Databricks Lakebase vs Snowflake Postgres

Summary

• Lakebase is "a fully managed Postgres database integrated into the Databricks platform", built on Neon technology, with autoscaling compute, scale to zero and branching. Snowflake Postgres lets you "create, manage, and use Postgres instances directly from Snowflake", each on a dedicated virtual machine that Snowflake manages.
• Both are generally available, support Postgres 16 through 18, accept standard Postgres clients, and list PostGIS and pgvector among their extensions.
• Lakebase connects to the lakehouse through a read-only Unity Catalog catalog (GA), synced tables (GA), data that is readable through open table formats such as Delta and Iceberg, and a change feed into Delta tables (Public Preview). Snowflake Postgres connects to Snowflake through pg_lake and Iceberg tables, and through data mirroring (Preview).
• One documented difference: Lakebase lists native Postgres logical replication as not yet available, while Snowflake Postgres documents logical replication to or from an external source.

Databricks Lakebase vs Snowflake Postgres

Both are generally available, managed Postgres services that you reach with standard Postgres clients, so the difference is the platform around the database. Lakebase runs inside the Databricks platform with autoscaling compute, scale to zero and copy-on-write branching, and it connects to Unity Catalog and the lakehouse's open table formats. Snowflake Postgres runs each instance on a dedicated virtual machine that Snowflake manages, and it connects to Snowflake through pg_lake and Iceberg tables. Choose the one that sits next to the data platform you already use.

What are Lakebase and Snowflake Postgres?

Lakebase is built on Neon technology, with separated compute and storage so each scales independently. Databricks announced its agreement to acquire Neon in May 2025. You connect to Lakebase with psql or any Postgres driver.
Snowflake Postgres runs a Postgres database server for each instance and is available in selected AWS and Azure regions. You connect to an instance with any Postgres client.

Databricks Lakebase vs Snowflake Postgres: how they compare

Capability	Lakebase	Snowflake Postgres
Postgres versions and extensions	Postgres 16, 17 (default) and 18. PostGIS and pgvector listed on Postgres 17	Postgres 16 through 18. PostGIS and pgvector listed
Compute	Autoscaling compute and scale to zero	You pick a compute family and storage size when you create an instance
Branches and recovery	Copy-on-write branches, including a branch from a point in time within the history window	A fork is a new instance that reflects an existing one at a point in time within the retention window
High availability and replicas	Secondary computes across availability zones, automatic failover, same connection string. Separate read replicas	A standby in another availability zone, automatic failover, no connection change. Optional, and not on burstable sizes. Separate read replicas
Logical replication	Native Postgres logical replication is not yet available	Logical replication to or from an external source
Access control	Postgres roles and GRANT, including OAuth roles for Databricks identities. Unity Catalog privileges apply to the registered read-only catalog	Postgres credentials or short-lived Snowflake tokens. A network policy is required before an instance accepts connections
Link to analytics	A read-only Unity Catalog catalog (GA), synced tables (GA), storage readable through Delta and Iceberg, and a change feed into Delta tables (Public Preview)	pg_lake with Iceberg tables, and data mirroring (Preview)

How Databricks approaches transactional Postgres workloads

Lakebase puts the operational database next to the lakehouse. With synced tables, Unity Catalog tables sync into Postgres "so applications can query lakehouse data directly with low latency". Registering a Lakebase database in Unity Catalog creates a read-only catalog that you query through a serverless SQL warehouse. Lakebase materializes row-oriented Postgres data into Parquet's columnar layout as the data lands in the lake, where it is readable through open table formats such as Delta and Iceberg. Databricks calls this architecture LTAP, short for Lake Transactional/Analytical Processing. Unity Catalog governs analytical access, while Postgres privileges govern connections from applications. Lakebase Change Data Feed, in Public Preview, stores every insert, update and delete as a row in a Unity Catalog managed Delta table.
Lakebase is a managed service, so some Postgres features differ. You can't connect as the Postgres superuser or reach the host operating system, native logical replication isn't yet available, and tablespaces aren't supported. Read the Lakebase documentation for the full list.

What does Snowflake document for connecting Postgres to Snowflake?

Snowflake describes two ways to move data between a Postgres instance and Snowflake. pg_lake is "selective, on-demand data movement between Postgres and Iceberg tables, stages, or customer-managed cloud storage", and tables that Snowflake reads through the catalog integration are read-only in Snowflake. Data mirroring, a Preview feature, replicates whole tables continuously. See the Snowflake Postgres documentation.

FAQs

Are Lakebase and Snowflake Postgres both generally available?

Yes. The Databricks release notes record Lakebase general availability on January 22, 2026. Snowflake's release notes record Snowflake Postgres general availability on February 24, 2026.

Does Lakebase store its data as Delta tables?

Applications work with Postgres tables over standard Postgres connections. Databricks documents that Lakebase materializes the data into Parquet's columnar layout, readable through open table formats such as Delta and Iceberg. You can also register the database in Unity Catalog for read-only queries, or turn on the change feed (Public Preview) to store changes in Unity Catalog managed Delta tables.

Can I restore to an earlier point in time?

Yes on both. Lakebase creates a branch from a point in time within the history window. Snowflake Postgres creates a fork, a new instance, from a point within its retention window.

Which extensions do they support?

Lakebase lists PostGIS and pgvector for Postgres 17, and Snowflake Postgres lists PostGIS and pgvector in its extensions catalog. Both let you run CREATE EXTENSION for supported extensions.