.png)
Key takeaways
1
Model the configuration, not a headline rate. Edition, machine series, region, HA, storage type, replicas, and database engine determine the applicable prices.
2
Separate commitment-eligible and ineligible spend. Cloud SQL CUDs cover eligible regional CPU and memory usage, while storage, backups, networking, IP addresses, licensing, and extended support remain outside the discount.
3
Commit against a stable hourly baseline. A large discount does not compensate for a commitment that exceeds the CPU and memory usage your workloads consistently consume.
Google Cloud SQL pricing extends beyond the instance’s advertised compute rate. A reliable estimate must account for the selected edition and machine series, high availability (HA), provisioned storage, replicas, backups, networking, licensing, and any eligible committed use discount.
This guide explains those components and applies them to a worked MySQL example in us-central1. Use the same framework with the current rates for your database engine, machine series, region, and configuration.
HA, replicas, backups, data transfer, public IP usage, licensing, and extended support can materially increase the total. Cloud SQL CUDs can reduce eligible CPU and memory costs by 25% for one year or 52% for three years, but they do not discount the other components.
Enterprise Plus adds higher performance, availability, observability, and data-protection capabilities through its supported machine series.
The following calculations use Cloud SQL Enterprise’s general-purpose dedicated-core machine series in us-central1. Its list rates are $0.0413 per vCPU-hour and $0.007 per GiB-hour of memory. Other machine series and regions use different rates, so identify the machine series before applying an example.
For an instance with 4 vCPUs and 15 GiB of memory running for 730 hours:
This guide explains those components and applies them to a worked MySQL example in us-central1. Use the same framework with the current rates for your database engine, machine series, region, and configuration.
The Short Answer
Cloud SQL charges primarily for provisioned vCPU, memory, and storage. For a Cloud SQL Enterprise general-purpose dedicated-core instance in us-central1, 4 vCPUs and 15 GiB of memory cost about $197.25 per month before storage and other additions.HA, replicas, backups, data transfer, public IP usage, licensing, and extended support can materially increase the total. Cloud SQL CUDs can reduce eligible CPU and memory costs by 25% for one year or 52% for three years, but they do not discount the other components.
How Google Cloud SQL Compute Pricing Works
Cloud SQL has two editions. Enterprise provides the core Cloud SQL capabilities and supports N4, general-purpose shared-core, and general-purpose dedicated-core machine series.Enterprise Plus adds higher performance, availability, observability, and data-protection capabilities through its supported machine series.
The following calculations use Cloud SQL Enterprise’s general-purpose dedicated-core machine series in us-central1. Its list rates are $0.0413 per vCPU-hour and $0.007 per GiB-hour of memory. Other machine series and regions use different rates, so identify the machine series before applying an example.
For an instance with 4 vCPUs and 15 GiB of memory running for 730 hours:
vCPU: 4 × $0.0413 × 730 = $120.60
Memory: 15 × $0.007 × 730 = $76.65
Monthly compute: approximately $197.25
Memory: 15 × $0.007 × 730 = $76.65
Monthly compute: approximately $197.25
That figure excludes HA, storage, replicas, backups, networking, IP addresses, database licensing, and extended support.
Google states that an HA-configured instance costs twice as much as a standalone instance, including CPU, memory, and storage. Therefore, the example’s compute rises from approximately $197.25 to $394.49 per month.
Standard SSD storage in us-central1 is approximately $0.17 per GiB-month for a standalone instance and $0.34 per GiB-month for HA. With 500 GiB, storage is approximately $85 per month without HA or $170 with HA.
The resulting HA instance costs approximately $564.49 per month for compute and storage before backups and other charges. For development or staging, disable HA only when the workload’s availability requirements permit it; the decision should follow reliability requirements, not cost alone.
Cloud SQL does not reverse automatic storage increases when usage falls. Eligible instances support manual storage shrink to a safe minimum, but the operation restarts the instance and requires downtime. Google recommends migration to a smaller instance when downtime must be limited.
Read replicas must have at least as much storage as their primary. Shrinking a primary does not shrink existing replicas automatically, so each eligible replica must be evaluated and reduced separately.
Storage also depends on the machine series: general-purpose Enterprise configurations can use SSD or HDD, while N4 uses Hyperdisk Balanced.
Backup consumption depends on database size, changes, and retention. PITR log storage also varies by workload and configuration, so no universal percentage should be added.
Database versions enrolled in extended support incur a separate charge. Check the engine’s version policy before buying a multi-year CUD.
CUDs apply to eligible CPU and memory usage across Cloud SQL Enterprise and Enterprise Plus instances in the committed region. They support MySQL, PostgreSQL, and SQL Server, but exclude shared-core machine types.
