Railway vs Render (2026): Free Tiers, Real Pricing & Postgres Costs

Quick Verdict: Metered Credits or Fixed Instances?

Choose Railway when you want per-second usage billing with a small monthly credit, want databases that cost only the RAM, CPU and disk they consume, and your team wants unlimited seats on one $20 workspace. Choose Render when you want a fixed monthly price per instance, a Hobby workspace with no subscription fee, and managed Postgres with documented point-in-time recovery and high-availability options. Neither is free for a production API with a database; the free routes are an evaluation path on both.

The pricing models differ at the root. Railway charges a plan fee that includes an equal usage credit—$5 on Hobby, $20 on Pro—and then meters RAM at $10 per GB-month and CPU at $20 per vCPU-month, billed per second. Render charges nothing for the Hobby workspace itself and $25 per month for Pro, then bills each service at a fixed instance price such as $7 for 0.5 CPU and 512 MB, prorated to the second. The same small API can cost more or less on either platform depending on how much of its reserved capacity it actually uses.

Research method: official pricing pages and documentation checked October 24, 2026. This is a documentary comparison, not a hands-on deployment, latency or reliability test. Dollar budgets use regular USD rates or explicitly labeled planning assumptions, excluding taxes and regional variations. The Railway review, Render review and hosting and DevOps category provide context; the dated primary sources below govern this comparison's current claims.

Plans and Subscription Fees

Railway's pricing page lists four plans plus a trial: Free at $0 per month with $1 of monthly usage credits, Hobby at $5 per month including $5 of monthly usage credits, Pro at $20 per month per workspace including $20 of usage credits with unlimited seats, and Enterprise with custom pricing. Its plans documentation is explicit that Hobby is not free: you pay the $5 subscription even if you use less than $5 of resources, and unused credit resets each billing cycle rather than accumulating.

Render's pricing page lists Hobby at $0 per month plus compute, Pro at $25 per month plus compute, Scale at $499 per month plus compute, and Enterprise at custom pricing. Hobby includes up to 25 services, 5 GB of bandwidth and single-service previews; Pro adds unlimited seats and services, 25 GB of bandwidth, full-stack previews and horizontal autoscaling; Scale adds multiple workspaces, 1 TB of bandwidth, HIPAA-compliant workspaces and SAML SSO. If a workspace has no services and no activity during a month, Render waives that month's workspace fee.

Older guides still describe Render's Professional plan at $19 and an Organization tier; the current page uses Pro at $25 and Scale at $499, although some registration links and documentation still use the legacy names. Railway's Pro fee is also often misquoted as per seat; the pricing page states the fee is per workspace with seats unlimited. The startup hosting guide still carries older figures for both platforms, so treat this article's dated numbers as the current reference.

Collaboration is where the plan structures diverge. Railway's pricing table lists 3 members per project on Hobby and unlimited on Pro, while its plans FAQ says Pro and Enterprise are the plans designed for collaboration—both are official statements, so confirm the Hobby member behavior in your own workspace. Render's Hobby workspace has a single seat; Pro removes the seat limit. A two-person team therefore needs Railway Pro ($20) or Render Pro ($25) before adding any compute.

Compute Pricing: Per-Second Metering vs Instance Tiers

Railway's rates are published as $10 per GB of RAM per month, $20 per vCPU per month, $0.05 per GB of network egress and $0.15 per GB-month of volume storage, with object storage at $0.015 per GB-month and free egress from it. The pricing page says all usage is metered per second except egress and object storage, so stopped services cost nothing. Builds are free: Railway does not charge for build CPU, memory, base-image downloads or image storage.

Render's compute plans are fixed tiers: $7 per month for 0.5 CPU and 512 MB, $25 for 1 CPU and 2 GB, $85 for 2 CPU and 4 GB, $135 for 2 CPU and 8 GB, $175 for 4 CPU and 8 GB, $225 for 4 CPU and 16 GB and $450 for 8 CPU and 32 GB. Render introduced new plan IDs in August 2026 with no price changes, so Starter, Standard and Pro now map to 0.5c-512mb, 1c-2g and 2c-4g. Billing is prorated to the second: a paid service turned off after 24 hours is billed for 24 hours.

