Cloud databases have become the default choice for scaling applications, but pricing models vary dramatically across AWS, Azure, and Google Cloud. Understanding the real cost differences—and what you're actually paying for—is essential before committing to a platform. This guide breaks down the pricing structures so you can make an informed decision for your workloads.
AWS RDS and DynamoDB Pricing
AWS dominates the cloud database market, but its pricing structure rewards careful planning. RDS (relational databases) charges by instance type, storage, and data transfer. A db.t3.medium instance runs roughly $0.067 per hour on-demand (about $50/month), while larger production instances like db.r6i.2xlarge cost $1.50+ hourly. You'll also pay $0.10–$0.23 per GB for storage, depending on database type, plus backup and data transfer fees.
DynamoDB flips the model to pay-per-request. On-demand pricing runs around $1.25 per million write units and $0.25 per million read units, making it ideal for unpredictable workloads. If you have steady traffic, provisioned capacity saves 50–70% of costs—typically $0.00013 per write capacity unit per hour.
The real trap: data transfer out of AWS costs $0.09 per GB after the first GB. If you're syncing databases across regions or pulling large datasets, this adds up quickly.
Azure SQL Database and Cosmos DB
Azure pricing depends heavily on the service tier you choose. Basic tier runs $5–15 per month for development, while Standard (S0–S12) ranges from $15–$220 monthly. Premium tiers for production workloads start at $400 and climb significantly with compute and storage guarantees.
Azure SQL Database charges separately for compute (DTUs or vCores) and storage ($0.115 per GB/month for standard). A modest vCore setup (2 vCores, 32 GB) costs around $300–400 monthly, compared to AWS's equivalent at $200–250.
Cosmos DB introduces multi-region replication into the equation. Provisioned throughput costs $0.008 per 100 request units per hour, meaning a 10K RU setup runs roughly $60 per day. Serverless mode charges $0.25 per million operations—cheaper for spiky workloads but unpredictable at scale.
Google Cloud Firestore and Cloud SQL
Google Cloud tends to undercut competitors on storage but charges more for operations. Cloud Firestore (their NoSQL offering) costs $0.06 per 100K reads, $0.18 per 100K writes, and $0.18 per 100K deletes, plus $0.18 per GB stored. Light usage might cost $20–30/month; medium traffic runs $200–400.
Cloud SQL (MySQL, PostgreSQL, SQL Server) uses a per-hour compute model with instance types like db-f1-micro at $0.03/hour (about $22/month) up to db-n1-highmem-4 at $0.49/hour. Storage is simple: $0.18 per GB/month for SSD. Automated backups are included, which saves money versus AWS.
Google's egress pricing matches AWS at $0.09 per GB, but their storage costs are genuinely cheaper—a differentiator if you're managing large datasets.
Key Comparison Points
| Factor | AWS | Azure | Google Cloud | |--------|-----|-------|--------------| | Entry-level database | ~$50/month (RDS) | ~$5/month (Basic) | ~$22/month (Cloud SQL) | | Production small setup | $150–300/month | $300–500/month | $150–250/month | | NoSQL per-operation cost | High (DynamoDB) | Moderate (Cosmos) | Lowest (Firestore) | | Data transfer egress | $0.09/GB | $0.09/GB | $0.09/GB | | Reserved instances (discount) | Up to 55% | Up to 55% | Up to 70% |
Hidden Costs to Watch
Backups, snapshots, and restore operations often surprise teams. AWS charges $0.095 per GB for incremental backups; Azure includes limited backups but charges for extended retention. Network replication across regions is almost never "free"—budget $0.02 per GB transferred between zones at minimum.
Monitoring and logging tools add costs too. AWS CloudWatch, Azure Monitor, and Google Cloud Logging all charge for data ingestion. Enable detailed logging on a busy database and you could see an extra $100–300/month in monitoring fees alone.
Commitment-based discounts matter. AWS RIs and savings plans cut costs 40–55% for one-year commitments. Azure has similar programs. If you're managing databases for a year or longer, buying reserved capacity almost always makes financial sense.
Making Your Decision
Calculate your actual needs: How many transactions per second? How much storage? How often do you need geographic redundancy? Use each provider's pricing calculator with real numbers, then compare with a three-month, one-year, and three-year horizon. Many teams find different providers win for different workloads—AWS for relational scale, Google Cloud for analytics, Azure for organizations already deep in the Microsoft ecosystem.
Mercoly helps you compare Database Design & Administration providers and find trusted experts who've optimized costs across these platforms in your specific context.
Frequently Asked Questions
Q: Should I use on-demand or reserved instances? Reserve instances if your workload is consistent and you're planning for 1+ years; switch to on-demand only if you're testing configurations or expect traffic to vary wildly month to month.
Q: Why is my cloud database bill higher than expected? Most overages come from data transfer out (egress), unoptimized read/write operations in NoSQL databases, or backup storage accumulation—audit these three areas first.
Q: How do I know which cloud platform is cheapest for my database? Build a test instance on each platform with your actual schema and typical workload, run it for one week, and compare costs directly—calculators are useful but real-world traffic is the only true measure.
Ready to find a Database Design & Administration expert who can optimize your cloud database setup? Compare providers on Mercoly today.