- The five categories of Postgres hosting and which workloads each one fits
- How pricing models differ: usage meters, flat plans, instance-hours, and 'free' hardware
- What each option handles for you versus leaves on your desk - backups, failover, upgrades
- A provider-by-provider comparison table with entry prices and catches
- The decision rules that pick a host in five minutes
PostgreSQL hosting is a solved problem with too many solutions. The engine is identical everywhere - that is the whole appeal of Postgres - so what you are really choosing between is pricing models and operational splits: who takes the backups, who answers the 3am page, and what shape of bill you get. This guide compares the real options by use case, because 'the best Postgres host' is a property of your workload, not a brand.
The five categories
Serverless Postgres (Neon). Compute scales with load and suspends when idle. Brilliant for dev/preview/spiky workloads; expensive left running 24/7.
Backend-in-a-box (Supabase). Postgres plus auth, storage, realtime and APIs. You are adopting a platform, not just a database.
Flat-priced managed Postgres (Swyftstack, DigitalOcean). A provisioned database with backups and pooling at a fixed monthly price. The predictable-bill option.
Hyperscalers (AWS RDS, Cloud SQL, Azure). Postgres inside the cloud's own operational model - VPCs, IAM, compliance. Powerful, and priced across five meters.
Self-hosted (your VPS). Free software on hardware you already rent. Every operational task is yours.
Provider comparison
Provider | Model | Entry price | Backups | The catch |
|---|---|---|---|---|
Neon | Serverless, usage-based | $0 free tier; ~$0.106/CU-hr paid | History-based restore | Always-on compute gets expensive; cold starts |
Supabase | Platform, plan + meters | $0 free tier; $25/mo Pro floor | Daily (Pro) | Per-project compute add-ons; platform lock-in gravity |
Swyftstack | Flat-priced managed | $19/mo (14-day free trial) | Automated, included | No free tier; smaller ecosystem than AWS |
DigitalOcean | Flat-priced managed | $15/mo (1 GB / 10 GB) | Daily, included | Entry node is small; HA doubles the price |
AWS RDS | Instance + 4 more meters | ~$12/mo (t4g.micro, single-AZ) | Included to 100% of DB size | Multi-AZ doubles cost; VPC/IAM homework |
Railway | Usage-based platform | ~$5 credit, then metered | Manual/limited | Compute meters run while the app idles |
Self-hosted VPS | DIY on rented hardware | $0 extra on an existing box | Yours to build and test | Upgrades, security, failover, 3am - all yours |
Full disclosure: Swyftstack is our product. We put it in the table on the same terms as everyone else, and the guidance below tells you honestly when the others are the better pick.
What you are actually buying
Hosting Postgres involves the same work everywhere: provisioning, connection management, backups that verifiably restore, minor-version patching, major-version upgrades, monitoring, and failover. A managed provider does not make that work disappear - it moves the work to people who do it hundreds of times a day. When you compare a $6 VPS against a $19 managed plan, the $13 difference is the market price of those tasks. Decide what your hour costs before deciding that is expensive.
Recommendations by use case
Side projects and prototypes
Start free: Neon or Supabase, whose free tiers are genuinely useful (we compared every option in our free PostgreSQL hosting guide). Export a dump weekly if the data matters. Upgrade the day real users arrive.
A steady production app
This is the flat-price sweet spot. An always-on database with predictable traffic gains nothing from usage meters - it just makes the bill a variable. Swyftstack ($19/month, backups and pooling included) and DigitalOcean ($15/month entry) are the two cleanest options; the difference is that our plan includes object storage and app hosting in the same flat price, while DO composes them as separate line items.
AWS-native (or compliance-bound) teams
If your stack already lives in a VPC and the security team speaks IAM, RDS is the correct answer and the ecosystem is the product. Budget with all five meters - we broke them down in our RDS pricing teardown - and expect the real number to be 2-4x the instance price.
Spiky, bursty, or preview-heavy workloads
Neon's model was built for this: previews per PR, dev databases that sleep 20 hours a day, demo environments. Our Neon pricing breakdown has the worked examples; the short version is that suspending compute is the entire economic argument, so make sure nothing keeps it awake.
You want the whole backend, not just Postgres
Supabase. Auth, storage, and realtime out of one console is a real productivity win for solo builders - see our Supabase pricing analysis for where the $25 floor actually lands at scale, and Neon vs Supabase for how the two philosophies differ.
You enjoy operating databases
Then self-host - genuinely. Postgres in Docker on a VPS you already rent is free, educational, and completely yours. Just write the backup script before the first real row of data, and test the restore before you trust it.
What about Heroku, Railway and Render?
Three names people expect in this list, and why they sit outside the main table. Heroku Postgres still works, but its pricing has aged badly against every option above - the entry paid tiers cost more and include less, and the free tier is long gone. Railway and Render both host Postgres competently as part of hosting your whole app; the database inherits the platform's billing model (usage meters on Railway, per-service pricing with a 30-day free-database limit on Render). If your app already lives on one of them, keeping the database there is defensible for simplicity - just compare what the same money buys from a database-first provider once the bill passes $20 - 30/month. We compared the two platforms directly in Railway vs Render and unpacked Railway's meters in Railway pricing explained.
Questions that decide it faster than benchmarks
Benchmarks between managed Postgres providers mostly measure the instance size you paid for. These four questions separate providers faster: When did the provider last lose someone's data, and what did they publish about it? Is the backup restore path documented and testable by you, today? What happens at 2x your current scale - a bigger plan, or a pricing model change? And can you get a human when something breaks? The answers differ far more than query latency does.
The five-minute decision
Prototype? Free tier (Neon/Supabase). Done.
Always-on production app, want a predictable bill? Flat-priced managed.
Already deep in one cloud? That cloud's managed Postgres.
Compute idles most of the day? Serverless.
Need auth + storage + realtime too? Supabase - or a flat plan that bundles storage.
None of the above and curious? Self-host, with backups from day one.
Whatever you pick, Postgres itself keeps the exit door open:
pg_dump "$OLD_URL" | psql "$NEW_URL"
Prices quoted here were checked in July 2026. Vendors change them; treat the structure as the durable part and re-check the current numbers before you budget.
Frequently asked questions
What is the best PostgreSQL hosting provider?
For steady production apps, a flat-priced managed provider; for spiky or dev workloads, Neon's serverless model; for AWS-native stacks, RDS; for a full backend-in-a-box, Supabase. 'Best' is a property of your workload, not of any vendor.
How much does PostgreSQL hosting cost?
Free tiers cover prototypes (0.5 - 5 GB). Real production hosting starts around $15 - 25/month flat, RDS lands at $30 - 140+ once you add Multi-AZ and storage, and usage-based platforms bill anywhere between pocket change and hundreds depending on compute hours.
Is managed Postgres worth it versus self-hosting?
Managed hosting is buying back the hours you would spend on backups, upgrades, monitoring and failover - plus the 3am pages. If a $19 - 50/month plan saves even one hour a month, it pays for itself for most working developers.
Can I move between Postgres hosts later?
Yes, and this is Postgres's superpower: pg_dump/pg_restore or logical replication work between any two providers on compatible versions. Avoid provider-proprietary extensions and the exit door stays open.