Skip to content
FinOps workspace
COST LEDGER / DECISION GUIDE

BigQuery cost control: pricing, queries and storage

Updated: 18 November 2026Request a quote

BigQuery costs come from query compute, stored data and any additional services used. Estimate on-demand scans separately from slot capacity and storage, then reconcile the Google BigQuery costs to the billing account and the business workload.

Server cabinets in the mint and dark green Cost Ledger visual style
Define the workload before comparing its cost.

When the workload includes models, check the distinct operation and billing routes in the guide to BigQuery ML training and inference pricing.

For the specific scan-size question, use the cost of a 12 gig BigQuery query to compare decimal GB with GiB and identify when the on-demand calculation applies.

BigQuery cost control: choose what each guardrail stops

Choose BigQuery cost controls against the processing model and the work that must continue. Google's cost-control guidance distinguishes on-demand query limits from reservation capacity settings. Maximum bytes billed rejects an on-demand query whose estimated scan exceeds the setting. Daily custom query quotas apply to on-demand usage per project or per user within a project. Google warns that these quotas are approximate: a query can occasionally exceed the quota, so they are not a strict bytes-processed or spending ceiling. For capacity workloads, review baseline slots, autoscaling limits and idle-slot sharing and scaling mode together. A capacity setting does not cap storage or the entire billing account.

DecisionEvidence to requestAcceptance check
Exploratory on-demand queriesJob settings and a representative rejected queryThe analyst understands the rejection and approval path
Daily project or user allowanceCritical schedules, approximate enforcement, quota scope and ownershipRequired reporting has an agreed escalation procedure
Reservation capacityBaseline, autoscaling limit, idle-slot sharing, scaling mode, assignments and demand profileThe analytics owner accepts throughput and completion behavior
Billing notificationsBudget scope, thresholds and recipientsA named owner investigates and records the response

An alerts-only Cloud Billing budget sends notifications but does not automatically cap spending or usage. Keep that monitoring function distinct from query rejection. Set the wider operating boundary in the GCP cost review and assign the investigation workflow through cost anomaly detection.

  1. Inventory the projects, processing models and business-critical outputs.
  2. Choose a control and record its owner, intended effect and exceptions.
  3. Test both permitted work and the rejection or notification case.
  4. Obtain analytics approval before enabling a restrictive setting.
  5. Review billed cost alongside delivered workload and unresolved alerts.

A cost-control review should leave a tested control register. Ask for configuration evidence, representative job references, approved exceptions and handover instructions. Use the FinOps proposal checklist to separate analysis from implementation. If models are involved, retain the separate BigQuery ML billing boundary.

Does a BigQuery budget stop queries when the budget is reached?

An alerts-only Cloud Billing budget does not stop queries or automatically cap spending. Use the applicable BigQuery query or capacity settings for the intended restriction, and give a named owner responsibility for budget alerts.

Can maximum bytes billed reject a clustered-table query that would cost less?

Yes. Google documents that the pre-execution estimate for a clustered table is an upper bound. A query can be rejected even when its actual billed bytes would have stayed within the setting. Have the analytics owner review that tradeoff before selecting a limit.

How to estimate BigQuery storage and query costs

Separate the compute estimate from the storage estimate. The official BigQuery pricing page publishes on-demand analysis per TiB and capacity per slot-hour. It lists the first 1 TiB of on-demand query processing per month as free. Check the selected location, currency, contract and remaining allowance before applying a rate. A file size or the size of a query result is not a billed scan estimate.

Cost lineCalculation to prepareEvidence to retain
On-demand queriesBillable TiB after applicable allowance × regional rate per TiBQuery estimate, bytes billed, account allowance and query frequency
Capacity computeBilled slot-hours × applicable edition and purchase rateReservation configuration, autoscaling, commitments and workload profile
Dataset storageBillable GiB-time in each storage category × its regional rateDataset billing model, retained volume and active or long-term category
Additional servicesApplicable usage × the relevant SKU rateIngestion, transfer, extraction or other separately billed features used

For 1 TiB billed, multiply the eligible billable quantity by one regional TiB rate. A decimal 1 TB is about 0.9095 TiB; 100 GB is about 0.09095 TiB, while 100 GiB is 0.09765625 TiB. Those conversions describe quantities before free allowance, rounding and any query-specific minimum. BigQuery documents rounding to MB and minimum processing charges per referenced table and query on its price sheet. Preserve the bytes returned by the job rather than replacing them with a rounded file size.

The cost of storing data in BigQuery depends on the dataset model. Google distinguishes logical and physical storage. Time travel and fail-safe are included in the logical storage base rate but charged separately under physical billing. Active and long-term storage also have different rates. Query and storage are separate meters, so reducing scanned columns can lower on-demand query charges while leaving retained storage unchanged.

BigQuery cost management tools and their limits

  • Use the query validator or a dry run to inspect estimated bytes before execution, with the documented row-level security caveat.
  • Use maximum bytes billed for individual on-demand queries and daily query quotas where appropriate. These can reject required work, so agree an escalation path.
  • Inspect the INFORMATION_SCHEMA.JOBS view for job-level evidence, including bytes billed and slot usage. Bytes-billed fields are empty for queries that read tables with row-level access policies. Exclude parent SCRIPT rows when summing child-job metrics to avoid double counting. Validate access and region scope; slot usage is not itself billed reservation capacity.
  • Use Cloud Billing reports for billed cost trends, and the TABLE_STORAGE view to examine current logical and physical storage volumes. TABLE_STORAGE is a delayed snapshot, not a monthly usage history; reconcile storage over the billing period rather than pricing one snapshot as the whole month.