Storage, backups, IP addresses, outbound data transfer, licensing, and extended-support charges are not covered.
Applied to the standalone $197.25 compute example, a fully utilized one-year CUD reduces covered compute to approximately $147.93 per month. A three-year CUD reduces it to approximately $94.68, saving about $1,230.82 annually compared with on-demand compute.
High Availability: The Full Cost Impact
A current Cloud SQL HA configuration creates a primary and standby in different zones of the same region. Writes are synchronously replicated across their persistent disks, and the standby becomes the primary after a qualifying failure. The standby cannot serve read queries.Google states that an HA-configured instance costs twice as much as a standalone instance, including CPU, memory, and storage. Therefore, the example’s compute rises from approximately $197.25 to $394.49 per month.
Standard SSD storage in us-central1 is approximately $0.17 per GiB-month for a standalone instance and $0.34 per GiB-month for HA. With 500 GiB, storage is approximately $85 per month without HA or $170 with HA.
The resulting HA instance costs approximately $564.49 per month for compute and storage before backups and other charges. For development or staging, disable HA only when the workload’s availability requirements permit it; the decision should follow reliability requirements, not cost alone.
Storage Pricing and Storage Shrink
Cloud SQL bills provisioned storage rather than only used database space. For the general-purpose Enterprise configuration here, standard SSD costs approximately $0.17 per GiB-month without HA and $0.34 with HA; standard HDD is approximately $0.09 and $0.18, respectively.Cloud SQL does not reverse automatic storage increases when usage falls. Eligible instances support manual storage shrink to a safe minimum, but the operation restarts the instance and requires downtime. Google recommends migration to a smaller instance when downtime must be limited.
Read replicas must have at least as much storage as their primary. Shrinking a primary does not shrink existing replicas automatically, so each eligible replica must be evaluated and reduced separately.
Storage also depends on the machine series: general-purpose Enterprise configurations can use SSD or HDD, while N4 uses Hyperdisk Balanced.
Additional Cloud SQL Costs to Include
Add every applicable item below before approving a budget.Read replicas
A read replica uses standard Cloud SQL instance rates for its selected configuration. Each replica has separate compute and storage charges, while an HA replica adds HA pricing. An HA standby is included in the primary’s HA price and cannot serve reads.Public IPv4 addresses
Public connectivity assigns a static IPv4 address. Google currently charges $0.01 per hour while the instance is off or deactivated; the charge is not determined by whether traffic flows.Network transfer
Incoming transfer is free, but outgoing charges depend on the source, destination, regions, address path, network tier, and volume. Same-region traffic is not universally free across every path, so use current Google Cloud network pricing.Backups and PITR
Standard backup storage is billed by space used; this guide applies approximately $0.08 per GiB-month in us-central1. Enhanced backups use Backup and DR and have a different billing structure.Backup consumption depends on database size, changes, and retention. PITR log storage also varies by workload and configuration, so no universal percentage should be added.
Licensing and extended support
Cloud SQL for SQL Server bills licensing separately from compute and requires at least four licensed cores per instance. An instance below four vCPUs is therefore charged four times the applicable license rate; CUDs do not discount it.Database versions enrolled in extended support incur a separate charge. Check the engine’s version policy before buying a multi-year CUD.
Committed Use Discounts: Coverage and Readiness
Cloud SQL CUDs exchange a commitment to consistent hourly spend in a region for discounted eligible usage. A one-year term provides a 25% discount and a three-year term provides a 52% discount. The percentage is the same across regions, although the underlying regional prices vary.CUDs apply to eligible CPU and memory usage across Cloud SQL Enterprise and Enterprise Plus instances in the committed region. They support MySQL, PostgreSQL, and SQL Server, but exclude shared-core machine types.
Storage, backups, IP addresses, outbound data transfer, licensing, and extended-support charges are not covered.
Applied to the standalone $197.25 compute example, a fully utilized one-year CUD reduces covered compute to approximately $147.93 per month. A three-year CUD reduces it to approximately $94.68, saving about $1,230.82 annually compared with on-demand compute.
| Usage pattern | Practical decision |
|---|---|
| Stable production CPU and memory baseline | Size a CUD against the verified hourly floor, not the peak. |
| Intermittent development or uncertain demand | Keep bursty or uncertain usage on demand until a stable floor emerges. |
| Storage or licensing dominates the bill | Optimize those charges separately because a Cloud SQL CUD will not reduce them. |
Any usage above the commitment is billed at on-demand rates. If eligible usage drops below the committed amount, the commitment fee continues for the full term.
