AWS
Microsoft Azure
Google Cloud
Key takeaways
1
Look beyond the instance price. With Cloud SQL pricing, compute is only part of the bill. Storage, backups, HA, and replicas can make the total higher than you expect.
2
Check what the discount covers. CUDs can cut eligible CPU and memory costs by up to 52%, but other charges stay separate. Your overall savings will depend on the mix.
3
Think ahead before committing. A commitment can save money on steady usage, but you keep paying if that usage drops. Factor in planned changes and any management fees before buying.
How Cloud SQL pricing works
Cloud SQL pricing starts with the resources you choose for your database: processing power, memory, and storage. You then add the costs of your setup, such as high availability and read replicas.Think of it as renting capacity. A running database with little traffic can still incur compute charges because that capacity remains provisioned. Google’s Cloud SQL overview explains the main pricing inputs.
| Cost component | What you pay for |
|---|---|
| CPU and memory | Your selected processing capacity and RAM. A vCPU is a unit of processing power; RAM holds data the database is working with. |
| Storage | Provisioned disk capacity, with additional performance charges where applicable. |
| HA and replicas | The resources supporting your standby and any additional read replicas. |
| Backups | Backup storage, with pricing depending on the backup option. |
| Networking and IP | Chargeable outgoing data transfer and applicable IP address charges. |
| Licensing and extended support | SQL Server licensing and charges for database versions in extended support. |
Google’s Cloud SQL pricing page lists the applicable rates and conditions. Not every setup incurs every charge. Our guide to GCP cost optimization best practices explores that broader review.
Build your estimate in the Google Cloud Pricing Calculator
Build your estimate in the Google Cloud Pricing Calculator
Record your backup option, location, and retention assumptions. Estimate any backup charges not included in the calculator separately; enhanced backups use Backup and DR pricing.
1
Add Cloud SQL in the Google Cloud Pricing Calculator, then choose your database engine and edition.
2
Select the region, machine series, vCPU count, and memory.
3
Add provisioned storage and choose a standalone or HA deployment.
4
Record your backup option, location, and retention assumptions. Estimate any backup charges not included in the calculator separately; enhanced backups use Backup and DR pricing.
5
Account for replicas, transfer, IP, licensing, and extended support where applicable. Then model CUD coverage separately for eligible regional CPU and memory, comparing the commitment payment with your expected minimum hourly usage.
How engine, edition, and region affect Cloud SQL pricing
Two databases with the same CPU and memory can have different bills. Before comparing Cloud SQL pricing, check these three choices.Database engine. MySQL and PostgreSQL do not carry the additional Microsoft licensing charges that SQL Server does. SQL Server licensing has a four-core minimum, so even a smaller instance can incur four cores’ worth of license charges. See Google’s SQL Server pricing rules.
Edition and machine series. Enterprise and Enterprise Plus offer different performance and availability capabilities. For qualifying HA configurations, their availability SLAs are 99.95% and 99.99%, respectively; shared-core and single-zone instances are excluded.
Check the machine series, too. For PostgreSQL, N2 uses SSD storage, while N4, C4, and C4A use Hyperdisk Balanced. That changes which storage charges belong in your estimate. Google’s machine-series guide maps the options.
Region. The comparison below uses USD list rates for Enterprise general-purpose dedicated-core HA vCPUs in Iowa and Singapore, checked October 22, 2026. Both calculations assume 730 hours. So, check Google’s current pricing before budgeting.
| Region | HA vCPU rate/hour | Four HA vCPUs × 730 hours |
|---|---|---|
Iowa (us-central1)
|
$0.0826 | $241.19 |
Singapore (asia-southeast1)
|
$0.1156 | $337.55 |
These calculated totals cover only the vCPU portion, before discounts. Memory, storage, and other charges remain separate. Check current regional rates before budgeting.
Compare matching configurations first. A cheaper headline price is useful only when you know what it includes.
For the worked example, we’ll use MySQL 8.0 on Enterprise’s general-purpose dedicated-core series in Iowa (us-central1), with SSD storage.
We’ll use 4 vCPUs, 15 GiB of RAM, and 500 GiB of SSD storage. The estimate assumes 730 running hours and 300 GiB of billable standard backup storage throughout the month.
