Render's $25 Flat Team Plan vs Railway's $20 Seats vs Fly.io's $0.02 Egress: Pricing the Same 5-Person Workload in 2026

Heroku's move to sustaining engineering in early 2026 — no new features, no new enterprise contracts — finally closed the door on the platform that defined git-push deploys. Thousands of teams need a new home, and 2026's third-party testing has landed on a split verdict: Render wins for teams after its April 2026 move off per-seat pricing, Railway takes fastest developer experience with auto-managed PR environments, and Fly.io demands the most glue work for team workflows. All three claims are true. None of them answers the question a migrating team actually asks, which is what the same workload costs on each — and whether any of them beats leaving altogether.

So here is the receipt, up front. One concrete team workload — five engineers, one always-on web service in the 1 vCPU / 1 GB class, managed Postgres with 20 GB of storage, five ephemeral preview environments, 500 GB of North America/Europe egress a month — priced in July 2026 terms on all three platforms, plus the same workload as owned capacity:

Three things fall out of that table before any prose. First, the per-seat line is the single biggest swing on any team bill — bigger than compute, bigger than egress. Second, Fly.io is roughly 5–7× cheaper than Railway for this workload, and the gap widens as traffic grows. Third, owned hardware is an order of magnitude cheaper than all three — and the reason teams don't all leave is in the rows the table can't price: who runs Postgres backups, who wires preview environments, who gets paged. The rest of this post prices each claim in the split verdict, then prices the exit.

Render's per-seat removal: $70 a month back, then the flat-rate bill begins​

On April 23, 2026, Render introduced new workspace and organization plans that removed seat fees, added self-serve compliance features, and revised billing for outbound bandwidth and custom domains. Before that date, the math for our five-person team was $19 per member per month — $95 in seats before a single container ran. After it, the same team pays a flat $25 Pro workspace fee with unlimited members. That is $70 a month returned on the seat line alone, and it is the entire reason "Render wins for teams" became the consensus in 2026 comparisons.

Where Render still costs more is everything below the seat line. The flat-rate model that makes bills predictable makes them lumpy: a Standard web service is $25 whether it idles overnight or not, each additional service (worker, cron, private API) is another flat instance charge, and Postgres storage at $0.30/GB/month sits outside the $6 Basic database price rather than bundled in. Preview environments are full infrastructure copies with a real database — excellent fidelity for staging a migration, billed per service while they live. For a team running one service and one database, Render's post-April pricing is the cheapest managed seat-included option in the table above. For a team running eight small services, the flat per-service rates re-add what the seat removal took away, and that is the line to audit before migrating on the strength of the headline.

The honest read: Render sells predictability, and since April 2026 it sells it without a per-head tax. If your trauma is surprise usage bills, that is worth real money. If your architecture is many tiny services, count them before you switch.

Railway's PR environments: the DX lead is real, and it is metered​

Railway's claim in the split verdict is fastest developer experience, and its most concrete evidence is preview environments: every pull request gets an auto-created, auto-deleted copy of the stack, with persistent staging environments alongside, and nothing in the workflow asks the developer to think about it. Render matches the capability with heavier fidelity — full infrastructure copies including the database — at full per-service cost. Fly.io leaves it to you: preview VMs wired through GitHub Actions, $0 in platform charges, paid for in YAML and maintenance instead.

Three caveats ride along with Railway's lead. First, the meter runs on everything the automation spins up: compute bills per vCPU-second (about $0.00000772) and per GB-second (about $0.00000386), and egress at $0.05/GB is the highest of the three platforms — the same 500 GB that costs $10 on Fly.io costs $25 here, and at 2 TB it costs $100. Second, there is no ongoing free tier to experiment inside: the old free plan died in 2023, new accounts get a one-time $5 trial credit, and then the Hobby plan is $5 a month with usage on top. Third, the managed-Postgres gap that long favored Render is closing but not closed — Railway shipped experimental HA Postgres in March 2026, while Render's point-in-time recovery remains the safer story for customer data today.

So the DX verdict holds with an asterisk: Railway is the fastest path from pull request to reviewed preview, and the meter charges you for exactly that speed. Teams with chatty preview traffic or egress-heavy workloads should price the convenience at their own volumes, not the benchmark's.

