Cloud Database Pricing Compared: AWS vs Azure vs Google Cloud

Short answer

All three clouds bill managed relational databases mainly by instance time plus provisioned storage; AWS and Google list hourly rates, Azure lists monthly rates per instance size.
For a 2 vCPU, 8 GiB PostgreSQL database with 100 GB of storage in a single zone, the three list-price estimates are close to each other, with Google lowest and Azure highest in our example.
Among the instances with about 2 GiB of memory, Azure's B1ms works out cheapest; the very smallest options (db.t4g.micro on AWS, shared-core db-f1-micro on Google) cost less but have less memory, and Google's shared-core types are outside the Cloud SQL SLA.
For a database used only a few hours a day, usage-billed Neon costs less than the smallest always-on AWS instance in our example; high availability, backups beyond allowances and data transfer add to every bill.

How we know: Method: list prices only, on-demand (no commitments or savings plans), PostgreSQL, single availability zone, in USD, checked on 8 October 2026. Regions: AWS US East (N. Virginia), Azure East US, Google Cloud Iowa (us-central1), AWS figures come from the published price data behind the Amazon RDS for PostgreSQL pricing page. Worked examples are our arithmetic on those prices at 730 hours per month, excluding tax, data transfer and support. We have not run these instances, so nothing here is a measured bill.

How cloud database pricing works on each provider

Cloud database pricing uses the same building blocks everywhere, packaged differently by each provider.

Billing model for managed PostgreSQL on the three clouds (October 2026)
	Amazon RDS for PostgreSQL	Azure Database for PostgreSQL	Google Cloud SQL for PostgreSQL
Compute unit	Instance class (for example db.t4g.small), per hour	Instance size (for example B1ms), listed per month	vCPUs and memory priced separately, per hour; shared-core types priced per instance
Smallest options	Burstable T4g and T3 classes	Burstable B-series	Shared-core db-f1-micro and db-g1-small
Storage	Per GB-month by storage type	Per GiB-month, Premium SSD and other options	Per GiB-hour for SSD or HDD
Backups	A separate monthly cost component	No charge up to 100% of provisioned server storage	Charged per GiB-hour for backups used
High availability	Multi-AZ deployment at a higher instance and storage rate	Same-zone or zone-redundant standby option	HA vCPU, memory and storage rates (double the single-zone rates)
Commitment discounts	Reserved Instances and Database Savings Plans	Savings plan and reserved instances (1 and 3 years)	Committed use discounts (1 and 3 years)

What this means for your bill

Google's per-vCPU and per-GB pricing lets you size memory and CPU independently, which can help memory-heavy workloads. Azure's monthly figures are easy to read but still assume the server runs all month (you can stop a server to stop compute billing). AWS separates the most items, including CPU credits on burstable classes when sustained usage exceeds the baseline. For per-provider detail, see the AWS RDS pricing, Azure SQL pricing and Cloud SQL pricing pages in this section, and the head-to-head AWS RDS vs Google Cloud SQL, AWS RDS vs Azure SQL Database and Azure SQL vs Google Cloud SQL.

Equivalent small instances compared

No two clouds sell identical machines, so we matched on memory and vCPU count as closely as the price lists allow. The specifications come from each provider's own documentation or price data.

Instance sizes used in the worked examples
Size class	AWS (RDS for PostgreSQL)	Azure (PostgreSQL flexible server)	Google Cloud SQL
Smallest listed	db.t4g.micro: 2 vCPU, 1 GiB (burstable)	B1ms: 1 vCore, 2 GiB (burstable)	db-f1-micro: shared CPU, 0.6 GiB
About 2 GiB memory	db.t4g.small: 2 vCPU, 2 GiB (burstable)	B1ms: 1 vCore, 2 GiB (burstable)	db-g1-small: shared CPU, 1.7 GiB
2 vCPU, 8 GiB general purpose	db.m7g.large: 2 vCPU, 8 GiB	D2ds v5: 2 vCores, 8 GiB	Enterprise edition custom: 2 vCPU, 8 GiB

Caveats on the match

Burstable and shared-core instances are not equal to dedicated ones: they suit light or uneven load. Google states that its shared-core types are not covered by the Cloud SQL SLA, so treat them as development options. Processor generations also differ (the AWS classes above use Graviton processors), and CodeWithSQL has not compared performance, so cheaper does not mean better value for your workload.

NoSQL and serverless database pricing

NoSQL and serverless databases are priced on usage rather than instance time, so they cannot be compared line by line with the examples above. DynamoDB pricing on demand is per million read and write request units plus storage; Firestore database pricing is per 100,000 document reads, writes and deletes plus storage, after a daily free quota; Azure Cosmos DB serverless charges for request units consumed. Serverless PostgreSQL such as Neon bills compute-unit hours, which is why it appears in the intermittent-workload example below. For the trade-offs, see serverless databases explained.

How to reduce cloud database costs

Right-size first. Start small; all three clouds let you change instance size later.
Commit only for steady workloads. Reserved Instances, savings plans and committed use discounts lower the rate in exchange for a 1- or 3-year commitment; the Azure page shows about 20% off for a 1-year savings plan on the sizes above.
Stop or pause what is idle. Stop development servers outside working hours where the service allows it, or use a serverless option.
Keep the database next to the application. Cross-region and internet data transfer is billed separately.
Upgrade old engine versions. AWS charges RDS Extended Support per vCPU-hour for major versions past standard support.
Check storage and backup settings. Over-provisioned storage and long backup retention are easy to forget.

For providers beyond the big three, see best database hosting providers.

Pricing

Prices from the provider's official pricing pages, checked 8 October 2026, region AWS US East (N. Virginia); Azure East US; Google Cloud Iowa (us-central1), in USD, excluding tax. List prices only; discounts, commitments and your actual usage change the bill.

