Railway pricing explained: what $5 and $20 actually buy

How Railway's per-second billing works

Railway does not sell instance sizes. Each service is metered per second on CPU, memory, volumes and egress, and the pricing page prints the unit rates: CPU at $0.00000772 per vCPU-second, memory at $0.00000386 per GB-second, volumes at $0.00000006 per GB-second, and service egress at $0.05 per GB.

Spread across a 30-day month of 2,592,000 seconds, those rates convert to about $20.01 per vCPU-month, about $10.01 per GB-month of memory, and roughly $0.16 per GB-month of volume storage. Railway quotes the same arithmetic in its sandbox table: sandbox CPU and memory are $0.00001929 per second, which the page describes as about $50 per month while running. Sandboxes are a separate, pricier meter from regular services, so agent and untrusted-code workloads do not price like a web app.

Egress is the other line worth watching. At $0.05 per GB, serving 200 GB of outbound traffic adds $10 to the bill, and object storage is metered separately at $0.015 per GB-month with free egress.

Free, Hobby, Pro, and Enterprise at a glance

• Free ($0 per month): 30-day trial with $5 of credits, up to 1 vCPU and 0.5 GB RAM per service, 0.5 GB of volume storage, no credit card required.
• Hobby ($5 minimum usage, includes $5 of credits): up to 48 vCPU and 48 GB RAM per service, up to 5 replicas at 8 vCPU / 8 GB each, 5 GB of storage, a single developer workspace, 7-day log history, 99.9% availability target.
• Pro ($20 minimum usage, includes $20 of credits): up to 1,000 vCPU and 1 TB RAM per service, up to 42 replicas at 24 vCPU / 24 GB each, 1 TB of storage, unlimited workspace seats, 30-day log history, 99.99% availability target.
• Enterprise (custom): 90-day log history, single sign-on, role-based access control, HIPAA BAAs, support SLOs, 99.999% availability target and up to 2,400 vCPU per service.

Read the fee as a floor, not a surcharge. If your Hobby usage comes to $3 in a month, you pay $5. If it comes to $60, you pay $60 and the $5 credit is already inside that number.

What $5 and $20 buy once the meters run

Always-on services are the expensive case, because every second is billable. Using the monthly equivalents above:

• 0.5 vCPU / 0.5 GB, running all month: 0.5 x $20.01 plus 0.5 x $10.01, or about $15. That clears the Hobby minimum, so the invoice is $15, not $5.
• 1 vCPU / 1 GB, running all month: about $30.
• 1 vCPU / 2 GB, running all month: about $40.
• A staging service at 1 vCPU / 1 GB that runs three hours a day: three twenty-fourths of $30, or about $3.75, which floors at the $5 Hobby minimum.
• 200 GB of egress on top of a busy API: $10, which is a separate line from the compute meters.

That fourth line is the actual selling point of usage billing, and also why two teams on the same plan can report wildly different bills. Your number depends on idle time, not on a plan tier. Pull a week of metrics from the dashboard, average the vCPU and memory curves, multiply by 730 hours, and you have a forecast that matches Railway's math within a few dollars. Add volumes in GB and egress in GB, then compare the total to the $5 or $20 floor to see whether the minimum is even binding.

Railway vs Render pricing, verified side by side

Render runs the opposite model: a flat workspace fee, then fixed-size instances with published monthly prices. Both pages were checked on October 22, 2026, so these figures are the current rate cards rather than list prices from a review written two years ago.

The line that decides most migrations is one shape: an always-on web service with 1 vCPU and 2 GB of RAM. Render lists 1 CPU / 2 GB at $25 per month, and the Hobby workspace costs nothing. On Railway the same shape is CPU plus memory metered around the clock, roughly $40. Railway only closes that gap when the service sits idle, which is exactly the workload a fixed instance prices badly.

Bandwidth and seats are the quiet variables. Render's Hobby workspace includes 5 GB of egress, Pro 25 GB and Scale 1 TB, while Railway charges $0.05 per GB with no allowance. On seats it reverses: Railway Pro includes unlimited workspace seats inside the $20 minimum, and Render charges $25 per month for the Pro workspace.

Where the price curves cross

Render's instance ladder is not smooth, and that makes "which is cheaper" unanswerable without a size. Compare always-on shapes against the published Render prices:

• 0.5 vCPU / 0.5 GB: Railway about $15, Render $7 for 0.5 CPU / 512 MB.
• 1 vCPU / 2 GB: Railway about $40, Render $25.
• 2 vCPU / 4 GB: Railway about $80, Render $85.
• 4 vCPU / 8 GB: Railway about $160, Render $175.
• 8 vCPU / 16 GB: Railway about $320, Render $300.

