Postgres Hosting in 2026: RDS vs Supabase vs PlanetScale Cost Comparison
Choosing the wrong database host costs you either money or engineering time, sometimes both. This guide compares AWS RDS, Supabase, and PlanetScale at real scale points with actual 2026 list prices, total cost of ownership breakdowns, and clear decision rules. It is written for founders, CTOs, and engineers who are about to pick a database and need to understand what the bill looks like twelve months from now, not just at launch.
Why Database Hosting Costs Are Harder to Compare Than They Look
Every managed database vendor publishes a headline price. Almost none of them tell you what the actual monthly bill looks like once you add storage, backups, bandwidth, high availability, and connection pooling. A $25/mo plan can turn into $180/mo in production. A $143/mo RDS instance can balloon past $400/mo when you add Multi-AZ, storage snapshots, and data transfer out.
The comparison problem is structural. AWS RDS bills by compute instance, storage IOPS, backup retention, and data transfer, each as a separate line item. Supabase bundles compute, auth, storage, and realtime into a platform fee but still charges per gigabyte overages once you exceed plan limits. PlanetScale charges per row read and row write, which maps cleanly to query patterns but is nearly impossible to estimate without production traffic data.
Before you pick a host, you need answers to four questions: How many gigabytes of data will you store in 12 months? What is your read-to-write ratio? Do you need high availability from day one? And who on your team will handle database operations when something breaks? The answers push you toward different providers, often decisively. This article walks through each provider with 2026 list pricing and gives you the math to build your own estimate.
AWS RDS: The Default Choice and Its Real Price
RDS remains the default choice for teams that want maximum compatibility, deep AWS integration, and predictable instance-based pricing. The entry point is a db.t4g.micro at $11.68/mo on-demand in us-east-1. That instance is fine for development and low-traffic staging. It is not fine for production: it shares CPU burst capacity with other workloads and has no path to zero-downtime failover.
A realistic production starting point is a db.m5.large at roughly $143/mo on-demand. Add Multi-AZ for automatic failover and you roughly double the compute cost. Over a year that Multi-AZ markup runs to about $3,432, before storage. gp3 storage starts at $0.115/GB/month, which is cheap, but IOPS above the baseline 3,000 cost extra and snapshots are billed separately.
The single biggest lever on RDS cost is reserved instances. A one-year reserved db.m5.large with no upfront payment saves roughly 30 percent versus on-demand. A three-year all-upfront reservation saves close to 60 percent. For a database you know you will run for two years, that is real money: a $143/mo on-demand instance becomes roughly $57 to $100/mo reserved, depending on term and payment option.
The operational cost you cannot escape on RDS is time. Parameter groups, IOPS tuning, connection limits, and major version upgrades all require engineering attention. Budget 4 to 8 engineering hours per quarter for a single RDS cluster if you have no dedicated DBA.
Supabase: What the Pro Plan Actually Costs in Production
Supabase Pro is $25/mo but that number almost never appears on a real bill without additions. The plan includes 8GB of database storage, 250GB of bandwidth, and 100K monthly active users. Every number above those limits triggers per-unit overages: $0.125/GB for database storage past 8GB, $0.09/GB for bandwidth past 250GB, and $0.00325 per MAU past 100K.
Compute is the variable that surprises most teams. The default Pro plan gives you a shared compute instance comparable to a small cloud VM. If your app needs dedicated compute, you upgrade to a larger compute add-on. A Large compute add-on runs roughly $100/mo on top of the base $25. Supabase offsets this with $10/mo in compute credits included with Pro, so your net compute cost starts at around $15/mo for the base tier.
For a small production app with 3GB of data, under 100GB of bandwidth, and moderate traffic, expect a real monthly bill of $35 to $75. For a higher-traffic app with 20GB of data, 400GB of bandwidth, and the Large compute add-on, you are looking at $100 to $200/mo.
One critical operational note: Supabase free-tier projects pause after 7 days of inactivity. This catches every developer who spins up a free project for a demo and forgets about it. On Pro that behavior does not apply, but the free tier is genuinely unsuitable for any public-facing endpoint.
Supabase Team at $599/mo removes per-member pricing, adds SOC 2 controls, and includes daily backups with PITR. That is the tier where regulated workloads or enterprise customers belong.
PlanetScale: Row-Based Pricing and Who It Favors
PlanetScale runs on Vitess, the MySQL sharding layer that YouTube built and open-sourced. It is not Postgres. If your stack requires Postgres-native features (extensions, row-level security, JSONB operators, logical replication) PlanetScale is not a drop-in replacement and you should skip to the next section.
For teams on MySQL or willing to adopt it, PlanetScale's pricing model is fundamentally different: $1 per billion row reads, $1.50 per million row writes, and $2.50/GB of storage. There are no compute instances to size and no Multi-AZ checkbox. Branching, schema migrations without locks, and sharding are built in.