Fly.io's glue-work tax: the cheapest bill with the longest setup​

Fly.io's side of the verdict cuts both ways: most glue work for team workflows, cheapest bill at scale. Both halves survive contact with the numbers. Billing is per-component — Machines by the VM-second (a small shared-CPU VM with 256 MB runs about $2.02 a month always-on), volumes at $0.15/GB/month, IPv4 addresses, and egress in three regional tiers: $0.02/GB in North America and Europe, $0.04 in Asia Pacific, Oceania, and South America, $0.12 in Africa and India. There is no seat fee and no per-service flat rate, which is why the same workload lands around $25–35 a month — the lowest managed total in the table.

The tax is operational. Deployment is Dockerfile-and-fly.toml rather than git-push-and-done; staging and preview environments are GitHub Actions workflows you write and own; there is no managed Postgres in the Render sense — databases are Machines you operate, with backups and failover as your configuration, not a plan feature. And the free tier is gone for new organizations: where Railway still offers a $5 trial credit and Render keeps a (cold-starting) free tier, Fly.io's old monthly allowance is history and new orgs get only a short trial before a card goes on file.

That is why the "cheapest at scale" and "most glue work" claims are the same claim. Fly.io's $0.02 NA/EU egress undercuts Railway's $0.05 by 60%, so every terabyte of growth widens its lead — but the lead accrues to teams with the platform time to spend it. A solo founder who never touches fly.toml keeps more of the savings than a five-person team wiring previews, Postgres standby, and multi-region deploys by hand. Price the bill, then price your week.

The exit column: what ~€16 of owned hardware actually buys​

The fourth column in the opening table is the one the benchmarks never include. Two shared-vCPU Hetzner boxes in the CPX22 class (2 vCPU, 4 GB each, roughly €8 a month apiece) run the web service, Postgres, and all five preview environments with room to spare, and the 20 TB of included traffic swallows 500 GB of egress without a meter twitching. No seats, no per-GB-RAM line, no per-vCPU line: about €16 a month flat against ~$195 on Railway.

Sensitivity first, because a single point flatters every argument. Push egress to 2 TB and Railway adds $75 while Fly.io adds $30 and owned hardware adds nothing — the ranking doesn't change, the gaps do. Shift traffic to Asia Pacific and Fly.io's edge narrows to $0.04 against Railway's flat $0.05 while Render's flat-rate structure barely notices. Scale to twenty concurrent previews and Render's full-copy billing scales worst in dollars, Railway's meter scales linearly, and Fly.io's DIY stays near $0 plus your Actions minutes. The exit column wins every sensitivity run on dollars. It loses on the rows no benchmark prices: Postgres point-in-time recovery you configure, preview environments you automate, kernel patches and 3 a.m. pages you absorb.

That asymmetry is the whole migration decision in one sentence. The platforms charge a large multiple of hardware cost to convert ops time into a bill; the exit converts the bill back into ops time. Teams should make that trade with open eyes — and notice that the most expensive line in the managed table isn't compute or egress, it's seats, which is pure pricing policy, not physics. Any platform that removes the seat meter without adding an ops burden wins on both axes at once.

Which team picks what​

For the prototype stage — one developer, iterating daily, no customer data — Railway's usage billing and instant previews remain the best value despite the highest egress rate, because at low volumes the meter barely moves. For the production-SaaS stage with a real team, post-April-2026 Render is the default answer: flat workspace pricing, managed Postgres with point-in-time recovery, preview environments with real databases, and a bill you can forecast to the dollar. For the scale-and-edge stage — high egress, multi-region latency requirements, Docker-native workflows, and someone who enjoys owning fly.toml — Fly.io's per-component pricing is structurally cheaper and pulls away with every terabyte. And for the team with platform time and sovereignty or cost reasons to own the machines, the exit math is overwhelming on dollars and honest about the ops cost — which is exactly the gap a self-hosted, Render-compatible PaaS exists to close: the hardware bill with the managed workflow.

Heroku taught a generation that git-push deploys are worth paying for. The 2026 lesson is narrower: pay for the workflow, not the seats, not the egress meter, and not the per-service flat rates you didn't audit. Run the table above with your own traffic, count your services, and pick the winner — or the exit — with the receipt in hand.

Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. Star the repo on GitHub or deploy your first app today.