Google drops BigQuery fees for its own data sources
The service loads data into the warehouse without code. Alphabet sources incur no orchestration fee, while moving data out requires another product.
Google is dropping the fee previously charged for each transfer into BigQuery for most of its own data sources. Teams loading advertising data from Google Ads or Analytics into the warehouse will now pay only for storage and queries. The trade-off lies elsewhere: the service writes exclusively to BigQuery. There is no way back.
Key takeaways
- The orchestration fee is being removed for Google Ads, Analytics 4, YouTube, object storage and migration connectors. Paid sources such as Salesforce or Oracle are still billed by slot-hour, while storage and queries remain chargeable for every source.
- Google presented the pricing change and current features in a roundup on 7 August 2026. The price is new. Iceberg destinations, MCP servers and connectors already existed.
- The service writes only to BigQuery and offers no route in the opposite direction. A second warehouse or another cloud requires a different tool, such as Fivetran or Airbyte.
- Preview features come without a service-level commitment. Iceberg ingestion, the MCP server and the new SaaS connectors remain outside the Service Level Agreement.
Related:SAP to BigQuery: A data bridge with architectural implications / Iceberg won the format war. Now the catalog matters
Google makes data transfers cheaper
Previously, warehouse teams paid an orchestration fee for every transfer into BigQuery, on top of storage and queries. That is over for Google’s own sources. Product Manager Vignesh Rajamani and Engineering Manager Pallavi Palleti outlined the current Data Transfer Service, or DTS, in the Cloud Blog on 7 August 2026. They also introduced the new pricing rule. The service brings data from advertising accounts, SaaS applications and object storage into BigQuery without anyone coding a pipeline or running a second orchestration tool. It replaces the nightly export script that many marketing departments have used for years to load advertising figures into their database.
The main change is the price. The service has loaded Google Ads data for years and Analytics 4 data since last year. The two components Rajamani and Palleti present as previews were also already running: files from Amazon S3 or Azure land directly in Iceberg tables, the open table format that multiple warehouses can read. A server using the Model Context Protocol lets AI agents create transfers themselves without a person opening the console.
Who benefits from the discount
The pricing change divides sources into two groups. Google no longer charges an orchestration fee for its own services, object storage or migration connectors. For generally available paid sources, DTS bills by slot-hour, meaning the compute time BigQuery reserves for the transfer. A slot is a unit of compute time in BigQuery; Google reserves up to 20 per hour of operating time for a transfer. A daily Salesforce extract taking one hour therefore incurs several hundred slot-hours per month, while the Ads import alongside it now costs nothing.
| Source group | Orchestration fee | Examples |
|---|---|---|
| Google services (except Google Play) | None | Google Ads, Analytics 4, Campaign Manager, YouTube |
| Object storage and migration connectors | None | Cloud Storage, Amazon S3, Azure Blob, Redshift, Teradata, Snowflake |
| Preview connectors | None until general availability | SQL Server, Shopify, Klaviyo, HubSpot, Mailchimp |
| Generally available paid sources | By slot-hour, with a guideline of up to 20 per hour of operating time | Salesforce, ServiceNow, Facebook Ads, MySQL, PostgreSQL, Oracle |
| Google Play | Fixed monthly fee per app package name | Play reports |
Source: Google Cloud, Data Transfer Service pricing and announcement dated 7 August 2026. BigQuery storage and queries incur additional charges for every source.
The question for managers is therefore: does most of the data come from Google’s own advertising and analytics ecosystem or from paid third-party systems? In the first case, DTS has become the cheapest choice: transfers now cost nothing. In the second, slot-hours and merge jobs for incremental updates eat up the discount. BigQuery itself remains on every bill: storage and queries, plus any data transfer charges another cloud provider levies for extracting the data.
There is no way back
DTS writes in one direction only. The service cannot reach a second warehouse, another cloud or the original source. The affected users are easy to identify: a corporation running a Snowflake or Databricks warehouse alongside BigQuery, an agency whose next customer uses Azure, a team that may one day want to move its advertising data to another system. For them, free transfers provide a way in without a way out.
Then there is what the service does not do in the warehouse overnight. It fetches and loads data, nothing more: teams that need to transform rows, propagate deletions or replicate changes every second require other Google tools. Google also provides an availability commitment only for the established components; the new ones run in preview without a guarantee.
Fivetran and Airbyte sell a way out
The corporation with a second warehouse, the agency with Azure customers, the team that wants to take its data elsewhere: they turn to the providers whose business model Google is now challenging. Fivetran and Airbyte deliver to every major warehouse and charge for it: Fivetran per row moved, Airbyte by volume or as self-hosted software. A way out here means paying for broader coverage.
How long Google’s pricing will hold once the preview connectors become generally available remains unclear. Google says only that preview sources remain free until general availability. It will decide what follows when they launch. Anyone choosing the way in today should know the price of getting out.
Frequently asked questions
What’s new in the Data Transfer Service?
The price. The orchestration fee is being removed for Google’s own sources, object storage and migration connectors. Iceberg destinations, MCP servers and the new SaaS connectors were already in preview before the blog post on 7 August 2026.
What is a BigQuery slot?
A unit of compute time in BigQuery. Queries and transfers reserve slots, and Google bills paid transfers by slot-hour.
What’s the PostgreSQL transfer limit without a primary key or index?
Without primary-key or indexed columns, a transfer moves at most two million records per table. An order or event table can quickly reach that limit.
Editor's Picks
cloudmagazinDuckDB and MotherDuck kill the warehouse theatercloudmagazinReal-Time Demand Forecasting: The Pipeline DecidescloudmagazinCross-cloud Lakehouse and Egress-Caching: What Google Cloud Next 2026 Changes for DACH ArchitectsMore from the MBF Media Network
MyBusinessFutureOne year of mandatory ePA: connected, but barely usedDigital ChiefsThe quiet infrastructure crisisSecurityTodayKillSec dismantled: 16-year-old suspected of leading groupImage Source: AI-generated (August 2026)
Translated from the German original using artificial intelligence. The German version is authoritative.
Also available in
