Qovery's September 2026 Price Table Says Render $7, Railway $5, Fly.io $2.02 — Here's the Second-Year Bill It Hides

The cheapest PaaS on the comparison table is never the cheapest PaaS on your second invoice. Qovery's September 2026 Heroku-alternatives roundup lists Render from $7/month, Railway from $5/month, and Fly.io from $2.02/month — and every one of those numbers is technically true and practically unreachable for a production app with a database, a worker, and real traffic. The honest question is not what the entry tier costs, but what the same small production workload costs in month twelve, multiplied by twelve.

Here is that answer up front, for one typical app (1 vCPU/2 GB web service, 0.5 vCPU/1 GB worker, 10 GB Postgres, 50 GB egress, preview environments for a small team). All figures are September 2026 list prices, per month in USD unless noted:

The ranking completely inverts. Fly.io's $2.02 entry is the lowest number on Qovery's table and its realistic bill is still the lowest of the three managed options — but it is 14x the entry price. Railway's $5 Hobby headline turns into an $84 run-rate on Pro once compute meters.

And the flat Hetzner box, which has no entry tier to advertise, undercuts all three by month twelve because its meter is the machine, not the request. The rest of this post shows the line items, the four meters that do the damage, and where the ranking flips.

What Qovery's table actually says​

The source is Qovery's "5 PaaS Platforms Like Heroku, Compared" (qovery.com/blog/paas-platforms-like-heroku-compared, prices stated as pulled from each vendor's own page as of September 2026; accessed September 2026). The entry prices it quotes check out against the vendor pages:

• Render: web services from $7/month, managed Postgres from $6/month. That $7 is the Starter tier (0.5 CPU / 512 MB). It is a fixed per-service price — every service you add multiplies it — and the vendor page is render.com/pricing.
• Railway: $5/month Hobby, $20/month Pro, then usage at roughly $20 per vCPU and $10 per GB of RAM per month. The subscription is a minimum commitment that includes the same dollar amount of usage credit ($5 includes $5, $20 includes $20). Compute, memory, storage, and egress meter on top. Vendor page: railway.com/pricing.
• Fly.io: from $2.02/month per Machine for shared-cpu-1x with 256 MB RAM. That is one tiny always-on VM with no volume, no second region, and no database. Everything scales per second from there. Vendor page: fly.io/docs/about/pricing.

Each entry number implies a different cost model — fixed per service, usage-based pool, per-Machine meter — and that model matters more than the number. A fixed $7 that multiplies per service behaves nothing like a $2.02 base with per-GB RAM and per-GB egress meters attached. Entry-price tables flatten three different billing geometries into one column, which is exactly why they mislead.

One more caveat before the math: I cross-checked the Qovery-quoted figures against the vendor pages and the September 2026 third-party benchmark at techsy.io (Railway vs Render vs Fly.io) rather than trusting any single table. Rates move — Render restructured workspace plans in April 2026, Railway's per-minute rates re-derive to the same monthly figures — so treat the line items below as list-price reconstruction, not a quote.

The four meters that eat year two​

Entry tiers cover an idle toy. Production apps pay four meters that entry tables either omit or footnote.

1. Compute scaling past the starter size. Nobody runs production on 512 MB. Render's step from $7 Starter to $25 Standard (1 CPU / 2 GB) is a 3.5x jump for one service, and workers bill as separate services. Railway charges about $20/vCPU/month plus $10/GB RAM/month, so a 1 vCPU/2 GB web service is ~$40/month of usage before anything else exists. Fly.io's shared-cpu-1x/256 MB at $2.02 grows roughly $5/GB/month of added RAM plus larger CPU sizes, so a 2 GB web Machine lands near $11/month. The pattern is uniform: the production shape costs 4–6x the advertised shape on every platform.

2. Managed Postgres plus storage. The database is the cruelest line item because no entry price includes a real one. Render's managed Postgres starts at $6/month on Basic, with storage billed separately at $0.30/GB/month — 10 GB adds $3, and that storage ratchets only upward. Railway has no separate managed-DB SKU; Postgres runs as a service from a template and consumes the same $20/vCPU and $10/GB meters, so a modest 0.5 vCPU/1 GB database is ~$20/month of usage plus $0.15/GB storage. Fly.io Postgres runs on Machines you size yourself: ~$6/month for a small shared node plus $0.15/GB volumes — cheapest of the three single-node, but high availability means a second node and doubles it. Budget $6–20/month for the database alone, before storage.

3. Egress. This is the meter with the widest spread and the most footnote energy. Fly.io splits egress into three regional tiers: $0.02/GB in North America and Europe, $0.04/GB in Asia-Pacific, Oceania, and South America, $0.12/GB in Africa. Railway charges a flat $0.05/GB. Render includes a bandwidth allowance on paid workspaces (5 GB on lower tiers, more higher up) and bills overage around $0.10–0.15/GB. At 50 GB/month in NA/EU that is $1 on Fly.io, $2.50 on Railway, and $0 on Render — but at 500 GB/month it becomes $10, $25, and a real overage line respectively. Egress is the variable most likely to flip the ranking, which is why the sensitivity section below exists.

