This post is maintained by Bytebase, an open-source database governance platform that can manage PostgreSQL hosted in any providers. We update the post every year.
PostgreSQL remains the default choice for relational databases in 2026. This article compares hosting from the major clouds (AWS, GCP, Azure) and specialized providers (DigitalOcean, Aiven, Neon, PlanetScale, Prisma, Supabase, TigerData, Heroku), focusing on pricing models and key cost factors.
Understanding Pricing Models
Three pricing models dominate:
Instance-Based: Pay hourly for fixed vCPU/RAM; storage and I/O billed separately. Common with major clouds.
Usage-Based/Serverless: Pay only for what you use — compute, storage, I/O, and data transfer. Ideal for variable workloads.
Tiered Plans: Fixed-price bundles from specialized providers offer predictable costs but less flexibility.
Comparison Table
| Provider | Entry-Level (Monthly) | Mid-Range (Monthly) | Enterprise-Level (Monthly) | Free Tier Available |
|---|---|---|---|---|
| AWS RDS | $12.41 (db.t4g.micro: 2 vCPU, 1 GiB) | $99.28 (db.m6g.large: 2 vCPU, 8 GiB) | $794.24 (db.m6g.4xlarge: 16 vCPU, 64 GiB) | $200 credits (new accounts) |
| AWS Aurora | $59.86 (db.t4g.medium: 2 vCPU, 4 GiB) | $211.70 (db.r6g.large: 2 vCPU, 16 GiB) | $1,693.60 (db.r6g.4xlarge: 16 vCPU, 128 GiB) | $200 credits (new accounts) |
| Google Cloud SQL | $11.32 (db-f1-micro: 0.2 vCPU, 0.6 GiB) | $122.49 (db-n1-standard-2: 2 vCPU, 7.5 GiB) | $980.03 (db-n1-standard-16: 16 vCPU, 60 GiB) | Yes ($300/90 days) |
| Google AlloyDB | $143.68 (2 vCPU, 16 GiB) | $301.44 (4 vCPU, 32 GiB) | $1,209.60 (16 vCPU, 128 GiB) | No |
| Azure PostgreSQL | $14.60 (B1ms: 1 vCPU, 2 GiB) | $99.28 (GP_Gen5_4: 4 vCPU, 16 GiB) | $794.24 (GP_Gen5_32: 32 vCPU, 128 GiB) | Yes (12 months) |
| DigitalOcean | $15.00 (1 vCPU, 1 GiB) | $60.00 (2 vCPU, 4 GiB) | $244.00 (6 vCPU, 16 GiB) | No |
| Aiven | Free (2 vCPU, 1 GiB) | $110 (4 vCPU, 8 GiB) | $435 (8 vCPU, 16 GiB) | Yes |
| Neon | Free (100 CU-hours, 0.5 GiB storage) | Usage-based: $0.106/CU-hour (Launch) | Usage-based: $0.222/CU-hour (Scale) | Yes |
| Supabase | Free (1 vCPU, 0.5 GiB) | $25 (2 vCPU, 1 GiB) | $99 (4 vCPU, 8 GiB) | Yes |
| TigerData (Timescale) | Free (750 MB) | Usage-based (hourly compute + storage) | Usage-based | Yes |
| PlanetScale Postgres | $5 (PS-5: 1/16 vCPU, 512 MiB) | $148 (PS-80 HA: 3 x 1 vCPU, 8 GiB) | $570 (PS-320 HA: 3 x 4 vCPU, 32 GiB) | No |
| Prisma Postgres | Free (100k operations, 500 MB) | $10 (Starter: 1M operations, 10 GB) | $49-129 (Pro/Business: 10M-50M operations) | Yes |
| Heroku Postgres | $5 (Essential-0: 1 GB storage) | $50 (Standard-0) | $500+ (8 vCPU, 16 GiB) | No |
Major Cloud Providers: The Big Three
AWS, Google Cloud, and Azure all offer capable managed PostgreSQL; the tradeoff is complex pricing.
AWS
AWS offers two primary managed PostgreSQL options: Amazon RDS for PostgreSQL and Amazon Aurora (PostgreSQL-compatible edition).