The difference matters most for idle capacity. A service that reserves 2 GB on Render costs $25 whether it uses 300 MB or 1.9 GB. On Railway, a service that averages 300 MB of RAM and 0.1 vCPU for a full month costs roughly 0.3 × $10 + 0.1 × $20 = $5, while one that averages 2 GB and 1 vCPU costs 2 × $10 + 1 × $20 = $40. Railway rewards right-sized, bursty workloads; Render rewards steady workloads that fill the instance they pay for. The small-business hosting guide explains why both models beat unmanaged servers for teams without an operations engineer.

Per-service ceilings also differ. Railway's documented defaults are 1 vCPU and 0.5 GB per service on Free, 48 vCPU and 48 GB on Hobby and 1,000 vCPU and 1 TB on Pro, with replica limits of 1, 6 and 42. Render's largest listed tier is 8 CPU and 32 GB, with horizontal autoscaling available on Pro workspaces and above. Neither number is a performance guarantee; they are the limits you can configure.

Free Tiers: Credits That Expire vs Services That Sleep

Railway's free trial is a one-time grant of $5 in credits that expire in 30 days, limited to 1 GB of RAM, shared vCPU and 5 services per project, with no credit card required. After 30 days or $5 spent, the account reverts to the Free plan with $1 of credit per month, 1 project, 3 services, 0.5 GB of RAM and no custom domain. Railway deletes stateful volumes created by trial accounts 30 days after the credits expire, so a trial database is not a place to keep data.

Render's free instance documentation grants each workspace 750 free instance hours per calendar month; once consumed, Render suspends all free web services until the next month. A free web service that goes 15 minutes without inbound traffic spins down and spins back up on the next HTTP request, a process Render says takes about one minute. Free compute is 0.1 CPU and 512 MB, static sites are free, and free services draw from the Hobby workspace's 5 GB bandwidth and 500 pipeline minutes.

Free databases are time-limited on Render and credit-limited on Railway. Render allows one active free Postgres with 1 GB of fixed storage that expires 30 days after creation, followed by a 14-day grace period before deletion, and free databases do not support any form of backups. Railway has no separate free database product: a Postgres template on the Free plan simply consumes the $1 monthly credit through RAM, CPU and volume rates, and the 0.5 GB volume cap applies. A hobby project that must stay online for months fits neither free tier comfortably; the Node.js hosting guide covers where permanent free hosting still exists.

Sleep behavior is opt-in on Railway and automatic on Render Free. Railway's serverless setting, formerly App Sleeping, puts a service to sleep after roughly 5 to 10 minutes without outbound traffic and wakes it on the first request, which may return a 502 while waking. Render's documentation describes spin-down only for free web services; paid instances are billed continuously, which is what the "turn it off after 24 hours, pay for 24 hours" FAQ implies.

Worked Budget: One API Plus Postgres for a Two-Person Team

Assume a small production API that needs about 1 GB of RAM and half a vCPU on average, a Postgres database that needs 1 GB of RAM, half a vCPU and 10 GB of storage, 20 GB of monthly egress, and two engineers who need their own logins. These are planning inputs chosen to make the arithmetic visible, not measured requirements for any real application, and they ignore regional pricing, taxes and spikes.

On Railway Pro, the workspace fee is $20 and includes $20 of usage. Compute for both services is 2 × (1 GB × $10) + 2 × (0.5 vCPU × $20) = $20 + $20 = $40. Storage is 10 GB × $0.15 = $1.50 and egress is 20 GB × $0.05 = $1.00, so usage totals $42.50. Subtracting the $20 credit leaves $22.50 of overage, and the monthly bill is $20 + $22.50 = $42.50. If the API idles at 300 MB and 0.1 vCPU most of the month, compute drops to $3 + $2 = $5 for that service and the bill falls to $27.50.

On Render Pro, the workspace is $25. The nearest web-service tier with at least 1 GB of RAM is 1 CPU and 2 GB at $25. The nearest Postgres tier with 1 GB is 0.5 CPU and 1 GB at $19, and storage beyond the included 1 GB is 9 GB × $0.30 = $2.70. Egress of 20 GB sits within Pro's 25 GB allowance, so the bill is $25 + $25 + $19 + $2.70 = $71.70. Idle time does not reduce it, but neither does a traffic spike until you choose a larger instance or enable autoscaling.

