/

Console

• English
• Deutsch
• Español
• Español – América Latina
• Français
• Indonesia
• Italiano
• Português
• Português – Brasil
• עברית
• 中文 – 简体
• 中文 – 繁體
• 日本語
• 한국어

Sign in

• Cloud Billing

Start free

Structure of Detailed data export

This document provides reference information for the schema of Cloud Billing detailed usage cost data that's exported to each table in BigQuery.

The detailed usage cost data provides all of the information included in the standard usage cost data, along with additional fields that provide granular, resource-level cost data, like a virtual machine or SSD that generates usage. The detailed export includes granular cost information about the following services:

Identify granular cost data by service

To analyze granular cost information in a detailed export, use the following table to identify the column that contains information about specific resources.

See examples of querying granular data for your resources.

Detailed usage cost data schema

In your BigQuery dataset, your detailed Google Cloud usage cost data is loaded into a data table named gcp_billing_export_resource_v1_<BILLING_ACCOUNT_ID>.

When you use the detailed usage cost data in BigQuery, note the following:

• The table schema for the detailed Cloud Billing data export can change over time, such as when new fields are added to the detailed export. To protect your queries from future schema changes, rather than relying on queries that directly reference the exported columns, we recommend that you use BigQuery views to normalize the data so that the table presents the same schema to your queries. Learn more about this limitation.
• When selecting or creating a BigQuery dataset for your detailed usage cost data, you can select any dataset location that is supported for use with Cloud Billing data.
• When you enable the detailed usage cost data export for the first time in Cloud Billing, if you select a dataset configured to use a multi-region location (EU or US), Cloud Billing data will be available retroactively from the start of the previous month. Data is exported in chronological order. For the initial backfill of exported data, it might take up to five days for your retroactive Cloud Billing data to finish exporting before you start seeing your most recent usage data.
• If you enable the detailed usage cost data export and select a dataset that's configured to use a supported region location, your Cloud Billing data will be available starting from the date when you enabled the export.
• If you enabled, disabled, and subsequently re-enabled the detailed usage cost data export, the Cloud Billing data might not be available for the period when data export was explicitly disabled.
• Learn more about the frequency of the data loads into your BigQuery tables.
• See other limitations that might impact exporting your billing data to BigQuery, such as datasets with customer-managed encryption keys (CMEK) enabled.
• Consider the additional data volume that your BigQuery tables might need and the additional cost when enabling detailed usage cost data instead of the standard usage cost data export. The increased granularity of resource-level information can increase the number of rows, which are aggregated in the standard usage cost format. We recommend that you review Control costs in BigQuery for further best practices in managing your BigQuery costs.

The Cloud Billing account ID that the usage is associated with.

For resellers: For usage costs generated by a Cloud Billing subaccount, this is the ID of the subaccount, not the ID of the parent reseller Cloud Billing account.

The year and month (YYYYMM) of the invoice that includes the cost line items. For example: "201901" is equivalent to January, 2019.

You can use this field to get the total charges on the invoice. See Cloud Billing Export to BigQuery Query Examples.

Indicates the publisher associated with the transaction. This field supports the splitting of invoices between transactions made directly with Google (first party), and transactions made with a partner (third party), which also signals what regulations might apply to the transaction.

• GOOGLE: First-party, unregulated transaction by Google Cloud.
• PARTNER: A third-party, regulated or unregulated transaction by a partner.

The type of cost this line item represents: regular, tax, adjustment, or rounding error.

For more information, see the BigQuery documentation on timestamp data types. See also, Differences between exported data and invoices.

For more information, see the BigQuery documentation on timestamp data types. See also, Differences between exported data and invoices.

For example: /ParentOrgNumber/ParentFolderNumber/. Learn more about the Resource Hierarchy.

This field describes the structure and value of the resource hierarchy of a cost line item, including projects, folders, and organizations. Ancestors are ordered from node to root (project, folder, then organization).

Cost per the applicable consumption model inclusive of any negotiated discounts reflected in the custom pricing in the contract linked to your Cloud Billing account (if applicable).

If the credits.id field is empty, then the product SKU isn't associated with a credit.

