained CPU at $20/vCPU/month and egress at $0.05/GB on top of resident RAM. A scraping workload that pulls tens of gigabytes a month adds a meaningful line the VPS simply doesn’t have — most VPS plans include multiple terabytes of transfer.

The durability problem

Cost is the smaller issue. This is the bigger one.

Railway sleeps and redeploys containers. Deployments, platform maintenance and configuration changes all restart the container — and that interacts badly with n8n in ways a persistent VPS doesn’t:

• Long-running executions are interrupted. A workflow midway through a batch, a large HTTP request, or a multi-step sequence is cut off when the container cycles.
• Webhook events arriving during a restart are simply gone, unless the sender retries. The caller gets an error and there’s nothing to catch it.
• Anything not on a persistent volume is lost on redeploy, which is why the database and n8n’s data must be on a volume rather than container-local storage.

A VPS reboots when you reboot it. That predictability is the actual product difference, and it’s worth more than the price gap for anything load-bearing.

If you do run n8n on Railway, mitigate: put data on a persistent volume, use queue mode so interrupted jobs can be re-queued rather than lost, and make webhook senders retry. We covered the in-flight execution problem in more depth in DigitalOcean vs Hostinger for n8n.

The verdict

Railway to prototype. A VPS the moment it’s load-bearing.

Railway’s template is the fastest path from nothing to a working n8n, and for evaluating whether n8n solves your problem at all, ninety seconds and a few dollars is excellent value. Use it. The Hobby plan’s $5 included usage covers meaningful experimentation.

Move to a VPS when the workflows start mattering — when someone depends on the daily digest arriving, when a webhook dropping means a lost order, or when the bill crosses roughly $15 a month. At that point a flat VPS gives you more memory, unlimited executions, no metered egress, and a machine that restarts only when you say so.

Stay on Railway if what you value is never touching a server — no OS patching, no Docker, no backups to configure — and you accept both the metered bill and the restart behaviour. That’s a legitimate trade for a low-stakes automation.

Check Hostinger VPS pricing → · Read our Hostinger review

Check Railway → · Read our Railway review

Our n8n VPS guide covers sizing, queue mode and the execution-data pruning that stops a self-hosted instance filling its disk.

How we picked

Railway’s rates are its published usage pricing read in August 2026: $20 per vCPU per month, $10 per GB of RAM per month, $0.05/GB egress, $0.15/GB/month volumes, with Hobby at $5 and Pro at $20 per workspace (seats included) with matching usage credit. The $15–25/month figure for a running n8n instance is a published estimate for a ~1 GB footprint, not our measurement.

Hostinger pricing is from its VPS page: KVM 1 at $6.49 introductory and $11.99 renewal, KVM 2 at $8.79 and $14.99, on 24-month terms.

We have not run an n8n instance on Railway for a month and reconciled the invoice, which is the only way to establish your actual usage cost — it depends on your workflows’ memory footprint and egress. The structural point stands regardless: you’re billed for residency, and n8n is resident continuously.

Both links here are affiliate links. The recommendation for anything beyond prototyping points at the flat VPS.

FAQ

How much does n8n cost on Railway?

Roughly $15–25 a month for a ~1 GB instance once CPU and egress are included — RAM alone is $10/GB/month and n8n stays resident continuously. The Hobby plan’s $5 included usage covers light experimentation but not a persistent instance.

Is Railway cheaper than a VPS for n8n?

Generally no, once it’s running continuously. A ~1 GB Railway instance at $15–25 costs more than a Hostinger 8 GB VPS at $14.99 on renewal, which also gives unlimited executions and multiple terabytes of transfer. Railway wins on setup time, not cost.

Does Railway sleep my n8n instance?

Containers can sleep and are restarted on deployments and platform maintenance. That interrupts long-running executions and drops webhook events arriving during the restart. Use queue mode and persistent volumes to reduce the damage, and don’t rely on it for time-critical automations.

At what execution volume should I move off Railway?

Execution volume isn’t the right trigger, because Railway bills resource-time rather than executions. The triggers are the bill crossing roughly $15 a month, or the workflows becoming load-bearing — whichever comes first.

Can I keep n8n data on Railway?

Only on a persistent volume, billed at $0.15/GB/month. Anything written to container-local storage is lost on redeploy. Put both n8n’s data and its database on volumes, or use an external managed database.

What’s the fastest way to try n8n?

Railway’s template — about ninety seconds from click to a working instance, with no Docker knowledge required. It’s genuinely the best evaluation path; just plan the move before it becomes something you depend on.