RDS for PostgreSQL
- 💰 Pricing: Instance-based (e.g.,
db.t3.micro,db.m6g.large) - 📦 Storage: gp3 (scalable) or io1/io2 (high-performance)
- 🛡️ High Availability: Multi-AZ; doubles instance cost
- 💾 Backups: Free up to DB size; extra charged per GB-month
- 🧮 Discounts: Reserved Instances, Savings Plans
Aurora PostgreSQL-Compatible
- 💰 Pricing: Cluster-based, compute + storage + I/O billed separately
- ⚙️ Compute: Per instance-hour or per ACU-hour (Serverless v2)
- 📦 Storage: Auto-scaling, charged per GB-month
- 🔁 I/O: Pay-per-request (Standard) or free (I/O-Optimized)
- 🛡️ High Availability: Built-in, cross-AZ replication
- 💾 Backups: Continuous to S3
- 💵 Entry Point: Higher than RDS; designed for heavier workloads
Google Cloud
GCP provides two main managed PostgreSQL offerings: the standard Cloud SQL for PostgreSQL and the higher-performance AlloyDB for PostgreSQL.
Cloud SQL for PostgreSQL
- 💰 Pricing: Instance-based (
db-f1-micro,db-n1-standard-2, etc.) - ⚙️ Compute: Billed per vCPU + RAM hour
- 📦 Storage: SSD or HDD, per GB-month
- 🛡️ High Availability: Regional standby; doubles cost
- 💾 Backups: Charged per GB-month
- 🧮 Discounts: Committed Use Discounts (CUDs)
AlloyDB for PostgreSQL
- 💰 Pricing: Cluster-based with separate compute + storage
- ⚙️ Compute: Per vCPU + RAM hour
- 📦 Storage: Auto-scaling, includes I/O
- 🛡️ High Availability: Built-in, 99.99% SLA
- 💾 Backups: Continuous with PITR
- 💵 Entry Point: Higher (min: 2 vCPU, 16 GiB RAM)
Microsoft Azure
Azure's primary managed PostgreSQL offering is Azure Database for PostgreSQL, with Flexible Server as the recommended deployment model.
PostgreSQL Flexible Server
- 💰 Pricing: Compute + Storage
- ⚙️ Tiers: Burstable (B-series, low-cost), General Purpose (D-series), Memory Optimized (E-series)
- 📦 Storage: Per GB-month with baseline and optional IOPS
- 🛡️ High Availability: Optional zone-redundant (doubles cost)
- 💾 Backups: Configurable, per GB-month
- 🧮 Discounts: Reserved Capacity (up to 70% off compute)
Specialized Managed PostgreSQL Providers
Beyond the big three, specialized providers compete on developer experience, specific features (serverless, time-series, ORM integration), and simpler pricing, often built on the major clouds' infrastructure.
DigitalOcean
DigitalOcean offers Managed Databases for PostgreSQL, focusing on simplicity and predictable pricing.
- 💰 Model: Tiered monthly plans
- 💵 Entry Point: $15/month (1 vCPU, 1 GiB RAM, 10 GB SSD)
- 📦 Storage: Included in the plan
- 🌐 Bandwidth: 1 TB outbound included; overage billed
- 🛡️ High Availability: Optional standby node (doubles cost)
- 💾 Backups: Daily backups + 7-day PITR
- 🆓 Free Tier: ❌ Not available
Aiven
Aiven provides managed PostgreSQL (and other open-source data services) hosted on major cloud providers (AWS, GCP, Azure, DigitalOcean, UpCloud).
- 💰 Model: Tiered plans by cloud & region
- 💵 Entry Point: Free (2 vCPU, 1 GiB RAM, 1–5 GB SSD); paid plans from ~$19/month
- 📦 Storage & Network: Included in plan; generous allowances
- 🛡️ High Availability: Included in Business and Premium tiers
- 💾 Backups: Automatic backups included
- 🆓 Free Tier: ✅ Yes
Neon
Neon offers a serverless PostgreSQL platform with a unique architecture separating storage and compute. Databricks acquired Neon in May 2025 (its technology also powers Databricks Lakebase), and prices dropped substantially afterwards: storage fell from $1.75 to $0.35/GB-month and the monthly plan minimums were removed.