• FEE_UTILIZATION_OFFSET For the new Spend-based CUD discounts, this credit is used to offset the fees paid to purchase CUDs. With the new model, you purchase commitments at a discounted price. As you use the SKUs covered by the CUD, this credit offsets the fee. When you fully utilize your commitment, the offset reduces the fee to zero, with your eligible SKU usage charged at the discounted price.
• COMMITTED_USAGE_DISCOUNT_DOLLAR_BASE: For legacy spend-based committed use discounts (CUDs) that aren't part of the new pricing model, this is the credit earned in exchange for your commitment to spend a minimum amount for a service in a particular region.
• COMMITTED_USAGE_DISCOUNT: Resource-based committed use contracts purchased for Compute Engine in return for deeply discounted prices for VM usage.
• DISCOUNT: The discount credit type is used for credits earned after a contractual spending threshold is reached. Note that in the Cloud Billing reports available in the Google Cloud console, the discount credit type is listed as Spending-based discounts.
• FREE_TIER: Some services offer free resource usage up to specified limits. For these services, credits are applied to implement the free tier usage.
• PROMOTION: The promotion credit type includes spend-based milestone credits, Google Cloud Free Trial and marketing campaign credits, or other grants to use Google Cloud. When available, promotional credits are considered a form of payment and are automatically applied to reduce your total bill.
• RESELLER_MARGIN: If you're a reseller, the reseller margin credit type indicates the Reseller Program Discounts earned on every eligible line item.
• SUBSCRIPTION_BENEFIT: Credits earned by purchasing long-term subscriptions to services in exchange for discounts.
• SUSTAINED_USAGE_DISCOUNT: The sustained use discounts credit type is an automatic discount that you receive for running eligible Compute Engine resources for a significant portion of the billing month, with no commitment required.

adjustment_info values are only present if the cost line item was generated for a Cloud Billing modification. A modification can happen for correction or non-correction reasons. The adjustment_info type contains details about the adjustment, whether it was issued for correcting an error or other reasons.

The type of adjustment.

Types include:

• USAGE_CORRECTION: A correction due to incorrect reported usage.
• PRICE_CORRECTION: A correction due to incorrect pricing rules.
• METADATA_CORRECTION: A correction to fix metadata without changing the cost.
• GOODWILL: A credit issued to the customer for goodwill.
• SALES_BASED_GOODWILL: A credit issued to the customer for goodwill, as part of a contract.
• SLA_VIOLATION: A credit issued to the customer due to a service-level objective (SLO) violation.
• BALANCE_TRANSFER: An adjustment to transfer funds from one payment account to another.
• ACCOUNT_CLOSURE: An adjustment to bring a closed account to a zero balance.
• GENERAL_ADJUSTMENT: A general billing account modification.

How the adjustment was issued.

Modes include:

• PARTIAL_CORRECTION: The correction partially negates the original usage and cost.
• COMPLETE_NEGATION_WITH_REMONETIZATION: The correction fully negates the original usage and cost, and issues corrected line items with updated usage and cost.
• COMPLETE_NEGATION: The correction fully negates the original usage and cost, and no further usage is remonetized.
• MANUAL_ADJUSTMENT: The adjustment is allocated to cost and usage manually.

Fields that describe the tag, such as key, value, and namespace.

The short name or display name of the key associated with this particular tag.

The resources attached to a tags.key. At any given time, exactly one value can be attached to a resource for a given key.

Indicates whether a tag binding is inherited (Tags Inherited = True) or direct/non-inherited (Tags Inherited = False). You can create a tag binding to a parent resource in the resource hierarchy.

Represents the resource hierarchy that define tag key and values. Namespace can be combined with tag key and tag value short names to create a globally unique, fully qualified name for the tag key or tag value.

Cost at list price per the default consumption model.

• GOOGLE = 1: Services sold by Google Cloud.
• THIRD_PARTY_RESELLER = 2: Third party services resold by Google Cloud.
• THIRD_PARTY_AGENCY = 3: Third party services sold by a partner, with Google Cloud acting as the agent.

The legal name of the seller.

This field is populated for Marketplace-related cost line items, which is set for countries where Google or an affiliate can act as the vendor's agent. For more information, see Agency Jurisdictions.

Fields that describe the structure and values related to the prices charged for usage.