4. Build minutes, preview environments, and seats. The collaboration tax. Render's Pro workspace is $25/month flat for full-stack previews and autoscaling (Hobby is $0 but limited), and build minutes beyond the included allowance bill per minute — auto-deploy on every push in a monorepo burns them fast. Railway's Pro minimum is $20/month per workspace with unlimited seats, which is genuinely team-friendly, but every preview service meters compute like production. Fly.io has no seat fee at all; you pay for the Machines your previews run on, and stopped Machines bill only disk at $0.15/GB/month. For a team of three shipping daily with ephemeral previews, expect $0–25/month here depending on vendor — small next to compute, but it is pure overhead over the entry price.

Worked example: one small production app, four bills​

Same workload, priced four ways. Web: 1 vCPU / 2 GB always-on. Worker: 0.5 vCPU / 1 GB always-on. Postgres: 10 GB with 50 GB egress in NA/EU, plus preview environments for a team of three. All at September 2026 list prices.

Render: ~$66/month, ~$792 year two.

Annualized entry price ($7 × 12) is $84. The realistic year-two bill is $792 — 9.4x the headline. The workspace fee alone ($300/year) exceeds the entire annualized entry price.

Railway Pro: ~$84/month, ~$1,008 year two.

Annualized Hobby entry ($5 × 12) is $60; annualized Pro minimum ($20 × 12) is $240. The realistic bill is $1,008/year — 4.2x the Pro minimum, 16.8x the Hobby headline. RAM is the silent killer: every GB costs $10/month whether it serves traffic or idles.

Fly.io: ~$28/month, ~$330 year two.

Annualized entry ($2.02 × 12) is $24.24. The realistic bill is ~$330/year — 13.6x the headline, yet still the cheapest managed total here. Two honest caveats: single-node Postgres (HA doubles the DB line) and no second region (each region multiplies Machines). Scale-to-zero with min_machines_running=0 can push idle services toward storage-only cost — the one genuine escape hatch in this comparison.

Hetzner flat box: ~$18/month, ~$216 year two.

One shared-vCPU cloud node in the 4 vCPU / 8 GB class (CPX31 line) runs roughly €15/month plus €0.50 for IPv4 — about $18/month all-in at current rates, with generous included traffic and the 10 GB database living on local disk. No per-service fee, no per-GB RAM meter, no workspace seat, no build-minute counter. The same workload that costs $792–$1,008/year on Render or Railway costs ~$216/year on the box.

The honest tradeoff is operations: you run Postgres, backups, TLS renewal, and node upgrades yourself — or you run an open-source PaaS on top of the box and pay in setup time instead of rent. What the flat price buys is cost certainty: adding a staging copy of the app costs nearly $0 in marginal infra, where on Railway it meters a second copy of every resource.

Where the ranking flips​

One workload does not make a rule. The value of the worked example is showing which variable moves the winner:

• Low-traffic side project (scale-to-zero friendly): Fly.io wins outright. Stopped Machines bill only disk, so an app that idles 20 hours a day can land under $5/month. Render's fixed $7/service floor and Railway's $5–20 minimums cannot follow idle to zero.
• High-egress app (500 GB/month, image or API heavy): Fly.io still leads ($10 egress), Railway adds $25, and Render's overage line appears — but the real winner is the flat box, whose included traffic allowance absorbs it. This is the workload where metered egress hurts most and where "from $2.02" marketing is most misleading.
• RAM-heavy workload (caches, JVMs, ML-adjacent): Railway punishes this hardest at $10/GB/month — a 16 GB cache tier alone is $160/month of usage. Render's fixed tiers ($25–85) suddenly look reasonable, and the flat box looks best of all because RAM is already paid for.
• Team of ten shipping constantly: Render's $25 Pro workspace is flat regardless of headcount, Railway Pro is $20/workspace with unlimited seats, and Fly.io charges $0 for seats. All three are team-friendly in different ways — the seat-model horror stories belong to per-seat SaaS, not this trio. The cost driver here is preview-environment compute, which meters on Railway, multiplies fixed services on Render, and spins extra Machines on Fly.io.

The general rule: entry price predicts the bill only while the app fits the entry shape. The moment production adds a database, a worker, and traffic, the meter geometry — fixed per service, per-GB RAM, per-Machine, or flat box — decides the total, and the geometries rank differently for every workload shape.

What to do Monday morning​

If you are choosing a PaaS this week, skip the pricing hero section and model month twelve instead. Write down your web/worker/DB shape, your storage growth, your egress region mix, and whether you need previews — then price that, not the starter tier. For a side project, Fly.io's scale-to-zero is genuinely the cheapest managed path. For a steady small production app, budget 9–17x the entry headline on Render or Railway, or price one flat Hetzner-class box and decide whether the ops tradeoff is worth the 3–4x savings.

The deeper lesson is structural. Entry-price tables optimize for the click, and every vendor cooperates because every vendor's headline is technically true. The number that matters for planning — the annualized second-year run-rate for your actual workload shape — is never on the table. Build your own ledger before you commit, and re-run it every year: rates move, workspace plans restructure, and the cheapest platform for last year's shape may not be the cheapest for this year's.

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.