Dora NodaSomebody is willing to pay your switching costs, and that sentence should make you read the offer twice. Render's migration program offers up to $10,000 in enterprise migration credits for moving production infrastructure onto its platform, paired with official migrate-from-Heroku and migrate-from-Railway guides. The verdict worth stating up front: the credits cover the meter, never the migration — for a typical small production topology they buy the better part of a decade of compute, yet they pay for zero hours of the engineering work that actually decides whether the move succeeds. The rest of this post earns that sentence with a worked burn-down.
| Monthly Render burn | Example topology | What $10K buys |
|---|---|---|
| ~$48/mo | 1 Standard web ($25) + 1 Starter worker ($7) + Postgres Basic ($6) + Key Value Starter ($10) | ~17 years |
| ~$117/mo | 2 Standard web ($50) + 1 Standard worker ($25) + Postgres 1 GB ($32) + Key Value ($10) | ~7 years |
| ~$355/mo | 2 Pro web ($170) + 1 Pro worker ($85) + larger managed data (~$100) | ~28 months |
Bottom line: if your topology fits in the left two rows, the credit is effectively infinite runway and the decision is about the platform, not the price. If you sit in the bottom row, the credit is a two-year runway with a meter waiting at the end of it. Either way, the money covers compute burn-down — not a single hour of migration labor — so read on for what the move really costs before you apply.
What Render is actually offering, and why now
The offer lives on Render's migration-credits page: move production infrastructure to Render, get up to $10,000 in enterprise migration credits. Two words do heavy lifting there. "Up to" means the grant scales with the deal — a side project is not receiving ten grand. "Enterprise" means the program targets teams with production spend worth competing for, not hobbyists chasing free tiers. Alongside the credits sit two official guides — migrate-from-Heroku and migrate-from-Railway — plus a feature-by-feature Railway comparison that reads like a sales document because it is one.
The timing is the tell. Railway's pricing history runs in one direction: from deploy-anything-free in 2020 to a 30-day $5 trial, then a $5/month Hobby minimum with $5 of included usage credit, then $20/month per workspace for Pro with everything else metered — compute, memory, egress. In June 2025 Railway even throttled and paused Trial and Hobby deployments during a demand surge to protect Pro customers.
Meanwhile both vendors now publish step-by-step guides for migrating off each other — Railway's docs walk you from Render to Railway, Render's docs walk you back. When two direct competitors each maintain a guide for defecting from the other and one of them starts subsidizing the exit, the post-Heroku-sustaining-mode market has gotten genuinely contested. Credits are what a vendor offers when the product comparison alone stopped closing deals.
None of this makes the offer a trick. Credits that burn down against real compute are real money, and Render's per-second billing means the grant stretches exactly as far as the table above says. But "we pay your switching costs" is doing rhetorical work the fine print can't support, because compute was never the expensive part of switching.
The burn-down, worked honestly
Take the three topologies from the table and price them from documented 2026 numbers. Render web and worker instances run Starter at $7/month (512 MB / 0.5 CPU), Standard at $25 (2 GB / 1 CPU), Pro at $85 (4 GB / 2 CPU), Pro Plus at $175, up to Pro Ultra at $450 — billed per second, so these are steady-state ceilings. Managed Postgres starts at $6/month for the Basic 256 MB tier and around $32 for a 1 GB Standard instance; Key Value (Redis-compatible) starts at $10.
Railway's side of the ledger is structurally different: a $5 or $20 subscription plus usage meters, with databases as unmanaged Docker containers you maintain yourself rather than managed services.
Two observations fall out of the arithmetic. First, the smaller you are, the more absurd the credit looks — $10K against a $48/month burn is seventeen years, which is another way of saying small teams are cheap to acquire and the credit costs Render almost nothing until you grow.
Second, the credit is linear in the one dimension that matters least at migration time: it pays the destination meter while you are still paying the origin meter. Every serious migration double-runs for weeks — old stack serving traffic, new stack soaking, data replicating between them — and the credit covers Render's half of that overlap while Railway's final invoices land in full. That is fine, but it is not "switching costs." It is a discount on the destination.
