BigQuery Pricing FAQ
Straight answers on BigQuery Editions, Autoscaling, and Compressed Storage. What changed, who saves, and what to watch for.
On-demand price increase since July 2023
Discount with a 3-year Editions commit
Slot increments used by the Autoscaler
Minute minimum billing on Editions queries
The short version
What actually changed in BigQuery pricing
Three shifts to understand before you touch your reservations or storage settings.
On-demand got pricier
On-demand analysis pricing went up roughly 25%. Spiky and short-running workloads still tend to fit it best.
Editions replaced flat-rate
Monthly, yearly, and Flex slots are gone. Existing commits run to term, then move to Editions with Autoscaling.
Compressed Storage arrived
Bill on physical bytes instead of logical bytes. Big wins for string-heavy data, smaller wins for integer-heavy tables.
Editions and Autoscaling
Editions can be cheaper, or much more expensive
Autoscaling bills you for slots allocated, not slots used. It scales in increments of 100 with a one-minute scale-down. A 10-second query on Editions still costs you a full minute of slots.
Long-running, steady workloads with lots of data scanned tend to win on Editions. Short, spiky queries usually stay cheaper on-demand. The practical move: split workloads across projects and pick the pricing model per project.
Compressed Storage
Pay for physical bytes, not logical ones
Logical Storage bills you on the uncompressed size of your data. Compressed Storage bills you on the physical bytes on disk: $0.04/GB active and $0.02/GB long-term. The per-GB rate is higher, but the GB count is usually much lower.
Same APIs, same query behavior, no performance hit. One catch: on-demand queries are still billed on uncompressed bytes scanned, so the savings only show up on storage, not compute.
Frequently asked
questions
Is the new Editions model more expensive than legacy flat-rate?
It depends on the workload. If you had idle slots on flat-rate, Editions plus Autoscaling can drop your compute bill. Compressed Storage can lower the storage side. Look at it as TCO, not a single line item.
What is a BigQuery commit, and is it the same as a GCP commit?
It is separate. A BigQuery commit gives you a better slot/hour rate. A 3-year commit unlocks roughly 40% off analysis costs. Storage pricing is not discounted.
Should I run a POC before switching off on-demand?
Yes. Editions behavior, especially Autoscaling minimums and the one-minute scale-down, can change costs in non-obvious ways. Run a POC on a representative workload before committing.
Which workloads fit Editions, and which fit on-demand?
Long-running, steady workloads scanning lots of data are best for Editions. Short, spiky queries are usually cheaper on-demand because Autoscaling allocates 100-slot blocks and bills for a full minute minimum.
How do I know if Compressed Storage will save money?
It depends on compressibility. String-heavy data compresses well. Integer-heavy data compresses poorly. We have a query in this GitHub gist that compares your options: https://gist.github.com/sayle-doit/264d28dd990c478beb90b90ac3923681
Is there a tool to model the impact on my costs?
Yes. Duplicate our BigQuery Current Usage Analysis Tool sheet and follow the instructions: https://docs.google.com/spreadsheets/d/1Sv6SzATtAyGYID2Wg0qGdcpLfe7YdOsH4dFKHHtS59c/edit
Frequently asked
questions
Can I mix and match different Editions across projects?
Yes. One project can run on-demand, another on Standard Edition, another on Enterprise. Splitting workloads across projects is one of the best optimization levers you have.
Do I need to re-architect my data warehouse or change SQL?
SQL runs as-is. If you have code that buys Flex slots dynamically, that stops working. Terraform support is in progress, so use ignore_changes in the meantime. You may still want to refactor to reduce total slot consumption.
I had a flat-rate commit before July 5th. Can I still use Editions on top of it?
Yes. Flat-rate commits convert to Enterprise Edition at your committed price for the duration. Slots above your baseline are billed at the standard Enterprise rate. Autoscaler can scale up to a max you set.
Can I combine Compressed Storage with an annual flat-rate commit?
No. Google's tooling actively blocks Compressed Storage when a flat-rate reservation exists. You need Editions or pure on-demand projects.
On on-demand, am I billed on compressed or uncompressed bytes scanned?
Uncompressed. There is no compute cost advantage to querying compressed data from an on-demand project. The savings are on storage only.
Does switching to Compressed Storage break long-term storage discounts?
No. The long-term storage mechanism still applies to both compressed and uncompressed data.
Will short queries really be billed for a full minute on Editions?
Yes. Editions queries hit a one-minute minimum due to the Autoscaler's one-minute scale-down. A 10-second query costs a minute of slots.
How do I figure out how many slots I need before picking an Edition?
In Google Cloud Console, open Monitoring, go to Metric Explorer, and search for slot metrics. DoiT customers can open a ticket and review this with a BigQuery specialist.
Can I cap maximum spend with the Autoscaler?
Yes. Set a baseline of slots you always use and a maximum at your comfort ceiling. If you have idle stretches, set the baseline to 0 so you are not paying for unused allocated slots.
How are dbt costs affected?
dbt on BigQuery is compute-heavy. Compute is now more expensive on Editions, so dbt workloads typically cost more unless you tune slot usage and isolate them in the right project.
Time-travel and fail-safe storage charges on Compressed Storage. Can I turn them off?
Time-travel cannot be disabled, but you can drop the window from 7 days to 3. Fail-safe can be disabled. Restoring from fail-safe requires a Google Cloud support ticket.
Are some workloads more expensive on Editions than on-demand even when they were free before?
Yes. BigLake metadata refreshes, for example, use slot-hours but scan zero bytes, which means free on-demand. On Editions, you pay for the slots. Isolate those jobs in an on-demand project.
Have BigQuery pricing questions?
Work with a BigQuery specialist on your specific workloads.