Google documents the pre-execution controls in its cost estimation and control guide. Include these native tools in a FinOps tooling evaluation before buying additional software. Set the reporting boundary with the GCP cost review; use the Cloud Storage pricing guide for object storage outside BigQuery.

Questions about Google BigQuery costs

How much does it cost to process 1 TB in BigQuery?

For on-demand analysis, convert the actual billable bytes to TiB, subtract any applicable remaining free allowance and multiply by the regional rate in your billing currency. A decimal 1 TB is about 0.9095 TiB; 1 TiB is a different quantity. Under capacity pricing, use slot charges rather than a per-TB formula.

How much does it cost to run a 100 GB query in BigQuery?

If 100 GB means exactly 100 billion billed bytes, that is about 0.09095 TiB before applicable allowance and billing rules. Multiply the billable amount by your regional per-TiB rate. A 100 GB table may produce a different scan quantity, and a capacity-priced job requires a slot-cost calculation.

What is the effect of query cost on Google BigQuery pricing?

Under on-demand pricing, actual billed bytes across executions affect compute cost; cached query results incur no query charge. Under capacity pricing, the edition, capacity purchased or scaled and billed duration drive compute charges. Storage remains a separate line.

What is the effect of storage cost on Google BigQuery pricing?

Storage adds a separate charge based on retained data, the dataset billing model, the storage category and location. Logical and physical models treat time travel and fail-safe differently. Query tuning does not automatically reduce stored data.

Which BigQuery cost management tools should a review use?

Combine pre-execution estimates, appropriate query controls, job evidence, storage metadata and Cloud Billing reports. Ask the review to explain the gap between a workload estimate and billed cost rather than treating a dashboard subtotal as the invoice.

BigQuery pricing: two query processing models

Google documents on-demand and capacity-based query pricing. On-demand compute is billed for data processed; capacity pricing is based on compute resources represented by slots. Dataset storage has its own billing model. Record the current model before selecting an optimization measure.

DimensionEvidenceQuestion
On-demand processingBilled query usage and job historyWhich jobs repeatedly read unnecessary data?
CapacityEdition, reservations and observed workloadDoes capacity match concurrent demand?
StorageDataset billing model and retained dataWhat data must stay available?
Business outputReports, jobs and freshness requirementWhich work is needed and by when?

Begin with the Google Cloud cost review to reconcile projects to the billing account. For BigQuery cost optimization, keep scheduled queries and dashboard refreshes in the same workload inventory. Assign the owner of a recurring job before reducing its frequency. A report may have an operational obligation that is not visible in a query cost chart.

Put query controls next to the right pricing model

The official cost-control documentation explains dry runs and a maximum-bytes-billed setting for on-demand queries. It also notes that a LIMIT does not reduce the scanned data for a non-clustered table. Review query selection and filters with the analytics team. A small displayed result is not evidence of a small processing charge. For tables protected by row-level security, Google documents that dry runs can report zero bytes to prevent side-channel attacks; do not interpret that as a free query.

  • Identify repeated jobs and the data each job needs.
  • Check query estimates before execution for exploratory work.
  • Apply the appropriate cost limit without blocking required reporting.
  • Validate the output after changing data selection or scheduling.

Keep cost controls compatible with the reporting requirement. Rejecting an expensive query is useful only if the team understands which output will be delayed or unavailable. Record that tradeoff and the escalation owner. A finance report and an ad hoc analyst experiment may need different rules even when they use the same dataset.

Compare capacity with the actual demand profile

Use the current BigQuery price sheet for rates and conditions. Ask the review to distinguish a rate change from a workload change. For capacity choices, describe concurrency and delivery constraints rather than assuming all jobs can be spread evenly across the day. Preserve the project and reservation boundary in the financial model.

  1. Reconcile job and storage costs for the selected period.
  2. Group workloads by owner and required completion window.
  3. Test query or capacity changes against output correctness and service limits.
  4. Measure the resulting billed cost for comparable analytics work.

Use cloud cost management to agree ownership and budgets. The FinOps tools comparison helps define which data a cost platform must retain. Include the cost of operating those reports and tools in the cloud total cost of ownership calculation when making an investment decision.

Prepare billed usage, job history, dataset configuration and reservation information for a quote. Clarify who can approve query changes and which jobs are business-critical. Ask for a register of findings with the query or workload reference, the evidence used, the expected effect and the validation required. Keep query content and sensitive data within the agreed audit boundary.

Server cabinets in the mint and dark green Cost Ledger visual style
Connect analytical workload evidence with compute and storage charges.

Questions before a cost review

How much does BigQuery cost?

It depends on the processing model, usage, storage configuration and location. Calculate compute and storage separately from the current Google price sheet, then apply your contract terms.

Why is BigQuery so expensive?

Investigate repeated or broad scans, recurring jobs, retained data and capacity that does not match demand. Determine which dimension drives your own bill before choosing a control.

Is GCP cheaper than AWS?

There is no universal answer. Compare an equivalent workload, location, availability requirement and operating model. Include engineering and migration costs, and use your contract rates in both scenarios.

Get my FinOps quote