The winner flips three times along one axis. Small always-on services belong on Render's cheapest instances, mid-size ones between 2 and 4 vCPU are cheaper on Railway, and the very dense instances go back to Render at $300 for 8 CPU and 16 GB. Any comparison that answers "Railway is cheaper" or "Render is cheaper" without a shape is guessing.

Utilization is the second axis. Every number above assumes 730 hours of continuous compute. Cut the duty cycle to a third and Railway's mid-size figures drop toward $27, well under any fixed instance of the same shape. That is the trade: fixed instances punish idle, metered billing punishes sprawl.

The catches the pricing page does not spell out

Hobby is a one-person workspace. Collaboration, more than traffic, is what pushes a small team to Pro: $20 minimum usage with unlimited seats is cheaper than Render's $25 Pro workspace, but only if your compute stays spiky.

Seven days of logs on Hobby. Volume storage and log retention are the two meters people forget. If you need to reconstruct what a service did three weeks ago, Hobby's 7-day log history will not help, and Pro's 30-day window is part of what the $20 buys.

Variable invoices cut both ways. A $60 bill this month tells you nothing about next month unless you track utilization. Render's fixed instances are duller but easier to put in front of a finance team.

And the current incident. On October 21, 2026, Railway disclosed that it detected anomalous activity in systems storing customer environment variables, and that an unauthorized third party may have accessed encrypted environment variables and, for some of them, obtained decrypted contents. Railway asked every customer to roll all environment variables, API keys and private registry credentials. Its status page then showed all deployments paused, then reported that builds stayed paused into October 22 while bucket credentials were rotated for every workspace.

That sequence is a cost of vendor risk, not a line item. The recommended fix for a leaked secret is a redeploy with new values, and paused builds mean you cannot complete that fix on the platform itself while the investigation runs. Railway has not attached any pricing change to the incident, and the rate card above is unchanged as of October 22, 2026. Scope is still being narrowed, so check status.railway.com before you plan a migration in either direction.

When Railway is the cheaper bill, and when Render is

Railway's metered model wins when utilization is low or lumpy. Background workers that sleep between jobs, cron-driven pipelines, preview environments that exist for an afternoon, dozens of tiny services in one project: all of these pay Railway for the seconds they use and almost nothing else. If you are pricing agent sandboxes or bursty inference jobs, the separate sandbox meter is the number to model, and it is not cheap: about $50 per vCPU-month, more than double the service rate.

Render wins on predictability and on dense always-on services. A fixed 1 CPU / 2 GB instance at $25 per month is hard for a metered competitor to beat once the process never sleeps, and persistent SSDs at $0.25 per GB-month are simple to budget. Compliance pulls the same direction: HIPAA-compliant workspaces are listed under Render's Scale plan, while HIPAA BAAs sit on Railway's Enterprise tier.

For the wider field of platform hosts on fixed pricing, see our Vercel vs Netlify comparison and the Vercel alternatives roundup. If your real question is which free tier covers a side project, Oracle Cloud's free tier is the other comparison worth running against Railway's $5 floor.

At a glance

Railway and Render plans, both checked on October 22, 2026

FAQ

Is Railway free?

There is a $0 Free plan with a 30-day trial that includes $5 of credits, capped at 1 vCPU and 0.5 GB RAM per service and 0.5 GB of volume storage. After the trial, the cheapest paid path is Hobby at a $5 per month minimum that includes $5 of usage credit.

Is Railway cheaper than Render?

It depends on instance size. Render lists 1 CPU / 2 GB at $25 per month against roughly $40 of metered Railway compute, but at 2 vCPU / 4 GB Railway runs about $80 against Render's $85. Railway also wins whenever services idle, since billing stops mattering as utilization drops.

What happened to Railway on October 21, 2026?

Railway disclosed that an unauthorized third party may have accessed encrypted environment variables, some of them decrypted, and asked all customers to roll environment variables, API keys and registry credentials. Deployments were paused while it investigated and rotated bucket credentials, and builds remained paused on October 22 with no pricing change announced.

Related reading

Vercel vs Netlify comparison, Vercel alternatives roundup, Oracle Cloud free tier

Sources

• Railway pricing page
• Railway security disclosure, October 21, 2026
• Railway status page
• Render pricing page