The gap narrows for a solo developer. On Railway Hobby, the same two services at full utilization cost $42.50 of usage against a $5 credit, for a bill of $5 + $37.50 = $42.50; on Render Hobby the workspace is $0, so the bill is $25 + $19 + $2.70 = $46.70, with egress inside the 5 GB allowance only if traffic stays under it—20 GB would add 15 GB × $0.15 = $2.25. Use the SaaS spending worksheet to record your own measured RAM, CPU and egress before trusting either estimate.

Databases: Templates You Operate vs Managed Postgres

Railway's Postgres documentation describes a single node that Railway maintains, billed through the same RAM, CPU and volume rates as any service, with public access billed as egress through the TCP proxy. Volume backups run daily and are kept for 6 days, weekly for 27 days and monthly for 89 days, priced like volumes, with manual backups limited to 50% of the volume's size. Point-in-time recovery retains the last four full backups for a roughly four-week Postgres window and is billed only through storage and egress meters.

Render's Postgres is a separately priced managed product: $6 per month for 0.1 CPU and 256 MB, $19 for 0.5 CPU and 1 GB, $40 for 1 CPU and 2 GB, $55 for 1 CPU and 4 GB and higher tiers, each including 1 GB of SSD storage with expansion at $0.30 per GB. Its backup documentation gives Hobby workspaces a 3-day point-in-time window and Pro or higher a 7-day window, retains logical backups for seven days regardless of plan, and offers no recovery for free databases.

High availability is where the products most clearly differ. Render's HA documentation requires at least 1 CPU and bills the standby at the same compute plan and storage as the primary; read replicas need at least 10 GB of storage and 0.5 CPU, up to five replicas, each billed like the primary. Railway's HA Postgres converts a database into a Patroni, etcd and HAProxy cluster with two to seven replicas, each billed as additional services, and Railway states it does not publish an uptime SLA today.

Treat both as operational commitments rather than fire-and-forget storage. Verify a restore from a backup in a scratch environment, confirm the retention window you actually get on your plan, and keep an off-platform export for anything you cannot afford to lose; the free cloud storage guide explains why a platform snapshot is not the same as your own archive. Neither vendor's documentation reviewed here promises that a backup exists for a database you deleted.

Bandwidth, Regions, Domains and Networking

Egress is a flat meter on Railway and an allowance-plus-overage model on Render. Railway charges $0.05 per GB with no stated free allowance beyond the plan credit. Render includes 5 GB on Hobby, 25 GB on Pro and 1 TB on Scale, then charges $0.15 per GB according to its outbound bandwidth documentation, with $0.03 per GB through AWS Private Link; unused bandwidth does not roll over, and a workspace without a payment method is spun down when the allowance is exhausted. A 100 GB month costs $5.00 on Railway and $11.25 on Render Pro (75 GB over at $0.15).

Regions are comparable in count but not location. Railway's regions documentation lists US West (California), US East (Virginia), EU West (Amsterdam) and Southeast Asia (Singapore). Render's regions documentation lists Oregon, Ohio, Virginia, Frankfurt and Singapore. A team serving Germany gets Frankfurt on Render and Amsterdam on Railway; neither offers a documented region in South Asia, Australia or South America in the pages reviewed.

Custom domains carry free TLS on both platforms, but the counts are plan-gated. Railway's table lists 0 custom domains on Free, 2 on Hobby and 20 on Pro. Render includes 2 on Hobby, 15 on Pro and 25 on Scale, with additional domains at $0.25 each per month; automatic TLS, DDoS mitigation and a firewall are listed on all plans. Railway's private networking connects services over encrypted WireGuard tunnels with internal DNS, so a database that is never exposed publicly avoids the egress charge on the TCP proxy.

Scheduled work is supported on both with small operational caveats. Railway's cron jobs must be at least 5 minutes apart and are skipped if the previous run is still executing. Render's cron jobs are prorated by the second with a minimum monthly charge of $1 per cron job service. If your scheduler runs every minute, that is a Railway limit to design around; if you have twenty tiny crons, that is a Render minimum to budget for. The Vercel alternatives guide explains why these container platforms handle long-running and scheduled work that serverless hosts cannot.