SKU list price per the default consumption model.

SKU price per the default consumption model inclusive of any negotiated discounts reflected in the custom pricing in the contract linked to your Cloud Billing account (if applicable).

SKU list price per the applicable consumption model before any negotiated discounts reflected in the custom pricing in the contract linked to your Cloud Billing account (if applicable).

SKU price per the applicable consumption model inclusive of any negotiated discounts reflected in the custom pricing in the contract linked to your Cloud Billing account (if applicable).

The lower bound number of units for a SKU's pricing tier. For more information, see About pricing tiers. price.unit String

The unit of usage in which the pricing is specified and resource usage is measured.

The SKU's pricing unit quantity. For example, if the price is $1 per 1000000 Bytes, then this column shows 1000000.

Cost per the default consumption model inclusive of any negotiated discounts reflected in the custom pricing in the contract linked to your Cloud Billing account (if applicable).

Cost per the applicable consumption model before any negotiated discounts reflected in the custom pricing in the contract linked to your Cloud Billing account (if applicable).

Fields that describe the applicable consumption model.

The ID of the consumption model.

The description of the consumption model.

The fields that describe the structure and value of information relevant to service resources (like a virtual machine or a SSD) that generate usage.

A globally unique service identifier for the resource that generated relevant usage.

A service-specific identifier for the resource that generated relevant usage. This can be input generated by the user.

Fields that describe your spend-based or resource-based commitments. You can use these fields to analyze your fees for specific commitments.

The subscription ID linked to a commitment.

Understand standard and detailed usage cost data

The following sections describe the standard and detailed usage cost data exported to BigQuery.

About labels

The cost data for a specific label only shows usage from the date that the label was applied to a resource. For example, if you add the label environment:dev to a Compute Engine VM on January 15, 2024, any analysis for environment:dev includes only the usage for that VM since January 15.

You might also see label data at different times for different services, depending on when each service provides it.

Available system labels

System labels are key-value pairs for important metadata about the resource that generated the usage. The following system labels are automatically included on applicable usage.

Differences between exported data and invoices

Google Cloud and billing-eligible products such as Google Maps Platform, Google AI Studio, and Firebase report usage data to Cloud Billing at varying intervals. Due to the complexity of our billing and processing systems, you might see a delay between your use of services, the usage charges being applied to your Google payments accounts, and the usage and costs being available to view in the various cost reports and dashboards.

• Typically, your cost details are available within a day, but can sometimes take more than 24 hours.
• For self-serve, online accounts, it's possible that charges for some usage might occur rapidly, such as 5 to 15 minutes after usage.
• Because of this difference in timing, you might notice charges on your Postpay or Prepay payments accounts occur before you can view the details of those charges in a cost report or billing dashboard. These system delays also impact budget alerts, anomaly detection, and billing data exported to BigQuery.

At the end of a calendar month, late-reported usage might not be included on that month's invoice and instead might roll over to the next month's invoice.

When you query your costs using timestamp fields, your returned data might pick up late-reported usage that wasn't originally included on the invoice that was generated for the same usage month. As a result, the Cloud Billing data returned might not map directly to that invoice.

Timestamp fields include:

• usage_start_time
• usage_end_time
• export_time

To return Cloud Billing data that maps directly to an invoice, query on invoice.month instead of timestamp fields.

Taxes

As of September 1, 2020, your usage cost data shows your tax liability for each of your projects, instead of as a single line item. If you have queries or visualizations that depend on tax data, you might need to update the queries to account for these changes.

For example, for costs recorded before September 1, your usage cost data looks similar to the following example, which shows a total tax liability of $10.

For costs recorded after September 1, the $10 is broken down to $6 for example-project, and $4 for test-project:

Errors and adjustments

In the rare event that your Cloud Billing data contains an error or requires an adjustment, it's appended with corrective data. These adjustments fall under one of two categories: billing modifications or corrections.

Billing modifications

Billing modifications appear as separate line items. If you received a billing modification, a new line item in your Cloud Billing export to BigQuery shows the change. The adjustments shown correspond to the invoice, credit memo, and debit memo documents available in the Documents area of the Billing section in the Google Cloud console.

For more information on billing modifications and how they're applied, see Understand memos and adjustments.