The math works in your favor when reads dominate. A read-heavy analytics dashboard hitting 500 million rows/month and writing 5 million rows pays $0.50 for reads plus $7.50 for writes, so $8 before storage. The same workload on RDS would cost $143/mo minimum just for the instance, even if it runs at 5 percent CPU.
The math turns against you when writes are heavy. An event-logging app writing 50 million rows per day hits 1.5 billion writes per month, which costs $2,250 on PlanetScale alone. On RDS that same write volume is just IOPS and storage. You need production query metrics before you can make this comparison honestly. If you do not have those metrics, start with RDS or Supabase and migrate later with real data in hand.
Storage Costs at Scale: The Number Most Comparisons Skip
Storage is where long-running databases diverge sharply on cost, and most comparison articles ignore it because the numbers look small at launch. They do not stay small.
As of 2026 list pricing: Supabase charges $2.50/GB for database storage past your plan's included amount. CockroachDB Serverless charges $1/GB. Google Cloud SQL (which underlies many managed Postgres offerings) charges $0.34/GB for standard SSD storage. AWS RDS gp3 runs $0.115/GB.
A database that grows to 500GB of storage pays $1,250/mo for storage alone on Supabase overages, $500 on CockroachDB, $170 on Cloud SQL, and $57.50 on RDS gp3. That is a 21x spread between the most and least expensive option at the same data volume.
This matters most for time-series data, audit logs, media metadata, and any application that retains historical records. If your data model involves high-volume append patterns, do the 12-month storage projection before you choose a host. A product that stores 1GB today but writes 100MB per day reaches 36GB in a year. That is a $87.50 monthly storage bill on Supabase overages, a $12.24 bill on RDS gp3. Over three years the difference funds a meaningful engineering hire.
Backup storage adds another layer. RDS includes automated backups up to your retention window at no extra charge, but backups beyond the free allocation cost $0.095/GB/month. Supabase Pro includes 7-day backups. Point-in-time recovery requires the Team plan.
High Availability: What You Actually Pay for Reliability
High availability means the database survives a single infrastructure failure without manual intervention. The cost of HA varies enormously across providers.
On RDS, enabling Multi-AZ deploys a synchronous standby replica in a different availability zone and promotes it automatically on failure. The cost is roughly double the compute price of your primary instance. On a db.m5.large, that is an additional $143/mo, adding up to $3,432/yr compared to single-AZ. Most small SaaS products are not HA on RDS in year one because founders cannot justify doubling their database cost before they have paying users.
Supabase Pro includes high availability at the compute add-on tier, but the exact SLA depends on your compute size. The base shared compute does not carry the same uptime guarantees as a dedicated instance. If your app has a paid SLA with customers, verify Supabase's current uptime guarantees for your compute tier before committing.
PlanetScale's architecture is inherently distributed. Multi-region reads require their Scaler Pro plan, which is quota-based pricing. Failover is handled at the platform level and you do not pay a separate HA surcharge the way you do on RDS.
The operational math for HA on RDS is: accept the extra $3,432/yr or accept that a database failure requires human intervention. For apps where 15 minutes of downtime loses significant revenue, pay for Multi-AZ. For internal tools and low-traffic apps in year one, single-AZ with a good backup strategy is often the right call.
Choosing by Stage: Hobby, Startup, and Scale
Database selection decisions that make sense at one stage become expensive at another. Here is a direct decision framework by stage.
Hobby and pre-revenue: Supabase Free is genuinely excellent for development and demos. The 7-day pause for inactive projects is the only meaningful constraint, and you work around it by keeping the project active or upgrading to Pro before launch. Do not run a live user-facing app on the free tier.
Early startup, under $10K MRR: Supabase Pro at $25 to $75/mo all-in is the right default. You get auth, storage, realtime, and a solid Postgres database in a single bill. The platform eliminates entire service categories you would otherwise build or integrate separately. The compute credit system means your first real compute tier costs $15/mo net. Start here unless you have a specific reason not to.
Growth stage, $10K to $100K MRR: Evaluate RDS if your team includes engineers comfortable with AWS. A reserved db.m5.large with Multi-AZ costs roughly $250 to $350/mo with storage, which is competitive with Supabase once you factor in large compute add-ons and bandwidth overages. The tradeoff is operational complexity: you own connection pooling, parameter tuning, and upgrade coordination.
Scale, 500GB-plus data or high write throughput: Re-examine storage unit economics. At 500GB, RDS gp3 storage costs $57.50/mo versus $1,250/mo in Supabase overages. That gap alone justifies migration engineering time. PlanetScale is worth a rigorous write-cost analysis if your workload is read-dominated and MySQL-compatible.
Total Cost of Ownership: Building the Real 12-Month Number
The list price of any database host is a floor, not a ceiling. Building a defensible 12-month TCO requires accounting for five cost categories: compute, storage, bandwidth, backups, and engineering time.
