Snowflake launched public preview support for Apache Iceberg v3 on March 4, 2026, introducing advanced features including geospatial data types, deletion vectors, and row-level lineage for both Snowflake-managed and external Iceberg tables. This update expands interoperability for streaming, batch, analytics, and AI workloads, supporting compliance in regulated environments.

Background

Open data formats such as Apache Iceberg are designed to promote choice, portability, and collaboration across engines, clouds, and vendors. Snowflake supports full bidirectional interoperability via the Iceberg REST Catalog, enabling downstream engines to access Snowflake-exposed Iceberg tables and Snowflake to federate to external Iceberg catalogs, eliminating vendor lock-in. This approach creates a neutral, standards-based metadata layer and promotes governance portability without proprietary integrations.

Details

Snowflake's Iceberg v3 preview adds data types such as geography, geometry, nanosecond, and variant, as well as default column values. It introduces deletion vectors for improved write efficiency and row-level lineage for governance and audit purposes, available for both internal and external Iceberg tables. The preview release integrates Iceberg v3 features across the Snowflake ecosystem, including streaming and batch ingestion, analytics, machine learning, AI orchestration, business continuity, disaster recovery, and catalog integrations. Snowflake's Horizon Catalog delivers unified governance across open table formats like Iceberg and Delta Lake, alongside relational sources. The catalog centralizes access controls, lineage, tagging, and risk monitoring for external engines such as Spark, Trino, and Flink via standard Iceberg REST APIs and vended credentials, ensuring consistent governance. This architecture benefits enterprises deploying Copilot-style AI agents or orchestrating model workflows by providing transparent data provenance and consistent policy enforcement, key requirements for auditability and interoperability in regulated environments.

Outlook

As Iceberg v3 support advances toward general availability, organizations can run AI-driven pipelines on governed, open data platforms while maintaining data lineage and regulatory compliance. Future updates may further integrate AI catalog semantics and agent-aware governance to streamline enterprise AI deployment in regulated sectors.