- 💰 Model: Purely usage-based: $0.106/CU-hour (Launch) or $0.222/CU-hour (Scale), no monthly minimum
- 💵 Entry Point: Free (100 CU-hours/month, 0.5 GB storage per project)
- ⚙️ Compute: Billed only when active; scales to zero when idle
- 📦 Storage: $0.35/GB-month
- 🛡️ High Availability: Built into platform architecture
- 💾 Backups: Includes time-travel and PITR
- 🆓 Free Tier: ✅ Yes
PlanetScale Postgres
PlanetScale, known for years as a MySQL/Vitess platform, launched managed PostgreSQL in 2025 and made benchmark-topping performance its pitch. Pricing is resource-based, pro-rated to the millisecond.
- 💰 Model: Resource-based node types, pro-rated to the millisecond
- 💵 Entry Point: $5/month single node (PS-5: 1/16 vCPU, 512 MB); 3-node HA clusters from $15/month
- ⚙️ Metal: All-inclusive clusters with local NVMe storage from $50/month; storage included
- 🛡️ High Availability: 3-node clusters with automatic failover
- 💾 Backups: 2x per day included + PITR
- 🔀 Coming: Neki, PlanetScale's sharded Postgres (Vitess-style horizontal scaling)
- 🆓 Free Tier: ❌ No, but startup credits available case-by-case
Prisma Postgres
Prisma (the company behind the popular TypeScript ORM) launched Prisma Postgres in 2025, a serverless PostgreSQL built on unikernel microVMs with no cold starts. Its pricing model is unique: you pay per ORM operation, not for compute or bandwidth.
- 💰 Model: Operations-based; an operation is one Prisma ORM query
- 💵 Entry Point: Free (100k operations, 500 MB storage, 50 databases); Starter $10/month (1M operations, 10 GB)
- ⚙️ Higher Tiers: Pro $49 (10M operations, 50 GB), Business $129 (50M operations, 100 GB)
- 📦 Overages: Metered per 1,000 operations and per GB, rates fall on higher tiers
- 💾 Backups: Daily; 30-day retention on Business
- 🆓 Free Tier: ✅ Yes
Supabase
Supabase positions itself as an open-source Firebase alternative, providing a suite of backend tools, including a dedicated PostgreSQL database for each project.
- 💰 Model: Tiered plans + usage-based overages
- 💵 Entry Point: Free (shared DB, 500 MB); Pro starts at $25/month
- ⚙️ Compute & Storage: Varies by tier; add-ons available
- 🛡️ High Availability: Included in Team and Enterprise tiers
- 💾 Backups: PITR; retention depends on tier
- 🆓 Free Tier: ✅ Yes
TigerData (formerly Timescale)
Timescale renamed itself TigerData in June 2025, repositioning from time-series specialist to "modern PostgreSQL for the analytical and agentic era." Tiger Cloud remains fully compatible with standard PostgreSQL, with TimescaleDB's hypertables and columnar storage built in.
- 💰 Model: Pure usage-based (hourly compute + actual storage used)
- 💵 Entry Point: ~$0.02/hr (smallest compute unit)
- 📦 Storage: Includes hypertables + regular tables
- 🛡️ High Availability: Optional (adds to cost)
- 💾 Backups: Continuous with PITR
- 🆓 Free Tier: ✅ Yes: free plan with 750 MB storage, database forks, and pgvector (read-only past the cap)
Heroku Postgres
Heroku Postgres is one of the original and well-known managed PostgreSQL services, tightly integrated with the Heroku PaaS. The Essential plans now run on Amazon Aurora under the hood.
- 💰 Model: Tiered hourly/monthly billing
- 💵 Entry Point: $5/month (Essential-0, 1 GB storage); Essential-1 $9 (10 GB), Essential-2 $20 (32 GB)
- 📦 Storage: Included based on plan
- 🛡️ High Availability: Premium and above tiers only
- 💾 Backups: PG Backups (free); PITR on paid tiers
- 🆓 Free Tier: ❌ Removed in 2022
Conclusion
The big three offer power and scale with complex pricing; specialized hosts offer free tiers and prices you can predict. Match the choice to workload, budget, and team expertise, and start on a free plan.
The specialized tier is also where the industry consolidation landed: Databricks bought Neon, Snowflake bought Crunchy Data (Crunchy Bridge still operates under Snowflake), Timescale renamed itself TigerData, and two new entrants (PlanetScale and Prisma) arrived with opposite bets: benchmark-topping hardware versus operations-based serverless. Postgres hosting has never had more competition, and prices have dropped accordingly.