The honest version of the offer reads: free compute runway, sized to your deal, burning down from day one. Worth taking on those terms. Just don't mistake it for somebody else funding your migration.
What the $10K never touches
Real switching costs are measured in engineering weeks, and the credit denominates in compute dollars. Walk the actual checklist from Render's own migrate-from-Railway guide and price each line in labor:
- Configuration mapping. Railway splits config across service variables, shared variables, and
${{ }}reference variables; Render wants environment variables and environment groups. Somebody reads every service, re-maps every reference, and re-verifies every secret. Half a day if you are disciplined, two days if you find the variable nobody documented. - Data migration, unmanaged to managed. Railway databases are Docker containers on volumes; Render Postgres is a managed service with point-in-time recovery, read replicas, and high availability. That upgrade in operational quality is genuine — it is also a
pg_dump, a restore window, a connection-string rotation, and a rollback plan if the restore misbehaves. Budget a day, plus the anxiety. - Preview and deployment semantics. Railway environments map to Render preview environments, but autoscaling bands, health-check paths, and zero-downtime deploy behavior all need re-validation per service. Render does not pause deployments under load the way Railway did to Hobby users in 2025 — an improvement you still have to prove to yourself service by service.
- The cutover itself. DNS moves, traffic shifts, old-project teardown, and the week of watching dashboards you scheduled nothing else during. Nobody's credits cover an on-call rotation.
A fair estimate for a five-service stack with one database: one to three engineering weeks end to end. At any plausible fully-loaded salary, that dwarfs the first year of the compute delta the credits erase. The credit subsidizes the cheapest line item on the migration invoice and leaves the dearest one — your team's attention — entirely on you.
The self-hosted subtraction
There is a third quote on the table that neither vendor's guide mentions. The small topology above — a web service, a worker, Postgres, Redis — fits comfortably on a single Hetzner CX22 (2 vCPU, 4 GB RAM, roughly €4/month) with room to spare, or a CPX31 (4 vCPU, 8 GB, roughly €14/month) if you want headroom. The mid topology fits on two CPX31s for under €30/month total. Over three years, the small topology costs roughly $1,700 on Render against roughly $150–$500 on owned Hetzner hardware — a gap the $10K credit erases for years, then hands back in full the month it exhausts.
Self-hosting re-assumes everything the credit conversation politely skips: you become the managed-Postgres team (backups, point-in-time recovery, failover), the preview-environment team, the on-call rotation. A Cluster-API-managed fleet makes that burden declarative — machines reconciled from Git rather than clicked into existence — but it does not make it zero. The honest framing is a three-way trade, not a two-way one: Railway's usage meter with unmanaged data, Render's per-instance meter with managed data plus a credit runway, or a flat hardware bill with you as the platform team. Anyone telling you one of the three has no costs is selling you one of the three.
Who should take the credits, and who should walk past
Take them if you already decided to leave Railway on the merits — managed Postgres with real backups, native log drains, no deployment pausing, per-second billing you can model. The credits then do exactly what runway should do: remove price from the decision for long enough that the platform has to win on its own. A two-year runway at Pro-tier burn is a genuine gift; spend it verifying the operational claims, not celebrating the discount.
Walk past if the credit is the reason for the move rather than the sweetener. A team with no complaint about Railway that re-platforms for $10K in compute is trading one to three weeks of engineering focus — plus a second migration's worth of lock-in learning — for money that was never the constraint. And if your topology already fits on one Hetzner box with nothing exotic in it, price that option first: the cheapest migration is the one to infrastructure whose meter can never resume, because there isn't one.
The deepest signal in this story isn't the dollar figure. It is that hosted PaaS competition in late 2026 runs on switching-cost subsidies — guides both directions, credits for the jump. That is a buyer's market, and buyers in a buyer's market should do what buyers do: take the subsidy, keep the exit portable, and make every platform re-earn the workload every year. Render-compatible APIs, infrastructure as code, and data you can pg_dump on demand are what keep a credit a gift rather than a leash.
Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. If the credit math above has you pricing the third option, star the repo on GitHub or deploy your first app today.
