pg_duckdb Deep Dive: Postgres + DuckDB for Fast Analytics

Postgres is the world's favourite OLTP database, but the moment you ask it to crunch a billion rows of analytics it crawls. The classic answer was to ship data to a warehouse — Snowflake, BigQuery, ClickHouse. But that means a second system, a CDC pipeline, security audits, and the inevitable two-truths problem.

pg_duckdb is a Postgres extension, jointly developed by the DuckDB and MotherDuck teams, that embeds a full DuckDB engine inside your Postgres process. You keep one database, one connection string, one set of credentials, and you get a vectorized columnar engine that is 10x to 100x faster than Postgres's row-store executor for analytical queries. It can also read Parquet, CSV, JSON, Iceberg, and Delta Lake straight from S3, GCS, Azure Blob, and Cloudflare R2 — without ever moving the data into Postgres.