Supabase

What Supabase actually is

Supabase is a managed PostgreSQL platform with the backend services most applications need bolted on: authentication with row-level security, auto-generated REST and GraphQL-style APIs over your schema, realtime change streams, file storage with an S3-compatible API, edge functions, vector search, and database branching. The pitch is Firebase's developer experience on top of a real relational database — you get instant provisioning and client libraries, but underneath it is Postgres you can connect to directly, extend with extensions, and export whenever you like.

The company was founded in 2020 by Paul Copplestone and Ant Wilson, went through Y Combinator's Summer 2020 batch, and operates remote-first with roughly 120 people. Each project is an isolated Postgres cluster, so scaling one app never touches another. The codebase is open source, and that matters twice here: you can inspect what the platform does, and you can self-host the stack if the managed service ever stops fitting. In our view, that combination — a managed platform with a genuine exit path — is the core reason Supabase has become the default backend for a generation of new applications.

Pricing: what you really pay

Four plans, and the pricing page is one of the more complete ones in this industry. Free is $0/month: 50,000 monthly active users, a 500MB database on shared CPU with 500MB RAM, 5GB egress plus 5GB cached egress, 1GB of file storage, unlimited API requests, community support. The catches are real but reasonable — free projects are paused after a week of inactivity, there is a limit of two active projects, and no automatic backups.

Pro is from $25/month per organization: 100,000 monthly active users (then $0.00325 per MAU), 8GB of disk per project (then $0.125 per GB), 250GB egress (then $0.09 per GB), 100GB file storage, email support, daily backups kept 7 days, and $10/month in compute credits — which covers one Micro compute instance ($10/month, shared compute, 1GB RAM). Additional projects each carry their own compute cost; Supabase's own example shows a Pro org with two Micro projects billing $35/month after credits. Compute is billed hourly per instance, and the ladder runs from Micro at $10 to a dedicated 16XL at $3,730/month (64 vCPUs, 256GB RAM), with Large (2 dedicated vCPUs, 8GB) at $110.

Team is from $599/month: SOC 2 and ISO 27001 reports, SSO for the dashboard, project-scoped and read-only access, priority email support with SLAs, 14-day backup retention, 28-day log retention, and HIPAA as a paid add-on. Enterprise is custom-priced with technical account management, uptime SLAs, AWS PrivateLink, and 24×7×365 support. Add-ons are metered across plans: point-in-time recovery at $100/month per 7 days of retention, custom domains at $10 per domain per month, database branching at $0.01344 per branch per hour, log drains at $60 per drain per month, SAML/SSO auth beyond 50 MAU at $0.015 per MAU.

Our read: the entry economics are excellent and the growth economics require attention. A production app on Pro at $25/month plus included compute is a bargain for managed Postgres with auth, storage, and APIs. But the add-on prices are where budgets drift — PITR at $100/month per 7 days of retention is serious money, and moving from Micro to a dedicated instance multiplies compute cost quickly. The pricing page is honest about all of it; you just have to model your usage instead of assuming the flat price holds.

Performance & infrastructure

Projects deploy to AWS regions — 17 specific ones at the time of writing, from North California, Oregon, North Virginia, Ohio, and Canada through Ireland, London, Paris, Frankfurt, Zurich, and Stockholm to Mumbai, Singapore, Tokyo, Seoul, Sydney, and São Paulo, plus general "Americas / Europe / APAC" groupings that let Supabase pick a region for you. The docs are explicit that region choice is a data-location control, not proof of regulatory compliance — for example the Europe grouping includes London and Zurich, which are outside the EU. Every project is its own Postgres cluster, with dedicated instances from the Large compute size up, direct and pooled connection limits per instance size, and disk tiers (general purpose up to 64TB with 3,000 IOPS included, or high-performance at $0.195 per GB with IOPS billed separately).