If you use enhanced backups, estimate them separately using Backup and DR pricing.
Now, using Google’s Cloud SQL pricing tables, the standalone calculation is:
Compare matching configurations first. A cheaper headline price is useful only when you know what it includes.
Cloud SQL Pricing Examples for Standalone Instances, HA, and Replicas
Let’s put the charges together. We’ll start with one database, then see what happens when we add availability protection and extra read capacity.For the worked example, we’ll use MySQL 8.0 on Enterprise’s general-purpose dedicated-core series in Iowa (us-central1), with SSD storage.
We’ll use 4 vCPUs, 15 GiB of RAM, and 500 GiB of SSD storage. The estimate assumes 730 running hours and 300 GiB of billable standard backup storage throughout the month.
If you use enhanced backups, estimate them separately using Backup and DR pricing.
Now, using Google’s Cloud SQL pricing tables, the standalone calculation is:
| Component | Calculation | Monthly cost |
|---|---|---|
| CPU | 4 × $0.0413 × 730 hours | $120.60 |
| Memory | 15 × $0.007 × 730 hours | $76.65 |
| SSD storage | 500 GiB × approximately $0.17 | $85.00 |
| Standard backups | 300 GiB × approximately $0.08 | $24.00 |
| Standalone total | Calculated before rounding | $306.25 |
These are our calculations from Google’s published rates, not a fixed monthly package. Actual charges depend on runtime and billable storage over time.
Now add high availability. Google’s HA configuration provides a standby in another zone and doubles the equivalent standalone CPU, memory, and storage charges. The standby helps recover from qualifying failures; it cannot serve read queries.
Keeping our backup assumption unchanged:
Now add high availability. Google’s HA configuration provides a standby in another zone and doubles the equivalent standalone CPU, memory, and storage charges. The standby helps recover from qualifying failures; it cannot serve read queries.
Keeping our backup assumption unchanged:
| Configuration | Compute | SSD storage | Backups | Monthly total |
|---|---|---|---|---|
| Standalone | $197.25 | $85.00 | $24.00 | $306.25 |
| HA primary, including standby | $394.49 | $170.00 | $24.00 | $588.49 |
| HA primary + one non-HA read replica | $591.74 | $255.00 | $24.00 | $870.74 |
Monthly estimates use 730 running hours and approximate storage rates. Totals are calculated before rounding, so displayed amounts may differ by $0.01 when added.
The read replica adds another $282.25 monthly. We’ve given it the same CPU, memory, and storage as the primary, without HA. Google bills read replicas at standard instance rates; they provide additional read capacity separately from the HA standby.
This Cloud SQL pricing example excludes chargeable networking, IP addresses, additional log-storage charges, extended support, taxes, and discounts. The backup amount is an assumption, not a percentage automatically derived from database size.
The useful comparison is the complete configuration: $197.25 of standalone compute becomes approximately $870.74 once this example includes HA, a replica, storage, and backups.
According to Google’s Cloud SQL CUD documentation, the discounts are 25% for one year and 52% for three years. Here’s where they apply.
Cloud SQL CUDs cover
The read replica adds another $282.25 monthly. We’ve given it the same CPU, memory, and storage as the primary, without HA. Google bills read replicas at standard instance rates; they provide additional read capacity separately from the HA standby.
This Cloud SQL pricing example excludes chargeable networking, IP addresses, additional log-storage charges, extended support, taxes, and discounts. The backup amount is an assumption, not a percentage automatically derived from database size.
The useful comparison is the complete configuration: $197.25 of standalone compute becomes approximately $870.74 once this example includes HA, a replica, storage, and backups.
What Cloud SQL committed use discounts cover
If your database needs stay predictable, you can reduce Cloud SQL pricing through a committed use discount. You agree to a minimum hourly spend for one or three years, and Google discounts the eligible usage it covers.According to Google’s Cloud SQL CUD documentation, the discounts are 25% for one year and 52% for three years. Here’s where they apply.
Cloud SQL CUDs cover
Eligible CPU and memory, including HA and read-replica compute.