On-demand list prices for managed PostgreSQL, single zone. AWS prices are from AWS's published RDS for PostgreSQL price data for US East (N. Virginia); Azure prices from the Azure Database for PostgreSQL flexible server pricing page with region East US; Google prices from the Cloud SQL pricing page with region Iowa (us-central1). Neon and Supabase prices are from their pricing pages and are not tied to one region on those pages.

Usage-priced services for reference: DynamoDB on-demand is 0.625 USD per million write request units and 0.125 USD per million read request units in AWS's published examples; Firestore in Iowa is 0.03 USD per 100,000 document reads and 0.09 USD per 100,000 document writes beyond the free quota.

Managed PostgreSQL list prices used in the examples (8 October 2026)
Item	AWS RDS (US East, N. Virginia)	Azure (East US)	Google Cloud SQL (Iowa)
Smallest instance	db.t4g.micro: 0.016 per hour	B1ms: 12.41 per month	db-f1-micro: 0.0105 per hour
About 2 GiB instance	db.t4g.small: 0.032 per hour	B1ms: 12.41 per month	db-g1-small: 0.035 per hour
2 vCPU, 8 GiB instance	db.m7g.large: 0.168 per hour	D2ds v5: 129.94 per month	0.0413 per vCPU-hour plus 0.007 per GiB-hour of memory
SSD storage	gp3: 0.115 per GB-month	Premium SSD: 0.115 per GiB-month	0.000232877 per GiB-hour (about 0.17 per GiB-month)
Backup storage	Billed separately; rate not quoted here	Free up to 100% of provisioned storage	0.000109589 per GiB-hour used

Example 1: small always-on app database (about 2 GiB memory, 32 GB storage)

PostgreSQL, single zone, running 730 hours a month.
32 GB of SSD storage on each cloud.
Backups within free allowances; data transfer, tax and HA excluded.

Item	Basis	Estimated per month
AWS db.t4g.small	730 hours x 0.032 USD = 23.36; 32 GB x 0.115 = 3.68	27.04 USD
Azure B1ms	12.41 USD per month; 32 GiB x 0.115 = 3.68	16.09 USD
Google db-g1-small	730 x 0.035 = 25.55; 32 GiB x 0.000232877 x 730 = 5.44	30.99 USD
Estimated total		AWS 27.04 USD, Azure 16.09 USD, Google 30.99 USD per month (Google shared-core is outside the Cloud SQL SLA)

Example 2: production database (2 vCPU, 8 GiB, 100 GB storage)

PostgreSQL, single zone (no standby), 730 hours a month.
100 GB of SSD storage.
Backups beyond allowances, high availability, data transfer and tax excluded.

Item	Basis	Estimated per month
AWS db.m7g.large	730 x 0.168 = 122.64; 100 GB x 0.115 = 11.50	134.14 USD
Azure D2ds v5	129.94 USD per month; 100 GiB x 0.115 = 11.50	141.44 USD
Google Enterprise, 2 vCPU and 8 GiB	730 x 2 x 0.0413 = 60.30; 730 x 8 x 0.007 = 40.88; 100 GiB x 0.17 = 17.00	118.18 USD
Estimated total		AWS 134.14 USD, Azure 141.44 USD, Google 118.18 USD per month

Example 3: development database used 8 hours a day (10 GB storage)

Active about 176 hours a month (8 hours x 22 working days), idle otherwise.
Neon at its 0.25 CU minimum size while active, scaling to zero when idle.
Supabase Pro with one Micro project covered by the included compute credits.
AWS instance left running all month for comparison; data transfer and tax excluded.

Item	Basis	Estimated per month
Neon Launch	176 h x 0.25 CU = 44 CU-hours x 0.106 = 4.66; 10 GB x 0.35 = 3.50	8.16 USD
Supabase Pro	25 USD plan (includes 10 USD compute credit for Micro); 8 GB disk included, 2 GB x 0.125 = 0.25	25.25 USD
AWS db.t4g.micro, always on	730 x 0.016 = 11.68; 10 GB x 0.115 = 1.15	12.83 USD
Estimated total		Neon 8.16 USD, Supabase 25.25 USD, AWS always-on 12.83 USD per month

Worked examples are estimates calculated from the list prices above; they are not quotes or measured bills.

Frequently asked questions

Which cloud is cheapest for a managed database?

It depends on the size. In our list-price examples, Azure's B1ms was cheapest for a small always-on database with about 2 GiB of memory, and Google Cloud SQL was cheapest for a 2 vCPU, 8 GiB database. Region, storage, high availability and discounts can change the order, so price your own configuration.

How much does an AWS database cost per month?

It depends on the instance class, storage and region. The smallest RDS for PostgreSQL instance (db.t4g.micro) is one of the cheapest managed options on any cloud, and a 2 vCPU, 8 GiB instance costs several times more; the pricing section above shows the hourly rates and the monthly arithmetic. Free plan credits apply to new accounts.

How is DynamoDB pricing different?

DynamoDB on-demand charges per read and write request plus storage, not per instance hour, and has a monthly free tier, which suits uneven traffic.

Do the prices include high availability?

No. All examples are single zone. AWS and Google publish higher rates for Multi-AZ and HA configurations, and Azure adds a standby server in the same or another zone, so high availability raises the bill on all three.

Why are AWS prices quoted from a JSON file?

The Amazon RDS pricing page loads its price tables interactively by region. The figures shown in those tables come from AWS's published price data files, which we cite directly so each number can be checked.

Sources

Checked 8 October 2026.

How we research cloud database guides: our editorial method. CodeWithSQL earns nothing from the providers mentioned.