dbt Positioned as Multi-Cloud Control Plane
Tristan Handy, CEO of dbt Labs, on dbt’s multi-cloud tailwinds
Iceberg turned multi cloud from a budgeting nightmare into a real architecture choice. Before open table formats, a team that wanted Snowflake, Databricks, or BigQuery to work on the same data usually had to copy the same tables into each system, then pay storage and compute twice and reconcile drift when one copy updated before another. Iceberg gave vendors a shared table format so pipeline tools like dbt and Fivetran could write once and let multiple engines read the same underlying data.
-
This matters most for dbt because dbt sits in the transformation layer where business logic gets defined. If dbt writes models natively to Iceberg, the SQL, tests, lineage, and metadata can sit above any one cloud instead of being trapped inside Snowflake or Databricks specific storage formats.
-
The strategic signal was strong enough that even Databricks, which built Delta Lake, bought Tabular in June 2024 to bring Iceberg creators in house. Snowflake also made Iceberg storage generally available in June 2024. That is what made the format feel like a standard instead of an experiment.
-
The closest analogy is shipping containers. Before containers, moving goods between truck, rail, and ship meant repacking cargo each time. Iceberg is the container for analytics data. It standardizes the table layer so clouds compete on compute, governance, and developer workflow, not on locking up the bytes themselves.
From here, the winners are likely to be the tools that control logic rather than storage. As Iceberg adoption spreads, more value shifts to the control plane that defines models, permissions, lineage, and orchestration across clouds. That pushes dbt toward a stronger role as the neutral layer connecting warehouses, lakehouses, and governance systems.