What is Databricks UniForm and how does it make Delta tables readable as Iceberg?

Summary

• UniForm is Delta Lake Universal Format. It lets a Delta Lake table be read by Apache Iceberg clients from a single copy of the data, with no rewrite and no second copy of the files.
• It generates Iceberg metadata automatically. Both Delta Lake and Iceberg store Parquet data files plus a metadata layer. UniForm keeps the Parquet files as-is and asynchronously generates Iceberg metadata alongside the Delta metadata.
• It is read access for Iceberg clients. Any Iceberg REST-compatible engine can read a UniForm table; the underlying table is still written and managed as Delta Lake.
• Unity Catalog exposes it through the Iceberg REST Catalog. External engines read Unity Catalog tables (generally available) and write to managed Iceberg tables (public preview) while governance stays centralized.
• Available on Databricks Runtime 14.3 LTS and above. Enable it at table creation or with ALTER TABLE on an existing table.

What is Databricks UniForm and how does it make Delta tables readable as Iceberg?

Databricks UniForm (Delta Lake Universal Format) is a feature that lets Delta Lake tables be read by Apache Iceberg clients without rewriting the data. Both Delta Lake and Apache Iceberg are built the same way: a set of Parquet data files plus a metadata layer that describes them. UniForm takes advantage of that shared foundation. When you enable Iceberg reads on a Delta table, Databricks keeps a single copy of the Parquet files and automatically generates Iceberg metadata alongside the existing Delta metadata, so the same data can be read as either format. See How Iceberg reads work.

Why Databricks Unity Catalog and UniForm for open interoperability

• One copy of data, multiple formats. UniForm generates Iceberg metadata asynchronously without rewriting the underlying Parquet data files, so a single set of files serves both Delta Lake and Iceberg clients. This zero-copy approach avoids duplicating data. See How Iceberg reads work.
• Read from any Iceberg REST-compatible engine. Iceberg clients can read UniForm tables through a standard interface. Iceberg client access is read-only; the table continues to be written and managed as Delta Lake.
• Iceberg REST Catalog through Unity Catalog. Unity Catalog supports the full Iceberg REST Catalog API, so external engines can read Unity Catalog tables (generally available) and write to Unity Catalog managed Iceberg tables (public preview). Unity Catalog can be configured as an Iceberg catalog through an external connection.
• Centralized governance stays in place. Even when external engines access the data, governance and access control remain centralized in Unity Catalog. See What's new with Databricks Unity Catalog.
• Iceberg v3 features on managed Delta tables. Unity Catalog managed Apache Iceberg tables, foreign Apache Iceberg tables, and Apache Iceberg v3 features are generally available, including deletion vectors, the VARIANT data type, and row lineage.
• Efficient storage. Delta tables with Iceberg reads enabled use the Zstandard compression codec for their Parquet data files. Iceberg metadata generation runs asynchronously on the compute used to write data.

Getting started

• Read How Iceberg reads work for the full model and requirements.
• Enable UniForm on a new table during table creation, or on an existing table with an ALTER TABLE command. UniForm is available on Databricks Runtime 14.3 LTS and above.
• Confirm UniForm is enabled in Catalog Explorer under the Details tab, which shows the Iceberg metadata location.
• Review Read Databricks tables from Apache Iceberg clients to connect external engines through the Iceberg REST Catalog.
• Learn how teams adopt the format in Delta Lake UniForm: Iceberg compatibility now GA.

FAQs

Does UniForm copy or convert my data?

No. UniForm keeps a single copy of the Parquet data files and generates Iceberg metadata alongside the Delta metadata. There is no data rewrite or duplication.

Can Iceberg clients write to a UniForm table?

Iceberg client access to a UniForm (Delta) table is read-only. The table continues to be written and managed as Delta Lake. Separately, Unity Catalog managed Iceberg tables support external writes through the Iceberg REST Catalog API in public preview.

What version of Databricks Runtime do I need?

Iceberg reads (UniForm) are available on Databricks Runtime 14.3 LTS and above.

How does Unity Catalog fit in?

Unity Catalog exposes tables through the full Iceberg REST Catalog API, so external Iceberg engines can read them while access control and governance remain centralized in Unity Catalog.