Databricks' Upstream Advantage Over Snowflake
Charles Chretien, co-founder of Prequel, on the modern data stack’s ROI problem
Databricks is winning because it owns the earlier step in the data workflow where more expansion paths begin. Managed Spark put it where raw data gets ingested, cleaned, and processed, so it could add warehousing, governance, ML tooling, and AI on top of an existing compute relationship. Snowflake started at the warehouse, which is closer to the finished table, so moving upstream into pipelines and operational databases is a tougher product and go to market move.
-
Databricks built around managed Spark, notebooks, and scheduled pipelines, which means customers were already using it to run heavy data jobs before they bought analytics. That made Databricks SQL an attach product, not a brand new wedge. By 2024, Databricks SQL was already at about $400M annualized revenue.
-
Snowflake grew by making cloud warehousing simple for analysts, then expanded into nearby layers like app building with Streamlit and AI analysis with Cortex. But those sit downstream of the warehouse, where there are fewer large adjacencies than upstream infrastructure like pipelines, processing engines, and operational databases.
-
The revenue mix shows the difference. Databricks was around $4B run rate with only about $1B from warehousing, which means most growth came from products beyond the warehouse. Snowflake was around $5B revenue with the large majority still tied to warehousing, showing a narrower base to cross sell from.
This points to a data market where the biggest winners start closer to where data is created and moved, then bundle more workloads around that core spend. As AI pushes companies to unify processing, storage, governance, and model tooling, upstream platforms like Databricks have the clearest path to keep absorbing categories that used to be sold as separate tools.