Supabase Pipelines brings managed Postgres-to-BigQuery CDC into public alpha

Supabase Pipelines is now in public alpha as a managed CDC service powered by the open-source Supabase ETL engine. It performs an initial copy of selected Postgres tables and then continuously applies inserts, updates, deletes and truncates to BigQuery with at-least-once delivery. Builders can choose tables, schemas, columns and row filters; supported schema changes can be propagated automatically. During public alpha, Pro and Team pricing is $0.053 per configured pipeline-hour, $0.60 per GB for initial-sync data and $3 per GB for ongoing replicated row data, with destination-provider charges separate.

This gives Supabase users a managed path for moving operational Postgres data into an analytical warehouse without running heavy reporting queries against the primary database or operating their own CDC stack. The builder consequence is as much economic and operational as functional: configured pipelines keep accruing hourly charges even when stopped, ongoing change volume is billed separately, and BigQuery costs still sit on top. The service is also explicitly alpha, so teams gain a useful workload-isolation primitive but should design around destination, region and schema limitations rather than treating it as mature warehouse infrastructure.

A pipeline first copies existing rows, then switches to Postgres logical replication for ongoing changes. Supabase says delivery is at least once, replication resumes from the last acknowledged position after restarts, and the Dashboard exposes status, lag, table state, metrics, logs and errors. That removes a substantial amount of connector operation for teams that otherwise would run Debezium, custom logical-replication consumers or another managed CDC service.

Pipelines builds on Postgres publications. Builders can replicate selected tables, schemas or all tables, restrict columns, filter rows with a WHERE clause and work with partitioned tables. BigQuery requires source tables to have primary keys and requires those key columns to be included in the publication, so existing data models can constrain what is eligible without changes.

The public alpha can detect and apply supported changes such as adding, removing and renaming columns and changing nullability or defaults. Supabase still labels schema-change support beta and notes that broader type-change support remains on the roadmap. Custom data types are replicated as strings, generated columns are not supported, and data is replicated as-is rather than transformed in flight.

A configured pipeline is charged hourly even when paused or inactive; deleting it ends the hourly charge. Initial copies cost $0.60 per GB of logical Postgres row data accepted by the destination, while later changes cost $3 per GB. Supabase’s own example shows that ongoing replicated data can dominate the bill at meaningful change volume, and BigQuery ingestion, storage and CDC compute charges remain separate.

Managed Pipelines currently run in AWS eu-central-1 in Frankfurt. Supabase recommends placing the destination close to that region to reduce replication lag and improve throughput. That means a project or BigQuery dataset far from Europe may experience a less favorable latency path even if the source database itself is located elsewhere.