dbt and Fivetran multi-cloud choke point
Tristan Handy, CEO of dbt Labs, on dbt’s multi-cloud tailwinds
This reveals that dbt and Fivetran sit in the narrowest choke point of the modern data stack, where raw data gets turned into the tables every downstream tool actually uses. Fivetran lands source data from SaaS apps and databases into warehouses and lakes. dbt then turns that raw data into cleaned, tested tables. When both layers support Iceberg, the same modeled data can be read across Snowflake, Databricks, BigQuery, Redshift, and S3 based lake setups instead of being rewritten cloud by cloud.
-
The practical point is not market share, it is workflow control. A typical team uses Fivetran to pull data from Salesforce, Stripe, or Postgres into a warehouse, then uses dbt SQL models to build tables like orders, MRR, or product usage. Those tables become the inputs for dashboards, AI, and reverse ETL tools.
-
That is why Iceberg matters so much. Before open table formats, multi cloud usually meant copying the same dataset into multiple systems, which added cost and created mismatches. In 2024, major data platforms moved toward Iceberg interoperability, and both dbt and Fivetran added Iceberg support, making shared access more realistic.
-
It also explains why dbt is pushing up from transformation into a broader data control plane. If business logic, lineage, metadata, and orchestration sit above any one cloud, dbt becomes the place teams manage how data is defined, while Snowflake and Databricks compete to be the place data is stored and queried.
Going forward, the center of gravity moves toward the pipeline and control layers that can stay neutral across clouds. The winners will be the tools that let a company load data once, define it once, and reuse it everywhere, across analytics, operational apps, and AI workloads, without rebuilding the same dataset for each platform.