Corrections

Corrections appear as new data that negates incorrect data on the source SKUs. In some cases, new data replaces the incorrect charge. All columns in the billing data export will match the original data, except for the following columns:

• cost
• credit
• usage.amount
• export_time

For example, imagine that you're charged $10 for your usage of SKU A on January 1. On your January invoice (issued in early February), you'll see a charge of $10 for SKU A. However, on February 2, Google Cloud issued a correction against SKU A, reducing the usage cost to $5. You'll receive two additional line items on your February invoice (issued in early March):

• -$10 for usage on January 1 (negating the original charge)
• $5 for usage on January 1 (stating the intended charge)

These new items have an adjustment_info column in the billing data export. The original January invoice, showing the overcharge, won't be adjusted. You can verify your charges in your billing data export by viewing your costs by usage_start_time and grouping by Day. In these views, any corrections or charges for late-monetized usage are accumulated, and you don't need to worry about any temporarily incorrect data.

If you want more detailed information on your corrections, view all charges in an invoice month, and look for charges where the usage date occurred before the invoice month. These charges are the results of corrections or late-monetized usage.

The following code sample shows how to create a basic query that returns the total cost of corrections or late-monetized usage:

SELECT
  SUM(cost)
    +   SUM(IFNULL((SELECT SUM(c.amount)
      FROM   UNNEST(credits) c), 0))
    AS total
FROM `project.dataset.gcp_billing_export_v1_XXXXXX-XXXXXX-XXXXXX`
WHERE
  invoice.month = '202311' AND
  DATE(TIMESTAMP_TRUNC(usage_start_time, Day, 'US</Pacific'))  '2023-11-01';

For a query example that returns a cost breakdown by service, for invoice charges, where the usage date occurred before the invoice month, see Query cost details to view corrections or late-monetized usage by service for a specified invoice month in "Example queries for Cloud Billing data export."

About promotional credits in custom pricing contracts

If you have a custom pricing contract, you might receive promotional credits to use on Google Cloud as part of the contract. For example, you might receive $1,000 to use on Compute Engine resources. Promotional credits are typically considered a form of payment. When available, promotional credits are automatically applied to reduce your total bill.

The terms of your contract specify whether the promotional credits apply to your costs calculated at the list price of a SKU, or the net price (after discounts).

If your promotional credits apply to costs that are calculated at the list price, in the Cost table report, there's a service called Invoice, with a SKU called Contract billing adjustment. This SKU adjusts your credits so that they apply to the costs at list price. To see the usage that the adjustment is for, query the system.labels columns. The key in system.labels.key is cloud-invoice.googleapis.com/sku_id, and the value in system.labels.value contains the SKU ID that the credit and the adjustment applied to.

About tags

Tags are resources in the form of key-value pairs that can be attached to resources directly or through inheritance. You can use tags to perform chargebacks, audits, and other cost allocation analysis. You can also use tags and conditional enforcement of policies for fine-grained control across your resource hierarchy.

Tags have a robust permissions model and can support inheritance, centralized management, nomenclature standardization, and policy engine integration, while labels are a separate tool that allow you to annotate resources.

Tags data appears in BigQuery exports for Resources, Projects, Folders, and Organizations.

Available tags

The Standard costs and Detailed costs exports for Resources, Projects, Folders, and Organizations include these fields for tags data: Tags Key, Tags Value, Tags Inherited, and Tags Namespace.

Resource-level tags in the Cloud Billing data export are available for the following resources:

• AlloyDB for PostgreSQL clusters, instances, and backups
• Artifact Registry repositories
• Bigtable instances
• Cloud Run services and jobs
• Cloud Storage buckets
• Compute Engine instances
• Memorystore for Redis instances
• Secret Manager global and regional secrets
• Spanner instances
• Firestore databases

Tags limitations

• Tags might take up to an hour to propagate to BigQuery exports. If a tag has been added or removed within an hour, or if a resource has existed for less than an hour, it might not appear in the export.

Detailed usage cost query examples

This section provides examples of how to query the Cloud Billing detailed usage cost data exported to BigQuery.