Cloud SQL Key Rates: us-central1, 2026
Rates checked: September 10, 2026. These figures apply to the assumptions stated in the table and reflect Google Cloud SQL pricing for us-central1. Check the current pricing page before using them in a budget.| Component | On-demand rate | CUD treatment |
|---|---|---|
| Enterprise general-purpose dedicated vCPU | $0.0413/hour | 25% off for 1 year; 52% off for 3 years |
| Enterprise general-purpose dedicated memory | $0.007/GiB-hour | 25% off for 1 year; 52% off for 3 years |
| HA compute and storage | 2× standalone rate | CUD applies only to eligible CPU and memory |
| Standard SSD, non-HA | Approximately $0.17/GiB-month | Not discounted |
| Standard SSD, HA | Approximately $0.34/GiB-month | Not discounted |
| Standard HDD, non-HA | Approximately $0.09/GiB-month | Not discounted |
| Standard backup storage | Approximately $0.08/GiB-month | Not discounted |
| Public IPv4 while instance is off | $0.01/hour | Not discounted |
A Worked Cost Example
Monthly Cloud SQL cost
=
CPU and memory, including any HA multiplier
+
provisioned storage, including any HA multiplier
+
replicas
+
backups
+
networking and IP charges
+
licensing or support fees
−
applicable discounts
Assumptions: MySQL 8.0; Cloud SQL Enterprise edition; general-purpose dedicated core; us-central1; 4 vCPUs; 15 GiB memory; 500 GiB SSD; HA primary; one same-spec non-HA replica; 300 GiB standard backups; and 730 running hours. Chargeable transfer, licensing, extended support, and off-instance IPv4 hours are excluded.
| Cost component | Calculation | Monthly cost |
|---|---|---|
| Primary vCPU with HA | 4 × $0.0413 × 2 × 730 | $241.19 |
| Primary memory with HA | 15 × $0.007 × 2 × 730 | $153.30 |
| Primary SSD with HA | 500 × $0.34 | $170.00 |
| Standard backup storage | 300 × $0.08 | $24.00 |
| Primary subtotal | $241.19 + $153.30 + $170.00 + $24.00 | $588.49 |
| Replica vCPU | 4 × $0.0413 × 730 | $120.60 |
| Replica memory | 15 × $0.007 × 730 | $76.65 |
| Replica SSD | 500 × $0.17 | $85.00 |
| Replica subtotal | $120.60 + $76.65 + $85.00 | $282.25 |
| Total on demand | $588.49 + $282.25 | $870.74 |
| One-year CUD saving on eligible compute | $591.74 × 25% | −$147.93 |
| Post-CUD total | Using unrounded values: $870.738 − $147.9345 | $722.80 |
A fully utilized one-year CUD saves approximately $1,775.21 annually while leaving storage and backup charges unchanged.
For a defensible Cloud SQL versus Amazon RDS comparison, match the engine, region, availability, storage, licensing, transfer, and commitment assumptions.
We connect at the billing layer, analyze billing and usage data, and generate recommendations that you can approve manually or manage through Autopilot. Approved purchases use cloud-provider APIs and limited permissions that do not provide access to workloads or production infrastructure.
With Flex Insured Commitments, teams can get the 52% savings of a three-year Cloud SQL CUD on eligible CPU and memory usage with none of the commitment risk.
Cashback protection is conditional. If an eligible Flex Commitment costs more than the equivalent On-Demand usage, we calculate the covered loss and pay cashback according to the applicable program terms.
We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.
For a defensible Cloud SQL versus Amazon RDS comparison, match the engine, region, availability, storage, licensing, transfer, and commitment assumptions.
How Usage.ai Optimizes Cloud SQL Costs
Cloud SQL CUDs can lower a stable compute baseline, but their one- or three-year terms create exposure when eligible usage falls.We connect at the billing layer, analyze billing and usage data, and generate recommendations that you can approve manually or manage through Autopilot. Approved purchases use cloud-provider APIs and limited permissions that do not provide access to workloads or production infrastructure.
With Flex Insured Commitments, teams can get the 52% savings of a three-year Cloud SQL CUD on eligible CPU and memory usage with none of the commitment risk.
Cashback protection is conditional. If an eligible Flex Commitment costs more than the equivalent On-Demand usage, we calculate the covered loss and pay cashback according to the applicable program terms.
We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.
Conclusion
Build a Cloud SQL estimate from the selected compute configuration, then add HA, provisioned storage, replicas, backups, networking, licensing, and support. Consider a CUD only after separating eligible CPU and memory spend and identifying a stable regional baselineFrequently 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
