Choosing between Snowflake and Databricks for a modern Lakehouse architecture implementation?

All answers to this question.

If your team is primarily SQL-heavy, Snowflake’s ease of use is hard to beat, especially with their recent Unistore and Iceberg table support which bridges the gap to a Lakehouse. However, if your long-term goal is heavy Spark-based ML and data science, Databricks provides a more "open" ecosystem with Delta Lake. We found that Databricks was slightly cheaper for massive batch processing, but the "virtual warehouse" scaling in Snowflake saved us more on the compute side for ad-hoc BI queries.

   Answered 2025-02-20 by Mary Collins

Have you evaluated the egress costs and vendor lock-in risks associated with moving all your raw data into a proprietary format versus using open standards like Apache Iceberg?

   Answered 2025-03-05 by Michael Taylor

• Michael, that's exactly why we are leaning toward Iceberg. We want the flexibility to switch engines if the pricing models change significantly next year. Both platforms are supporting it now, but the implementation nuances are what we're currently testing in our Proof of Concept phase.

     Commented 2025-03-12 by William Wright

Don't forget the "people" cost. Snowflake requires less DBA-style maintenance, while Databricks often needs more specialized engineering resources to manage clusters.

   Answered 2025-03-25 by Elizabeth Reed

• Elizabeth hits a great point. The "hidden" cost of a platform is often the salary of the four extra engineers you need to keep the complex Spark pipelines running smoothly.

     Commented 2025-03-28 by James Anderson