• Return the resource-level costs on an invoice
  • Sum costs for each resource, per invoice
  • Return details by cost type for each resource, per invoice month

• Get breakdown of Google Kubernetes Engine (GKE) cluster costs
  • Filter GKE cost breakdown

Because the detailed usage cost schema includes all of the fields from the standard usage cost schema, the query examples provided for the standard data exported to BigQuery also work with the detailed data that's exported. The standard query examples aren't written to retrieve any of the resource-level information that's provided with the detailed usage cost export option. When creating queries for the detailed data, you can use a standard query example as a template, update the Table name, and add any of the fields that are available in the detailed usage cost schema.

Common values used in the example detailed cost queries

The query examples in this section use the following value for Table name: project.dataset.gcp_billing_export_resource_v1_XXXXXX_XXXXXX_XXXXXX

Return the resource-level costs on an invoice

The following queries demonstrate two ways of viewing resource-level cost and credit values on an invoice using exported billing data.

• The total field directly sums the floating point cost and credit values, which can result in floating point rounding errors.
• The total_exact field converts costs and credit values to micros before summing, then converts back to dollars after summing, avoiding the floating point rounding error.

Sum costs for each resource, per invoice

This query shows the invoice total for each resource.name per month, as a sum of regular costs, taxes, adjustments, and rounding errors. Any costs not associated with a resource-level item are aggregated under the name null for the month.

Standard SQL

SELECT
  invoice.month,
  resource.name,
  SUM(cost)
    + SUM(IFNULL((SELECT SUM(c.amount)
                  FROM UNNEST(credits) c), 0))
    AS total,
  (SUM(CAST(cost * 1000000 AS int64))
    + SUM(IFNULL((SELECT SUM(CAST(c.amount * 1000000 as int64))
                  FROM UNNEST(credits) c), 0))) / 1000000
    AS total_exact
FROM `project.dataset.gcp_billing_export_resource_v1_XXXXXX_XXXXXX_XXXXXX`
GROUP BY 1, 2
ORDER BY 1 ASC, 2 ASC
;

For example, the result of the preceding query might be:

Return details by cost type for each resource, per invoice month

This query shows the totals for each cost_type for each resource.name per month. Cost types include regular costs, taxes, adjustments, and rounding errors. Any costs not associated with a resource-level item are aggregated under the name null for the month.

Standard SQL

SELECT
  invoice.month,
  cost_type,
  resource.name,
  SUM(cost)
    + SUM(IFNULL((SELECT SUM(c.amount)
                  FROM   UNNEST(credits) c), 0))
    AS total,
  (SUM(CAST(cost * 1000000 AS int64))
    + SUM(IFNULL((SELECT SUM(CAST(c.amount * 1000000 as int64))
                  FROM UNNEST(credits) c), 0))) / 1000000
    AS total_exact
FROM `project.dataset.gcp_billing_export_resource_v1_XXXXXX_XXXXXX_XXXXXX`
GROUP BY 1, 2, 3
ORDER BY 1 ASC, 2 ASC, 3 ASC
;

For example, the result of the preceding query might be:

Get a breakdown of Google Kubernetes Engine (GKE) cluster costs

This section provides examples of filtering GKE cluster costs in your BigQuery export reports. To learn more about GKE cluster costs, visit View breakdown of cluster costs.

Filter GKE costs

The following example queries show you how to filter and group your GKE costs for supported resource types by cluster name, namespace, and label.

GKE cluster costs before credits

SELECT
  SUM(cost) AS cost_before_credits,
  labels.value AS cluster_name
FROM  `project-ID.dataset.gcp_billing_export_resource_v1_XXXXXX-XXXXXX-XXXXXX`
LEFT JOIN UNNEST(labels) as labels
  ON labels.key = "goog-k8s-cluster-name"
GROUP BY labels.value
;

GKE costs after credits by namespace

SELECT
  labels.value as namespace,
  SUM(cost) + SUM(IFNULL((SELECT SUM(c.amount) FROM UNNEST(credits) c), 0)) AS cost_after_credits,
FROM `project-ID.dataset.gcp_billing_export_resource_v1_XXXXXX-XXXXXX-XXXXXX`
LEFT JOIN UNNEST(labels) as labels
  ON labels.key = "k8s-namespace"
GROUP BY namespace
;

GKE costs by SKU

SELECT
  project.id AS project_id,
  labels.value AS cluster_name,
  sku.id AS sku_id,
  sku.description AS sku_description,
  SUM(cost) AS cost
FROM `project-ID.dataset.gcp_billing_export_resource_v1_XXXXXX-XXXXXX-XXXXXX`
JOIN UNNEST(labels) AS labels
  ON labels.key = "goog-k8s-cluster-name"
GROUP BY
  cluster_name, project_id, sku_description, sku_id
;

Query examples with tags

The following examples illustrate ways to query your data with tags.

