Snowflake announced general availability of pg_lake at its Summit 2026 conference, an open-source Postgres extension, open sourced in November 2025 and made natively available inside Snowflake Postgres in March 2026, that gives a standard Postgres instance direct, zero-ETL read and write access to Apache Iceberg tables stored in a data lakehouse, without requiring a separate ETL pipeline to move data between the operational database and the lakehouse. This directly targets one of the more persistent architectural headaches in modern data stacks: teams running Postgres for transactional workloads and a separate Iceberg-based lakehouse for analytics have typically needed dedicated pipelines, often built with tools like Fivetran, Airbyte, or custom CDC connectors, to replicate data from Postgres into the lakehouse on some schedule, introducing both replication lag and yet another system that needs monitoring and maintenance. pg_lake collapses that gap by letting Postgres itself read and write Iceberg tables natively, meaning a query against the lakehouse and a query against Postgres's own tables can potentially live in the same SQL session rather than requiring separate tooling and a data-freshness compromise. Because pg_lake is a genuinely open-source extension rather than a Snowflake-proprietary feature, it's also installable on any standard Postgres instance, not just Snowflake Postgres, which is a meaningful signal about open table format standards: Iceberg interoperability is increasingly treated as infrastructure that any database vendor benefits from supporting broadly rather than a proprietary lock-in lever, a notable contrast to how data warehouse vendors have historically approached storage format compatibility.