Reliability commitments are tier-scoped and worth stating precisely: the 99.9% monthly uptime commitment in Supabase's SLA applies to Enterprise customers only (per product, GA features only, with third-party vendor outages like AWS or Cloudflare excluded). Team and Enterprise customers get support SLAs; everyone else gets email or community support. Free-plan pausing is the other availability quirk — a project nobody touches for a week goes to sleep until you wake it.

We have not benchmarked Supabase ourselves, so no measured numbers from us; the platform's Postgres underpinnings mean performance is mostly a function of the compute instance and disk tier you buy. For the server-level measurements we have published, see benchmarks. What we would say editorially: the pooled connection architecture (200 pooler connections even on Micro) handles typical web-app traffic patterns well, and the places teams get surprised are large joins on shared compute and egress-heavy workloads — both are sizing questions, not platform defects.

Where it falls short

Bills that grow legs. Organization-based billing plus per-project compute plus usage overages plus add-ons is a lot of moving parts, and the step from "$25/month" to a dedicated compute instance, PITR, and a log drain can triple your spend without any change in how the product feels. The team publishes the numbers clearly — this is transparency, not concealment — but the model rewards people who read pricing pages for fun and punishes those who do not.

Support depth is the second constraint. Real SLAs — uptime and support — sit on Team ($599/month) and Enterprise. On Free and Pro you are in community and email queues during exactly the incidents when you want a human. For a startup that is acceptable; for a business with customers on the line it is the main reason to evaluate the higher tiers early rather than after an outage.

Finally, the usual BaaS trade-offs. Portability is better than most (it is Postgres), but the auth, realtime, storage, and function layers are Supabase-specific; migrating off means re-implementing those services somewhere else. And while the self-host option exists as an escape hatch, running the full Supabase stack yourself is a project, not a button. None of this is unique to Supabase — it is what "backend as a service" means — but it belongs in the decision.

Who should pick this

Our pick for teams building web and mobile applications on a relational model: startups shipping an MVP this quarter, solo developers who want auth and storage without writing infrastructure, and any team that has been burned by a NoSQL schema wall and wants SQL, migrations, and row-level security from day one. Within our database hosting and serverless platform categories it is the option we find ourselves recommending most often, and the free tier is genuinely enough to build and launch a first version.

We would look at plain managed Postgres (or our benchmarks hosts) when you want infrastructure control and predictable pricing more than bundled services, and at Firebase-style competitors when you need their specific ecosystem. For regulated workloads, the Team and Enterprise tiers carry the compliance reports and SLAs — price those in from the start. And if you are comparing against running Postgres yourself, remember what the $25/month actually replaces: connection pooling, backups, auth, storage, API generation, and the on-call rotation that keeps them all alive.

Key takeaways

• Supabase (founded 2020, YC S20, remote-first) bundles managed Postgres with auth, realtime, S3-compatible storage, APIs, and branching — open source with a self-host path.
• Free tier: 50,000 MAU, 500MB database, 5GB egress, 1GB storage, two active projects, paused after a week of inactivity, no backups.
• Pro is $25/month with $10 compute credits (one Micro instance included), 8GB disk, 250GB egress, 100GB storage, 7-day backups; Team is $599/month with SSO, SOC 2/ISO 27001, and support SLAs.
• Compute runs Micro ($10, shared, 1GB RAM) up to 16XL ($3,730/month, 64 vCPU) and is billed hourly per project — add-ons like PITR ($100/month per 7 days) can dominate the bill.
• 17 AWS regions across the Americas, Europe, and Asia-Pacific; region choice is a data-location control, not compliance proof.
• The 99.9% uptime SLA is Enterprise-only; Free and Pro support is community and email.

Sources

• https://supabase.com/pricing
• https://supabase.com/docs/guides/platform/regions
• https://supabase.com/docs/company/sla
• https://supabase.com/support-policy
• https://supabase.com/docs/guides/storage/s3/compatibility
• https://www.ycombinator.com/companies/supabase