Calculate costs by invoice month with tags

The following query demonstrates how you can use return costs by invoice month for the cost_center tag.

SELECT
  invoice.month AS invoice_month,
  tag.value AS cost_center,
  ROUND((SUM(CAST(cost AS NUMERIC))
    + SUM(IFNULL((SELECT SUM (CAST(c.amount AS NUMERIC))
                  FROM UNNEST(credits) AS c), 0))), 2)
    AS net_cost
FROM `project-ID.dataset.gcp_billing_export_resource_v1_XXXXXX-XXXXXX-XXXXXX`, UNNEST(tags) AS tag
WHERE tag.key = "cost_center" AND tag.namespace = "821092389413"
GROUP BY invoice.month, tag.value
ORDER BY invoice.month, tag.value;

For example, the result of the preceding query might be:

View costs of untagged resources

This query shows the invoice total for untagged resources, grouped by invoice month.

SELECT
 invoice.month AS invoice_month,
 ROUND((SUM(CAST(cost AS NUMERIC))
   + SUM(IFNULL((SELECT SUM(CAST(c.amount AS NUMERIC))
                 FROM UNNEST(credits) AS c), 0))), 2)
   AS net_cost
FROM
 `project-ID.dataset.gcp_billing_export_v1_XXXXXX-XXXXXX-XXXXXX`
WHERE "color" NOT IN (SELECT key FROM UNNEST(tags))
GROUP BY invoice_month
ORDER BY invoice_month;

For example, the result of the preceding query might be:

Query using system labels

System labels for Google Cloud reservations provide enhanced visibility into your reservation usage. They allow you to track which instances are consuming which reservations, and analyze the costs associated with both used and unused capacity directly within BigQuery.

These examples use two system labels:

• compute.googleapis.com/reservation_name
• compute.googleapis.com/reservation_project_id

These labels are present on Cloud Billing records from both instances consuming a reservation and the unused portion of a reservation.

The following examples demonstrate how to query the BigQuery billing export to analyze reservation consumption using system labels.

Before you begin

To query using system labels, you must have Cloud Billing export to BigQuery enabled, specifically the Detailed usage cost data export. This export contains the resource-level information, including system labels, necessary to see the new reservation labels.

To set up or verify your BigQuery export, see the following documentation:

1. Overview of billing data export to BigQuery
2. Set up billing data export to BigQuery
3. Understand the Detailed data export schema

Ensure your BigQuery dataset is configured to receive the detailed export. The table name will typically be in the format gcp_billing_export_resource_v1_<BILLING_ACCOUNT_ID>.

View consumption for a given instance

This query lets you check if a given instance utilized a reservation during a particular period and, if so, which reservation it was.

SELECT
  t.usage_start_time,
  sl_reservation.value AS reservation_name,
  t.sku.description AS sku_description,
  SUM(t.usage.amount) AS total_usage_amount
FROM
  project-ID.dataset.gcp_billing_export_resource_v1_X AS t
  LEFT JOIN
  UNNEST(t.system_labels) AS sl_reservation
  ON sl_reservation.key = 'compute.googleapis.com/reservation_name'
WHERE
  t.service.id = '6F81-5844-456A' AND DATE(TIMESTAMP_TRUNC(t._PARTITIONTIME, DAY)) = '2025-09-16' AND t.sku.description LIKE '%Core%' AND
  t.resource.name LIKE '%tc-billing-case1-instance-1757983787%'
GROUP BY t.usage_start_time, reservation_name, t.sku.description
ORDER BY t.usage_start_time, reservation_name, t.sku.description;

Sample Result:

Note:

• SKU Filtering: Limiting the SKU to CPU cores offers clearer resource distribution. This can be adjusted to other reserved resources, such as RAM, LocalSSD, TPU, or GPU.
• On-Demand Consumption: When both reservation_name and reservation_project_id are null, the instance is consuming on-demand resources.