MySQL, PostgreSQL, and SQL Server across Enterprise and Enterprise Plus editions.
Cloud SQL CUDs do not cover
Storage, backups, IP addresses, outbound data transfer, licensing, or extended support.
Shared-core instances such as db-f1-micro and db-g1-small.
The commitment covers eligible Cloud SQL usage within the purchased region and billing account, across its projects. It is not tied to one named instance.
When purchasing, check how the amount is expressed. Google’s current spend-based model uses discounted hourly spending. For example, covering $1/hour of eligible on-demand usage with a one-year CUD means committing to $0.75/hour. Older examples express that commitment as the $1 on-demand equivalent.
The trade-off is the payment obligation. You cannot cancel the commitment, and you keep paying if eligible usage falls. Usage beyond its coverage is billed on demand.
For the broader differences between commitment types, see our GCP committed use discounts guide.
Applying Google’s Cloud SQL CUD discounts gives us the following comparison. We assume the commitment covers all eligible compute and is fully used every hour.
When purchasing, check how the amount is expressed. Google’s current spend-based model uses discounted hourly spending. For example, covering $1/hour of eligible on-demand usage with a one-year CUD means committing to $0.75/hour. Older examples express that commitment as the $1 on-demand equivalent.
The trade-off is the payment obligation. You cannot cancel the commitment, and you keep paying if eligible usage falls. Usage beyond its coverage is billed on demand.
For the broader differences between commitment types, see our GCP committed use discounts guide.
How Much Can You Save After Commitment Costs and Fees?
Let’s return to our $870.74 monthly example. Eligible CPU and memory account for $591.74; storage and backups make up the remaining $279.Applying Google’s Cloud SQL CUD discounts gives us the following comparison. We assume the commitment covers all eligible compute and is fully used every hour.
| Monthly cost | On demand | One-year CUD | Three-year CUD |
|---|---|---|---|
| Eligible compute | $591.74 | $443.80 | $284.03 |
| Storage and backups | $279.00 | $279.00 | $279.00 |
| Total | $870.74 | $722.80 | $563.03 |
| Savings before management fees | — | $147.93 | $307.70 |
| Reduction in the total bill | — | 17.0% | 35.3% |
Calculations use unrounded values; displayed amounts may differ by $0.01 when subtracted.
The 52% compute discount reduces this complete bill by approximately 35.3%. Storage and backups stay unchanged, which is why the headline discount and overall savings differ.
If you use a third party to manage commitments, account for its applicable fees separately:
The 52% compute discount reduces this complete bill by approximately 35.3%. Storage and backups stay unchanged, which is why the headline discount and overall savings differ.
If you use a third party to manage commitments, account for its applicable fees separately:
Net savings = equivalent on-demand cost − total cost with commitments − management fees.
For example, the three-year scenario leaves $307.70 in monthly savings before any management fee. Subtract the fee specified in your agreement to calculate what you retain.
This Cloud SQL pricing comparison also assumes steady usage. If eligible usage falls, the commitment payment continues, and the savings can shrink or disappear. That is the next calculation to check.
Using Google’s Cloud SQL discount rates, here is a simplified calculation for a commitment initially covering $10/hour of eligible usage:
This Cloud SQL pricing comparison also assumes steady usage. If eligible usage falls, the commitment payment continues, and the savings can shrink or disappear. That is the next calculation to check.
When Does a Cloud SQL Commitment Break Even?
A commitment breaks even when its payment equals what the covered usage would otherwise cost on demand. Above that point, you save money. Below it, you pay more than you would without the commitment.Using Google’s Cloud SQL discount rates, here is a simplified calculation for a commitment initially covering $10/hour of eligible usage:
| Monthly cost | On demand | One-year CUD | Three-year CUD |
|---|---|---|---|
| Eligible compute | $591.74 | $443.80 | $284.03 |
| Storage and backups | $279.00 | $279.00 | $279.00 |
| Total | $870.74 | $722.80 | $563.03 |
| Savings before management fees | — | $147.93 | $307.70 |
| Reduction in the total bill | — | 17.0% | 35.3% |
These are calculated thresholds, before management fees, assuming unchanged rates and no other eligible usage absorbs the unused commitment.
