GCP storage pricing: storage classes, operations and egress
Updated: 10 November 2026GCP storage pricing should be read as a set of workload charges. For Cloud Storage, compare the selected class and location with object operations, retrieval and data transfer instead of multiplying all stored data by one headline rate.
Request a quote
For an ML workload, reconcile retained inputs and outputs with the model execution costs in the BQML pricing review.
For objects used by serverless container services, the Cloud Run application cost review adds service capacity and billing-mode evidence to the storage comparison.
For disk recovery points rather than objects in buckets, use the GCP snapshot storage and retention cost checklist; Compute Engine snapshot schedules follow their own retention rules.
GCP storage pricing: separate the billing dimensions
The official Cloud Storage pricing page lists storage, operations, retrieval and data transfer charges, with rules that vary by class and location. Keep the actual billing units in the worksheet. This guide concerns Cloud Storage object data; persistent disks and database storage require their own service pricing and should not be mixed into a bucket-rate estimate.
| Dimension | Input | Review question |
|---|---|---|
| Capacity | Average volume by class and location | What must remain readily accessible? |
| Operations | Read, write and listing activity | Does the application create repeated requests? |
| Retrieval | Read volume from applicable classes | Does lower capacity cost survive access charges? |
| Retention | Object age and expected remaining life | Could early deletion charges apply? |
| Transfer | Source and destination of data movement | Where is the data consumed? |
The Google Cloud cost review establishes the billing account and project boundary. Read BigQuery cost analysis alongside object storage when pipelines move or retain data in both services. Assign each cost once and document shared data ownership.
Choose a class for the object population
The Cloud Storage class documentation describes Standard, Nearline, Coldline and Archive and their access and retention characteristics. Do not treat a colder class as offline storage by analogy with another cloud. Read the service-specific behavior and minimum storage duration rules. Model the object group you will actually move, including its expected reads and deletion dates. The current class guide also lists Rapid storage for zonal buckets; it has separate workload and location constraints. This class comparison focuses on the Standard-to-Archive options.
- Group objects by owner, age, size and access pattern.
- Include retained versions and incomplete lifecycle decisions in the inventory.
- Review the location where applications read the data.
- Check recovery, deletion and retention requirements with the data owner.
Validate lifecycle changes before rollout
The Object Lifecycle Management guide describes rules that can change classes or delete eligible objects. Define the condition and action for one bounded population. Review how retained versions and retention controls affect the intended outcome. Test the rule against an inventory before approving deployment and keep an owner for the data decision. The pricing page states that Object Lifecycle Management class changes do not incur early deletion charges. Manual rewrites and deletions have different rules; evaluate the specific action rather than charging every lifecycle transition as early deletion.
When storage supports containers, GKE cost optimization keeps the application and persistent resources in one review. Compare broader design alternatives using the cloud pricing comparison method. A storage class name is not proof that two clouds offer equivalent access or recovery behavior. Keep service differences visible in the decision sheet.
- Reconcile capacity and operation lines to the bucket inventory.
- Choose a population and calculate the complete class scenario.
- Approve retention and validate the lifecycle conditions.
- Measure capacity, retrieval, operations and traffic after the change.
Prepare the evidence for a quote
Provide aggregate bucket capacity, class and location, object-age distribution, operation counts and read volumes. State which data needs a recovery test and who can approve deletion. Ask the proposal to separate analysis, policy implementation and post-change measurement. The handover should include rate sources, dated assumptions and editable calculations by population. Keep future growth and one-time data moves separate from the recurring result so the next reviewer can reproduce the comparison.
Questions before you decide
Is Archive always the cheapest total cost?
Compare capacity with retrieval, operations, minimum duration and data movement. The cheapest class for the complete workload depends on how that population is used.
Can I use the same class assumptions as another cloud?
Check Cloud Storage documentation directly. Class labels do not establish equivalent retrieval behavior, retention rules or billing dimensions across services.
Keep object storage separate from the cost of storing and querying data in BigQuery. Reconcile both services when an analytics pipeline retains data in each.