Cloud data platform vendor Snowflake has made its set of PostgreSQL extensions open source in a bid to help developers and data engineers integrate the popular open source database with its lakehouse system.

Snowflake said pg_lake would allow developers and data engineers to read and write directly to Apache Iceberg tables from PostgreSQL, thereby cutting out the need to extract and move data. Iceberg is an open table format that advocates say allows users to bring their preferred analytics engines to their data without moving it; it is backed by Snowflake, Google, AWS and others.

Christian Kleinerman, Snowflake executive vice president, product, told The Register that making the extension open source would allow developers who use PostgreSQL in their stack to turn the database into an interface to manage an open lakehouse. The lakehouse concept was introduced by rival Databricks five years ago to describe a system that manages structured (data warehouse) and unstructured (data lake) workloads on a single system.

Kleinerman said: “One of the most common use cases for developers [will be] to build applications against PostgreSQL and then [move] or copy the data for analytics into either a data platform like Snowflake or increasingly, an open data lakehouse like Iceberg tables on S3 Tables in [AWS] or Microsoft Onelake [in Fabric]… that data now becomes available for analytics.”

Available under the Apache license, the extensions were developed by PostgreSQL specialist startup Crunchy Data before it was acquired by Snowflake for $250 million in June this year.

In a blog post, Craig Kerstiens, Snowflake software engineering director, said pg_lake would allow developers to manage Iceberg tables directly in PostgreSQL by introducing a new Iceberg table type where PostgreSQL acts as the catalog. It would also allow developers to query raw data files in the data lake or external Iceberg tables, Delta tables, and various geospatial file formats from PostgreSQL.

Robert Kramer, Moor Insights & Strategy vice president and principal analyst, said giving PostgreSQL users a direct path into Snowflake’s lakehouse and AI capabilities without forcing architectural disruption is a smart move.

“Most organizations are not ripping out PostgreSQL — and Snowflake clearly understands that. Pg_lake lowers the barrier for PostgreSQL teams to gradually adopt Snowflake for high-value analytics and automation, rather than treating it as an all-or-nothing platform decision. I anticipate incremental adoption, but real traction over time, especially as teams blend operational databases with governed AI execution.”

• Salesforce facing multiple lawsuits after Salesloft breach
• Google fuses SQL, Python, and Spark in Colab Enterprise push
• Snowflake builds Spark clients for its own analytics engine
• ‘It isn’t designed to solve privacy concerns,’ Grafana CTO says of Bring Your Own Cloud

Snowflake also announced the general availability of Snowflake Intelligence, an AI agent it says offers users the ability to answer complex questions in natural language and put insights at every employee’s fingertips. It has made additions to its Horizon data catalog too.

Across the board, Kramer said Snowflake may still need to improve in terms of scale, monitoring, and real-world costs for agent workloads.

“Buyers might need some help understanding how Snowflake is different from Databricks and other cloud platforms. Snowflake is designed to be a platform where AI can work reliably and responsibly, not just for testing purposes. For customers who want to move from experimenting with AI to using it in real-world operations, this mindset is really important.” ®