Supabase Pipelines Opens to Everyone, Bills You Even When Paused

Supabase Pipelines, the managed service that streams Postgres changes out to an external warehouse, is now in public alpha on every paid plan, according to the Supabase blog post published July 22, 2026. BigQuery is the only destination available to everyone at launch, and the metering starts the moment a pipeline exists, not when it starts moving data.

---

What shipped

Pipelines is Supabase’s packaged version of Supabase ETL, the open-source Rust change-data-capture engine the company introduced earlier this year. A pipeline takes a complete initial copy of the tables you select, then streams inserts, updates, deletes and truncates to the destination in near real time, with at-least-once delivery, per the Supabase docs. The public alpha adds schema change support, so a column added or altered on the Postgres side now propagates to the BigQuery table automatically instead of breaking replication, plus a faster initial sync.

Public alpha is a rollout, not a switch. Being on a paid plan makes you eligible, but Supabase says access is still being enabled gradually, and you request it from the Database > Replication page in the dashboard. ClickHouse, Snowflake and DuckLake aren’t generally available yet either — they sit behind an early-access request form while BigQuery is the only destination open to everyone.

The pricing has a gotcha

Pipelines bills $0.053 per hour per active pipeline, $0.60/GB for the initial table copy, and $3/GB for data replicated after that, according to the same announcement. A pipeline is billable from the moment it’s created, and the hourly charge keeps running even while replication is paused or otherwise idle. Pausing a pipeline stops new rows from moving. It does not stop the meter.

That’s a different model from most sync tools, where a paused job is a free job. A pipeline built for a seasonal reporting job, or one temporarily paused while a downstream warehouse schema is reworked, keeps accruing the $0.053/hour charge the whole time it sits idle. Multiply that across a handful of pipelines left paused “for later” and the bill adds up quietly, with nothing in the product surface forcing you to notice until the invoice does.

Who should look now

If your app already writes to Postgres and you’ve been hand-rolling a Fivetran-style job to mirror it into BigQuery, Pipelines now handles schema drift for you, which is the part most of those homegrown jobs get wrong first. Before turning one on, decide whether you actually need it running continuously or whether it’s just an occasional batch sync, because the billing model only rewards the former. And if you pause a pipeline instead of deleting it, put a reminder somewhere to actually delete the ones you don’t plan to resume.

---

Key Takeaways

• Supabase Pipelines, the managed CDC service built on Supabase ETL, reached public alpha on all paid plans as of the July 22, 2026 announcement.
• BigQuery is the only destination generally available; ClickHouse, Snowflake and DuckLake require an early-access request.
• This alpha adds automatic schema change propagation to the BigQuery destination and a faster initial sync.
• Pricing is $0.053/hour per active pipeline plus $0.60/GB for the initial copy and $3/GB for ongoing replicated data.
• The hourly charge runs from creation and continues while a pipeline is paused — pause it to stop replication, delete it to stop the meter.

---

The take

Shipping schema-change handling before general availability is the right call; drift breaking a warehouse sync silently is one of the more common ways these integrations fail in production. The billing model is the part that needs a second read before anyone flips a pipeline on. Charging from creation instead of from first byte replicated is a defensible way to price a managed service with fixed compute overhead, but calling a paused pipeline anything other than “still costing you money” would have saved someone a support ticket. Read the pricing page before the dashboard, not after.