When multiple rows with differing reservation_name and reservation_project_id correspond to a specific instance within a given timeframe, it indicates that the instance's consumed reservation changed during that period. The usage portion will detail the consumption.

Analyze usage distribution for a given reservation

This query helps you understand the distribution of usage for a given shared reservation, showing how much capacity was used or unused among different Google Cloud projects that can consume the given reservation.

SELECT
  t.project.id AS consumer_project_id,
  t.sku.description AS sku_description,
  CASE
    WHEN t.resource.global_name NOT LIKE '%instances/%' THEN 'Unused Capacity'
    WHEN t.resource.name IS NOT NULL AND t.resource.name != '' THEN 'Instance Usage'
    ELSE 'Other Reservation Cost'
  END AS usage_category,
  SUM(t.usage.amount) AS total_usage_amount,
  t.usage.unit AS usage_unit,
  SUM(t.cost) AS total_cost_usd
FROM
  project-ID.dataset.gcp_billing_export_resource_v1_X AS t
  LEFT JOIN
  UNNEST(t.system_labels) AS sl_reservation
  ON sl_reservation.key = 'compute.googleapis.com/reservation_name'
  LEFT JOIN
  UNNEST(t.system_labels) AS sl_project_id
  ON sl_project_id.key = 'compute.googleapis.com/reservation_project_id'
WHERE
  t.service.id = '6F81-5844-456A' AND DATE(TIMESTAMP_TRUNC(t._PARTITIONTIME, DAY)) = '2025-09-16' AND t.sku.description LIKE '%Core%' AND
  sl_reservation.value LIKE '%tc-billing-case2-res-1757983801%'
GROUP BY consumer_project_id, t.sku.description, usage_category,
  usage_unit
ORDER BY consumer_project_id, usage_category;

Sample Results:

List all the instances consuming a reservation

Similarly, we can write a query to understand all the instances, including first-party (1P) instances, that have consumed a given reservation during a given time period.

SELECT
  t.usage_start_time,
  t.usage_end_time,
  t.project.id AS consumer_project_id,
  t.resource.name AS instance_name,
  t.usage.amount AS usage_amount
FROM
  project-ID.dataset.gcp_billing_export_resource_v1_X AS t
  LEFT JOIN
  UNNEST(t.system_labels) AS sl_reservation
  ON sl_reservation.key = 'compute.googleapis.com/reservation_name'
WHERE
  t.service.id = '6F81-5844-456A' AND DATE(TIMESTAMP_TRUNC(t._PARTITIONTIME, DAY)) = '2025-09-16' AND t.sku.description LIKE '%Core%' AND
  sl_reservation.value LIKE '%tc-billing-case2-res-1757983801%' AND t.resource.global_name NOT LIKE '%allocation/%' AND
  t.resource.name IS NOT NULL AND t.resource.name != ''
ORDER BY t.usage_start_time, t.project.id, t.resource.name;

Sample results:

Analyze Vertex AI reservation usage using BigQuery

To analyze how first-party (1P) products, such as Vertex AI, consume your shared reservations across projects where Vertex AI jobs are active, use the following system labels in BigQuery:

• compute.googleapis.com/reservation_name
• compute.googleapis.com/reservation_project_id

For more information about how to share reservations with first-party (1P) products in Vertex AI, see the following:

• Use reservations with training
• Use reservations with online inference
• Use reservations with Vertex AI Workbench instances

The following example queries show how to use these labels to get detailed visibility into your reservation usage, monitor how instances consume these reservations, and perform in-depth cost analysis.

List SKU types consuming a reservation for a 1P product in Vertex AI

If you create and share a reservation with Vertex AI for a training job, the following query returns the list of SKU types for the current date.

