Railway pricing explained: what you'll actually pay in 2026

How Railway pricing actually works

Railway's pricing page shows two numbers — Hobby $5/month, Pro $20/month — and both are accurate. They're just not the whole bill. Railway charges a plan fee plus metered usage, and the usage is what people don't see coming.

There are three meters:

• Memory (RAM) — roughly $10 per GB-month, billed per second for however much your app holds.

• CPU (vCPU) — roughly $20 per vCPU-month, billed per second of actual compute.

• Network egress — roughly $0.05 per GB of outbound traffic.

On Hobby, the $5/month plan fee includes $5 of usage — so a truly idle app stays near $5. On Pro, the $20 is a base and usage stacks on top. The key mental shift coming from Heroku or a flat host: there's no fixed “dyno” price. Your bill is a function of how much your app runs and how much data it sends. Let's put real numbers on three common cases.

Scenario 1: a hobby app

A small personal project — a landing page with a form, a tiny API, or a weekend app you show friends. It holds ~256MB RAM, uses almost no CPU, and serves little traffic. It's idle most of the day.

The math. 256MB for a month is about $2.50 of RAM; CPU and egress are pennies. Total usage lands around $3–5 — inside the Hobby plan's included $5.

What you pay: ~$5/month. This is Railway at its best. For bursty, low-traffic projects the meter genuinely works in your favor.

Scenario 2: self-hosted n8n running 24/7

Now an always-on workflow tool like self-hosted n8n. The difference from Scenario 1 is the word always: n8n holds memory continuously, even between workflow runs, because it never sleeps.

The math. n8n comfortably holds ~1GB RAM in steady use — that's ~$10/month in memory alone, every month, whether or not a workflow fires. Add modest CPU for executions and egress for API calls and webhooks, and you're at $15–25/month.

What you pay: ~$15–25/month, and it scales with how busy your workflows get. This is the first place the Hobby “$5” framing breaks down — nothing is wrong, the app simply runs all the time, so the meter runs all the time.

Scenario 3: a small SaaS under real traffic

A real product: a web service, a background worker, and a managed Postgres database, with actual users hitting it through the day.

The math. The web service holds ~512MB–1GB and uses real CPU during requests. The worker holds memory continuously. Postgres holds memory and storage. Together that's commonly 2–3GB of RAM-equivalent running most of the time (~$20–30), meaningful CPU during traffic, and egress that grows with users and any media you serve.

What you pay: ~$40–70/month, trending up with growth. Still reasonable for a revenue product — but a long way from the $5 on the pricing page, and it moves every month.

Usage-based vs flat: the real comparison

None of these rates are unfair. The issue is predictability — you can't know the number until the month ends, and a traffic spike or a memory leak shows up as a bigger invoice rather than a slow site. That's exactly why people compare Railway to flat pricing.

On a flat host like livemy.app, the same three apps cost the same thing: $10/month, every month, with custom domain, free SSL, and monitoring included and no compute meter. The hobby app “overpays” slightly versus Railway's $5; the n8n instance and the small SaaS pay far less and — more importantly — pay a number they chose in advance.

• Hobby app: Railway ~$5 · livemy.app $10 flat

• n8n 24/7: Railway ~$15–25 · livemy.app $10 flat

• Small SaaS: Railway ~$40–70 · livemy.app $10 flat

The crossover is fast: anything always-on tends to favor flat pricing. If you're weighing the switch, see the full lineup in Railway alternatives.

How to lower your Railway bill

• Sleep idle services. Railway can sleep services that aren't serving traffic — use it for staging and side projects so the meter stops.

• Right-size RAM. Memory is the biggest line for most apps. Don't let a Node or Python process hold a gigabyte it doesn't need; tune your runtime's memory settings.

• Watch egress. Serving images, video, or large API responses adds up at $0.05/GB. Offload static assets to a CDN or object storage.

• Consolidate databases. A separate Postgres per project each holds memory 24/7. Share where it's safe to.

FAQ

Does Railway have a free tier?

There's a limited trial, but no permanent free always-on tier — the Hobby plan starts at $5/month including $5 of usage. For a permanent free option, Render's free tier (services sleep after 15 min) or livemy.app's free tier (1 project on a livemy.site URL, no card) are closer.

Why is my Railway bill higher than $5?

The $5 Hobby fee includes only $5 of usage. Any always-on service holding RAM around the clock, plus CPU and egress, bills on top. An app that never sleeps is the usual reason.

Is Railway cheaper than Render or Heroku?

For small, bursty apps, often yes. For steady, always-on apps, fixed-price hosts (Render $7/service, livemy.app $10 flat) are usually cheaper and always more predictable. See Render vs Railway.

What happens if I exceed my usage?

Railway keeps running and bills the overage — it doesn't cut you off mid-month on paid plans. That's convenient for uptime and the exact reason bills can surprise you. Set usage limits and alerts in the dashboard.

How do I make my Railway cost predictable?

Either cap usage with hard limits (risking downtime if you hit them), or move always-on workloads to a flat-priced host. If predictability is the goal, flat pricing solves it by design.

The pricing page is the floor, not the bill

Railway's rates are fair and its product is excellent — the thing to internalize is that you're renting a meter, not a fixed box. For bursty hobby work that's a bargain. For anything that runs all month, decide whether you'd rather optimize usage forever or pay one flat number and move on.

→ Try livemy.app free · No credit card · Flat $10/month, no compute metering.