Clean Rooms now support shared output tables across all collaborators

Databricks has made shared output in Clean Rooms generally available (GA). Notebook runs within a clean room can now write tables directly to a shared schema that every collaborator can query immediately, even before the run finishes.

Use cases

Imagine two organizations collaborating in a Clean Room to calculate audience overlap or campaign attribution. Previously, output tables were accessible only to the runner upon completion. With shared output, you can:

• Distribute aggregated results and metrics to all participants simultaneously.
• Expose intermediate tables and operational logs while the notebook is still running.
• Connect shared analytical outputs directly into automated Lakeflow Jobs.

How to use

In your clean room notebook, build the three-part namespace using the parameters :cr_output_catalog and :cr_shared_output_schema:

CREATE TABLE identifier(:cr_output_catalog || '.' || :cr_shared_output_schema || '.audiencia_sobreposta') AS
SELECT
  origem.segmento_id,
  COUNT(DISTINCT parceiro.email) AS total_sobreposicao
FROM parceiro.base.perfis AS parceiro
JOIN criador.base.perfis AS origem
  ON parceiro.email = origem.email
GROUP BY origem.segmento_id;

What happens in this code:

1. The :cr_output_catalog parameter references the output catalog provisioned in the Unity Catalog metastore.
2. The :cr_shared_output_schema parameter routes table creation into the schema accessible by all collaborators.
3. The resulting table becomes immediately queryable for everyone in the room as soon as the cell finishes, even if subsequent notebook cells are still executing.

To write tables that should remain private to the executor, use :cr_output_schema instead.

Limitations and requirements

• Maturity: Generally Available (GA).
• Account setup: On Enterprise tiers (and Azure Premium), account admins must enable Shared Output from the Previews console. All collaborators in the clean room must have identical settings enabled.
• Clean room creation: Shared output must be activated during clean room creation; existing clean rooms cannot be retrofitted.
• Output catalog: Collaborators must create an output catalog via the UI (Catalog > Clean Rooms > Output > Create catalog) using EXECUTE CLEAN ROOM TASK permissions.
• Compute: Reading output tables requires serverless compute.
• Storage limits: Only tables are supported (no views or volumes). Tables are read-only and automatically purged after 30 days unless replicated to local storage.

Using it in a pipeline

You can schedule clean room notebooks using Lakeflow Jobs. Once the shared tables are written, downstream workflows on each collaborator's side can ingest the output into internal Silver or Gold layers for local reporting.