Is Databricks abandoning Delta for Iceberg?

Summary

• Databricks' documentation, updated September 11, 2026, states that "Delta Lake is the default format for all operations on Databricks" and that tables on Databricks are Delta Lake tables unless otherwise specified.
• Databricks also supports Apache Iceberg in Unity Catalog. Managed Iceberg tables, foreign Iceberg tables and Iceberg v3 features became generally available on May 21, 2026.
• Databricks' stated aim is interoperability between the formats. When it announced its agreement to acquire Tabular in June 2024, it said it would work with the Iceberg and Delta Lake communities toward "a single, open, and common standard of interoperability."
• Iceberg reads (UniForm) let Iceberg clients read Delta Lake tables without rewriting the data files, and the Unity Catalog Iceberg REST catalog serves both Delta tables with Iceberg reads and managed Iceberg tables to external engines.
• The open source Delta Lake project kept releasing in 2026, from Delta Lake 4.0.1 in January to Delta Lake 4.4.1 on October 5.

Is Databricks abandoning Delta for Iceberg?

Databricks' public documentation describes Delta Lake as its default table format and Apache Iceberg as a supported format, with both governed in Unity Catalog. The Delta Lake docs, updated September 11, 2026, say "Delta Lake is the default format for all operations on Databricks," and "Unless otherwise specified, all tables on Databricks are Delta Lake tables." Since May 21, 2026, Unity Catalog managed and foreign Apache Iceberg tables have been generally available as well. In its blog, Databricks frames the direction as interoperability, so that Delta and Iceberg clients can work on the same tables.

Delta Lake vs Apache Iceberg on Databricks: how they compare

Every cell comes from Databricks documentation or the open source projects' own docs.

Capability	Delta Lake on Databricks	Apache Iceberg on Databricks
Role in the docs	"The default format for all operations on Databricks"	Managed and foreign Iceberg tables generally available since May 21, 2026
Create a table	CREATE TABLE without a format clause creates a Delta Lake table	CREATE TABLE ... USING iceberg creates a managed Iceberg table (Databricks Runtime 16.4 LTS or above)
Iceberg client access	Read-only, through Iceberg reads (UniForm) and the Iceberg REST catalog	Read and write through the Iceberg REST Catalog API
Existing tables elsewhere	Clone or convert Parquet and Iceberg tables to Delta Lake	Read Iceberg tables managed by a catalog outside Unity Catalog as foreign Iceberg tables, read-only
Iceberg v3 features	Deletion vectors, VARIANT and row lineage on managed Delta Lake tables with Iceberg reads	The same features on managed Iceberg tables
Automatic maintenance	Predictive optimization runs OPTIMIZE, VACUUM and ANALYZE on Unity Catalog managed tables	Predictive optimization covers managed Iceberg tables too
Open source project	Delta Lake, with releases 4.0.1 through 4.4.1 in 2026	Apache Iceberg; Databricks supports versions 1, 2 and 3 of the spec

What Databricks has said publicly

• June 4, 2024, announcing the Tabular acquisition. Databricks wrote that it intends "to work closely with the Iceberg and Delta Lake communities to bring format compatibility to the lakehouse; in the short term inside Delta Lake UniForm and in the long term by evolving toward a single, open, and common standard of interoperability." It added: "This is a long journey, one that will likely take several years to achieve in those communities."
• June 12, 2025, announcing Iceberg support in Unity Catalog. "In future Delta Lake and Apache Iceberg releases, we want to build on this foundation so that Delta and Iceberg clients can use the same metadata and thus, can share tables directly."
• May 28, 2026, on Iceberg v3 general availability. In a section on Iceberg v4, which the Iceberg spec lists as under active development, Databricks described its goal as raising the bar "in a way that brings Iceberg and Delta Lake closer together," and wrote that it is "proposing that the next version of Delta, Delta 5.0, adopts the adaptive metadata tree structure."

"Delta Lake UniForm" in the 2024 quote refers to the metadata layer behind the feature the docs now call Iceberg reads.

How Databricks approaches both formats

Unity Catalog is the shared layer. Its managed tables are "the default and recommended table type in Databricks for Delta Lake and Apache Iceberg," and the Iceberg REST catalog serves both to external clients such as Apache Spark and Apache Flink. The docs give a path for each case:

• Iceberg engines only read the table. Keep it in Delta Lake and enable Iceberg reads (UniForm), which generates Iceberg metadata "without rewriting the Parquet data files."
• External Iceberg engines write the table. Create a managed Iceberg table in Unity Catalog.
• Another catalog owns the table. Read it as a foreign Iceberg table.

CREATE TABLE main.sales.orders (order_id BIGINT, amount DECIMAL(12, 2));

CREATE TABLE main.sales.orders_iceberg (order_id BIGINT, amount DECIMAL(12, 2))
USING iceberg;

The first statement creates a Delta Lake table, the default. The second creates a managed Iceberg table, as shown in the Unity Catalog managed tables docs.

FAQs

Is Delta Lake still the default table format on Databricks?

Yes. The Databricks docs state that Delta Lake is the default format for all operations, and CREATE TABLE creates a Delta Lake table unless you specify USING iceberg.

Can I create Iceberg tables on Databricks?

Yes. Managed Iceberg tables in Unity Catalog have been generally available since May 21, 2026 and need Databricks Runtime 16.4 LTS or above.

Can Iceberg engines read my existing Delta tables?

Yes, read-only. Iceberg reads (UniForm) generate Iceberg metadata alongside the Delta Lake metadata and require Databricks Runtime 14.3 LTS and above.

Why did Databricks acquire Tabular?

Databricks' June 2024 announcement says the acquisition brought together the original creators of Apache Iceberg and of Linux Foundation Delta Lake, with the aim of data compatibility "so that you are no longer limited by which lakehouse format your data is in."