Snowflake announced in March 2026 the public preview of its support for Apache Iceberg v3, the latest version of the open-source table format that enables data interoperability across platforms. The update introduces key features including row-level deletes, clustering for managed tables, and a new Variant data type for semi-structured data, marking a significant expansion of Snowflake’s open data lakehouse capabilities.
The new features arrive as organizations increasingly demand flexibility in managing data across multiple platforms without vendor lock-in. According to Snowflake’s announcement, the v3 specification also introduces Copy-on-Write (CoW) and Merge-on-Read (MoR) strategies, offering users flexibility in handling data updates and deletes.
Beyond the headline features, the update includes row lineage capabilities that enable Change Data Capture (CDC) use cases by tracking modifications at the row level, providing organizations with a clear history of data changes. This functionality is particularly valuable for compliance and auditing requirements in regulated industries.
Two Distinct Management Models
Snowflake’s implementation divides Iceberg table support into two models with significant functional differences. Snowflake-managed tables offer full ACID transactions, multi-statement transaction support, and clustering capabilities through the CLUSTERING KEY feature for performance optimization. These tables benefit from Snowflake’s automated lifecycle management, including compaction.
In contrast, externally-managed tables integrate with external catalogs like AWS Glue or REST systems but come with limitations. These tables support autocommit only for write operations and require external tools like Spark for optimization and maintenance. Notably, while externally-managed tables can read v3 format data, write support remains limited to the v2 specification, according to Snowflake’s documentation.
The distinction becomes critical for organizations planning their data architecture. Snowflake-managed tables provide full v3 feature support including deletion vectors, while externally-managed tables with frequent row-level deletes require regular external maintenance to prevent performance degradation.
Preview Limitations and Future Outlook
As a public preview feature, the v3 support is intended for evaluation and testing rather than production use. Current limitations include lack of support for row-level equality deletes and certain table properties. Standard Snowflake features such as Fail-safe, hybrid tables, and native schema evolution remain unavailable for Iceberg tables.
Snowflake has not announced a timeline for general availability, with the roadmap likely depending on customer feedback during the preview period. Future development areas include expanding v3 write support to externally-managed tables and providing clear migration paths from v2 to v3 tables.
The announcement positions Snowflake more competitively in the open lakehouse market, where interoperability has become a key differentiator as enterprises seek to avoid vendor lock-in while maintaining high performance across diverse data workloads.
Sources
- Snowflake