For a typical early-stage SaaS on Supabase Pro: compute at $15/mo net (after credits), storage overages near zero at launch, bandwidth likely within the 250GB included allotment, 7-day backups included, and near-zero operational engineering time. Total: $35 to $50/mo in year one.
For the same app on RDS db.t4g.micro single-AZ, on-demand: $11.68 compute, roughly $5/mo storage at 50GB, $0.09/GB data transfer out after the first GB free. Total: $20 to $30/mo. Cheaper on paper, but you also need to set up pgBouncer, configure backups manually, and handle major version upgrades. If you value your time at $100/hr and spend 3 hours/quarter on operational tasks, that is $1,200/yr in hidden cost.
For a scaling app at 100GB of data, 1TB of bandwidth, and need for Multi-AZ: RDS db.m5.large Multi-AZ reserved (1yr) runs roughly $200 to $250/mo compute plus $11.50/mo storage plus bandwidth charges. Supabase equivalent with large compute add-on and overages runs $150 to $300/mo depending on exact usage. The gap narrows and the decision becomes architecture and team preference, not just price.
Always build the estimate with your actual expected data growth curve, not your current state.
How Pineido Scopes Database Infrastructure in a 30-Day Sprint
When Pineido scopes a new SaaS or app build, database hosting is a first-week architecture decision, not an afterthought. For most early-stage builds we default to Supabase Pro because it collapses auth, database, storage, and realtime into a single operational surface area and keeps the sprint focused on product logic rather than infrastructure wiring. For teams with existing AWS footprints or compliance requirements that mandate isolated compute, we spec RDS with reserved capacity and include connection pooling in the sprint scope.
The fixed-price model works because these decisions are made explicitly at the start, with real numbers from the client's expected traffic and data projections, not guesses. If you are about to build a product and want a concrete infrastructure cost estimate alongside the development quote, the App and Web Development sprint is the starting point. Pricing is scoped per project.
Get operations insights weekly
Join service-business operators who get one actionable insight on ops, automation, and growth every week.
No spam. Unsubscribe anytime.
You’re subscribed! Look for our next issue in your inbox.
That did not go through. Please try again, or email [email protected].
Ready to ship?
Tell us where you are stuck. We will review it and send a focused recommendation within one business day.
Frequently asked questions
When should I choose Supabase over AWS RDS?
Choose Supabase when you want a fast path to production with auth, storage, and realtime bundled into one bill. For most early-stage apps under $10K MRR, Supabase Pro at $25/mo base is operationally simpler and often cheaper than a comparable RDS setup once you account for the engineering hours RDS requires for connection pooling, backup configuration, and parameter tuning.
Does Supabase really pause free projects?
Yes. Supabase free-tier projects pause after 7 days of inactivity. Any user request to a paused project triggers a cold resume that can take 30 to 60 seconds. This makes the free tier unsuitable for any public-facing endpoint. Upgrading to Pro ($25/mo) disables automatic pausing. Run free projects only for development and demos where inactivity gaps are acceptable.
Is PlanetScale expensive for write-heavy apps?
It can be. PlanetScale charges $1.50 per million row writes. An event-logging or analytics pipeline writing 50 million rows per day hits 1.5 billion writes per month, costing $2,250 on writes alone. For read-heavy workloads (10:1 read-to-write ratio or higher) the economics are favorable. Pull production write metrics before committing. If you are guessing, start on RDS or Supabase and migrate with real data.
Is RDS Multi-AZ worth the extra $3,432 per year?
It depends on your downtime tolerance. Multi-AZ gives automatic failover in under 2 minutes with no data loss. Single-AZ with good backups gives you manual recovery in 15 to 45 minutes with potential for a few minutes of data loss. If your app generates revenue directly and a 30-minute outage during business hours costs more than $3,432, Multi-AZ is worth it. Internal tools and low-traffic apps in year one usually do not clear that bar.
Why is Supabase storage so much more expensive than RDS at scale?
Supabase charges $2.50/GB for storage overages past your plan's included amount. AWS RDS gp3 charges $0.115/GB. At 500GB of data that is a $1,250/mo versus $57.50/mo difference. Supabase's pricing model is designed for apps that stay within their plan limits. Once your data grows past 50 to 100GB, model out the storage cost trajectory before renewal and consider whether migration to RDS or Cloud SQL makes economic sense.
How much can I actually save with RDS reserved instances?
30 to 60 percent versus on-demand, depending on term and payment option. A one-year no-upfront reserved db.m5.large saves roughly 30 percent, bringing the $143/mo on-demand price to around $100/mo. A three-year all-upfront reservation saves close to 60 percent, around $57/mo. Reserved instances make sense when you have 12-plus months of committed runway and a stable instance class. Avoid reserving instances before you have validated your production load.