Warming up to Iceberg: 2026 makes Iceberg easier
A round of changes across Snowflake, dbt, and DuckDB in 2026 turned Apache Iceberg from a multi-week infrastructure project into something a small analytics team can set up in an afternoon.
Overview
Covers running dbt transformations without a Snowflake virtual warehouse, loading data into Snowflake-managed Iceberg tables and transforming via dbt-duckdb. Discusses DuckDB/dbt v2/Snowflake catalog improvements simplifying Iceberg setup, automatic governance (RBAC, masking, lineage), and DuckLake as an alternative.
What you'll take away
- 01
Becoming a user of Apache Iceberg no longer requires becoming an infrastructure expert first: Snowflake's built-in Horizon catalog and a new catalog_name entry in dbt_project.yaml removed most of the historical setup burden in 2026.
- 02
Iceberg gives you a new way to cut costs: instead of only tuning warehouses, you can run some queries somewhere else entirely. In the demo, DuckDB read the same governed tables dbt had built in Snowflake, used no Snowflake compute, and still followed Snowflake's row-level access policies and column masking.
- 03
For teams already on Snowflake, Horizon still has the edge over newer alternatives like DuckLake, since governance, cataloging, and AI tooling come built in rather than requiring a separate catalog to maintain.
Session speaker
Jeff Skoldberg
Green Mountain Data Solutions