Before purchasing, use Google’s CUD recommendations and scenario modeling to compare terms, coverage levels, and usage-history periods. Account separately for planned changes that historical data cannot predict.
What happens if usage falls by half? The remaining usage would cost $5/hour on demand. The one-year commitment still costs $7.50/hour, creating a $2.50/hour loss. The three-year commitment costs $4.80/hour, leaving just $0.20/hour in savings before fees.
Google checks commitment usage hourly, and unused entitlement does not roll over. A busy afternoon cannot recover an unused morning entitlement.
Resizing or stopping instances can reduce eligible spending. Moving workloads to another region leaves them outside the original regional commitment. The payment obligation continues, although other eligible usage in the original region and billing account can consume it.
These percentages describe commitment consumption, not CPU activity. A lightly loaded database can still consume its commitment through provisioned compute.
When evaluating Cloud SQL pricing, check both whether you need that capacity and whether its commitment saves money.
Use these questions to review your Cloud SQL pricing estimate:
Before purchasing, use Google’s CUD recommendations and scenario modeling to compare terms, coverage levels, and usage-history periods. Account separately for planned changes that historical data cannot predict.
What happens if usage falls by half? The remaining usage would cost $5/hour on demand. The one-year commitment still costs $7.50/hour, creating a $2.50/hour loss. The three-year commitment costs $4.80/hour, leaving just $0.20/hour in savings before fees.
Google checks commitment usage hourly, and unused entitlement does not roll over. A busy afternoon cannot recover an unused morning entitlement.
Resizing or stopping instances can reduce eligible spending. Moving workloads to another region leaves them outside the original regional commitment. The payment obligation continues, although other eligible usage in the original region and billing account can consume it.
These percentages describe commitment consumption, not CPU activity. A lightly loaded database can still consume its commitment through provisioned compute.
When evaluating Cloud SQL pricing, check both whether you need that capacity and whether its commitment saves money.
Check Your Eligible Savings Before Committing
Before choosing a term, check what your database will need after planned changes. Google’s Cloud SQL CUD guidance recommends aligning commitments with historical and expected minimum spending.Use these questions to review your Cloud SQL pricing estimate:
Will the setup change? Account for planned resizing, shutdowns, replica changes, and regional migrations.
What is already covered? Check existing commitments before purchasing more.
What remains predictable? Identify the eligible CPU and memory spending you expect to maintain each hour in the region.
What will you retain? Compare savings after applicable management fees, including a scenario where usage falls.
How Usage.ai Optimizes Cloud SQL Costs
Getting a lower rate is only part of managing Cloud SQL pricing. Google’s three-year Cloud SQL CUD offers a 52% discount on eligible CPU and memory, but your total-bill savings depend on eligible compute’s share of spending, commitment utilization, and applicable management fees.We analyze billing and usage data to recommend commitments, then use cloud-provider APIs to make approved purchases. Through our Flex Insured Commitment Program, eligible Cloud SQL commitments receive Cashback protection under the program terms.
If a covered commitment costs more than equivalent on-demand usage, we calculate the covered loss monthly. Cashback is paid 90 days after it accrues. For customers on a savings-based commercial model, our fee may be an agreed percentage of realized savings.
Frequently asked questions
How much does Google Cloud SQL cost monthly?
The total depends on the complete configuration. Here, the HA primary costs about $588.49 monthly, or $870.74 with one same-spec non-HA replica.
What do Cloud SQL CUDs cover?
They discount eligible regional CPU and memory by 25% for one year or 52% for three years, excluding storage, backups, IP, transfer, licensing, and extended support.
Are read replicas cheaper than primaries?
Not automatically. A replica uses standard rates for its selected compute and storage, although it can be sized differently from its primary.
Does HA double Cloud SQL cost?
Google prices HA at twice the standalone CPU, memory, and storage rates; separately metered charges still depend on the configuration.
Does the calculator include every Cloud SQL cost?
It estimates the inputs provided. Include replicas, backups, transfer, IP, licensing, extended support, and discounts rather than relying on compute alone.
Share
Facebook
X
LinkedIn
Reddit