SELECT DISTINCT
  t.sku.description AS sku_description
FROM
  project-ID.dataset.gcp_billing_export_resource_v1_X AS t
LEFT JOIN
  UNNEST(t.system_labels) AS sl_reservation
  ON sl_reservation.key = 'compute.googleapis.com/reservation_name'
WHERE
  TRUE
  AND t.service.id = '6F81-5844-456A'
  AND DATE(t.usage_start_time) = CURRENT_DATE()
  AND EXISTS(
    SELECT 1
    FROM UNNEST(t.Tags) AS tag
    WHERE
      tag.key = 'gcp-product'
      AND tag.value = 'vertex_ai'
      AND tag.inherited = TRUE
  )
  AND sl_reservation.value = 'RESERVATION_NAME'
  AND t.resource.name IS NOT NULL
  AND t.resource.name != ''
ORDER BY sku_description ASC;

Sample results

The results show that the training job uses three SKU types within the Vertex AI service.

Analyze SKU statistics by reservation for a 1P product in Vertex AI

The following query helps you understand the distribution of instances and their usage costs associated with a reservation for a first-party (1P) product in the current month.

SELECT
  t.sku.description AS sku_description,
  t.usage_start_time AS usage_start_time,
  t.usage_end_time AS usage_end_time,
  SUM(t.usage.amount) AS total_usage_amount
FROM
  project-ID.dataset.gcp_billing_export_resource_v1_X AS t
LEFT JOIN
  UNNEST(t.system_labels) AS sl_reservation
  ON sl_reservation.key = 'compute.googleapis.com/reservation_name'
LEFT JOIN
  UNNEST(t.system_labels) AS sl_project_id
  ON sl_project_id.key = 'compute.googleapis.com/reservation_project_id'
WHERE
  TRUE
  AND t.service.id = '6F81-5844-456A'
  AND t.invoice.month = '202602'
  AND EXISTS(
    SELECT 1
    FROM UNNEST(t.Tags) AS tag
    WHERE
      tag.key = 'gcp-product'
      AND tag.value = 'vertex_ai'
      AND tag.inherited = TRUE
  )
  AND sl_reservation.value = 'RESERVATION_NAME'
  AND sl_project_id.value = 'PROJECT_ID'
  AND t.resource.name IS NOT NULL
  AND t.resource.name != ''
GROUP BY sku_description, usage_start_time, usage_end_time;

Sample results

The following output shows how SKUs use the reservation and provides a breakdown of the corresponding usage costs.

Analyze reservation usage for Vertex AI in shared projects

If you use a reservation shared across multiple owner projects where Vertex AI jobs are active, you can use the following query to analyze and break down your usage costs by individual project.

SELECT
  t.project.id AS consumer_project_id,
  t.sku.description AS sku_description,
  t.usage.unit AS usage_unit,
  SUM(t.usage.amount) AS total_usage_amount
FROM
  project-ID.dataset.gcp_billing_export_resource_v1_X AS t
LEFT JOIN
  UNNEST(t.system_labels) AS sl_reservation
  ON sl_reservation.key = 'compute.googleapis.com/reservation_name'
LEFT JOIN
  UNNEST(t.system_labels) AS sl_project_id
  ON sl_project_id.key = 'compute.googleapis.com/reservation_project_id'
WHERE
  TRUE
  AND t.service.id = '6F81-5844-456A'
  AND EXISTS(
    SELECT 1
    FROM UNNEST(t.Tags) AS tag
    WHERE
      tag.key = 'gcp-product'
      AND tag.value = 'vertex_ai'
      AND tag.inherited = TRUE
  )
  AND sl_reservation.value = 'RESERVATION_NAME'
  AND sl_project_id.value = 'PROJECT_ID'
GROUP BY
  consumer_project_id, sku_description, usage_unit
ORDER BY consumer_project_id, sku_description, usage_unit;

Sample results

The following output shows the contribution of each project to total usage costs while consuming the reservation.

Related topics

For information related to exported Cloud Billing data, see the following:

• Set up Cloud Billing data export to BigQuery.
• Example queries for Cloud Billing data export to BigQuery.

For details of the cost and pricing reports available in the Google Cloud console, see the following:

• View your Cloud Billing reports and cost trends.
• View and download the cost details of your invoice or statement.
• View and download prices for Google's cloud services.
• Understand your savings with cost breakdown reports.
• Analyze the effectiveness of your committed use discounts.
• View your cost and payment history.