Deploys, Previews and Disks

Both platforms build from GitHub repositories and Dockerfiles. Render's pricing page lists zero-downtime deploys, deploy hooks, pre-deploy commands, Docker builds, auto-deploys and infrastructure as code through a render.yaml blueprint on all plans, with GitHub, GitLab and Bitbucket as providers; single-service previews are on every plan and full-stack preview environments are Pro and above, with rollback history of 5, 15 or 30 builds by plan. Railway documents GitHub autodeploys, Dockerfile builds and environments, and bills none of the build time.

Persistent disks change Render's deploy behavior. Its disk documentation prices SSDs at $0.25 per GB per month, snapshots them once every 24 hours with seven days of availability, and warns that adding a disk prevents zero-downtime deploys because Render stops the existing instance before starting the new one—and that a service with a disk cannot scale to multiple instances. Railway volumes cost $0.15 per GB-month and are capped at 0.5 GB on Free, 5 GB on Hobby and up to 1 TB on Pro.

Those constraints shape architecture more than price. If your API writes uploads to local disk, Render will trade zero-downtime deploys and horizontal scaling for that convenience; moving files to object storage restores both. Railway's object storage at $0.015 per GB-month with free egress is the documented route for the same pattern. The software evaluation checklist has a place to record which services need state and which can stay stateless before you compare instance prices.

Support and Who Answers at 2 a.m.

Railway's support documentation gives Trial, Free and Hobby customers community support through its Central Station forum, where responses are not guaranteed; Pro customers typically receive responses within 72 hours; Business Class support, with one-hour 24/7 response for P1 incidents, is available after $5,000 per month in spend. Railway states it does not provide support via email.

Render lists email and chat support on all plans, Premium Support as an add-on for Pro and Scale with a Slack channel and response SLAs, and contractual uptime SLAs, a dedicated technical account team and support response SLAs on Enterprise. For a small team, that means Render's base plans include a direct support channel where Railway's Hobby plan relies on the community. Neither base plan includes an uptime SLA; Railway explicitly says it does not publish one, and Render's SLA is an Enterprise term.

Price the human side of outages the same way you price compute. A $5 Hobby plan with community-only support may be fine for an internal tool; a customer-facing API that must be fixed at 2 a.m. needs either a paid support tier, your own on-call rotation, or both. The small-team project management guide is the wrong place to run incidents, but it is where the escalation checklist should live before the first one.

Migration Checklist Between the Two

Moving a containerized service is straightforward because both platforms build Dockerfiles; moving the database is where migrations stall. Export a logical dump from the source database, restore it into a scratch database on the target, and compare row counts and a sample of application queries before scheduling cutover. Both vendors' backup features are platform-scoped: Railway's volume backups and Render's PITR do not transfer to the other provider.

Reconcile the cost model before the DNS change, not after. On Railway, confirm which services should use the serverless sleep setting and which must stay warm, and check that private networking is used between the API and database so the TCP proxy egress meter stays quiet. On Render, choose instance tiers from measured RAM rather than guesses, decide whether any service needs a disk, and check the bandwidth allowance against last month's egress. Keep the old deployment running until the new one has served real traffic for at least one full billing cycle.

Document the plan fees as separate line items: Railway's $5 or $20 includes an equal usage credit, Render's $0 or $25 includes bandwidth and seats but no compute. Teams that compare only the headline compute prices routinely miss that difference. If the real question is whether to leave a serverless host at all, the Vercel vs Netlify comparison covers the other side of that decision.

Which Platform Fits Your Backend?

Choose Railway when your workloads are bursty or right-sized, when you want databases, workers and crons billed as the resources they use, when free builds and per-second metering matter, and when a team of any size can share one $20 Pro workspace. Accept community support below Pro, no published SLA, a 5-minute cron floor and the discipline to monitor usage so the credit does not quietly turn into overage.

Choose Render when you want a fixed invoice per instance, a $0 Hobby workspace for a solo builder, managed Postgres with documented PITR windows and HA billing, included bandwidth allowances, email and chat support on every plan, and blueprint-based infrastructure as code. Accept paying for reserved capacity you may not use, disk-attached services that lose zero-downtime deploys, and a free tier whose web services sleep and whose databases expire. Validate the choice with a measured week of usage on each platform's lowest paid tier rather than with either free tier alone.