Snowflake Postgres vs Databricks Lakebase: what's the difference?

Summary

• Both are managed, standard-Postgres databases from a data-platform vendor: Snowflake Postgres runs Postgres instances on VMs managed by Snowflake; Databricks Lakebase is a fully managed Postgres database built on Neon, the serverless Postgres engine Databricks acquired in 2025.
• The core difference is architectural: Lakebase separates storage from compute and keeps transactional data in the lakehouse, so Postgres and the analytics engines work from one governed copy under Unity Catalog.
• Snowflake Postgres runs on dedicated Snowflake-managed VMs and connects to the Snowflake platform through data mirroring, built on logical replication and the open-source pg_lake extensions.
• Lakebase connects to the lakehouse through Synced Tables (GA), a Change Data Feed that lands Postgres changes as Delta (Public Preview), and Lakehouse//RT for live Postgres queries (Beta); Unity Catalog registration is GA.
• Both run standard Postgres, so existing drivers and tools work with either; the choice usually comes down to which analytics and governance platform your transactional data should sit inside.

Snowflake Postgres vs Databricks Lakebase: what's the difference?

Both put a managed, standard-Postgres database next to an analytics platform, but they connect it differently. Databricks Lakebase is a fully managed Postgres database, built on Neon, that keeps transactional data in the lakehouse and governs it with Unity Catalog, so the operational database and the analytics engines work from one governed copy of the data. Snowflake Postgres runs Postgres instances on dedicated virtual machines managed by Snowflake and connects them to the Snowflake platform through data mirroring. The practical decision is about placement: if your stack already runs on Databricks, Lakebase keeps transactional data inside that governed lakehouse; if it runs on Snowflake, Snowflake Postgres keeps it inside Snowflake.

What is Snowflake Postgres?

Per Snowflake's documentation, "Snowflake Postgres lets you create, manage, and use Postgres instances directly from Snowflake," and each instance "runs a Postgres database server on a dedicated virtual machine managed by Snowflake." It supports recent Postgres major versions (16, 17, and 18) and is "fully compatible with existing Postgres tooling and workloads." It supports common Postgres extensions, including pgvector and PostGIS, per the extensions documentation, and runs on AWS and Azure. Snowflake Postgres followed Snowflake's June 2025 acquisition of Crunchy Data. To bring Postgres data into the Snowflake platform, Snowflake uses data mirroring built on logical replication and the open-source pg_lake extensions, which produce Iceberg change logs the Snowflake engine can read, per Snowflake's data-mirroring docs.

What is Databricks Lakebase?

Lakebase is a fully managed Postgres database integrated into the Databricks platform, documented at docs.databricks.com. It's built on Neon, the serverless Postgres engine Databricks acquired in 2025. Lakebase separates storage from compute: the Postgres compute layer is serverless, autoscales with demand, and scales to zero when idle, with instant branching, cloning, and point-in-time restore. Because it runs standard Postgres, existing drivers and clients connect to it directly. The distinguishing point is where the data lives: transactional data sits in the lakehouse in open formats under one governance model, available to analytics and AI without a separate database.

Snowflake Postgres vs Databricks Lakebase: how they compare

Capability	Snowflake Postgres	Databricks Lakebase
What it is	Managed Postgres instances on dedicated VMs, managed by Snowflake (docs.snowflake.com)	Fully managed Postgres database integrated into the Databricks platform (docs.databricks.com)
Underlying engine	Postgres on Snowflake-managed virtual machines	Neon, the serverless Postgres engine Databricks acquired in 2025
Postgres compatibility	Standard Postgres, versions 16, 17, and 18; compatible with existing Postgres tooling	Standard Postgres; works with existing Postgres drivers and clients
Storage and compute	Instance-based, on dedicated virtual machines	Storage separated from compute; serverless compute autoscales and scales to zero; branching and point-in-time restore
Analytics integration	Data mirroring via logical replication and pg_lake produces Iceberg change logs the Snowflake engine reads (docs.snowflake.com)	Transactional data in the lakehouse; Synced Tables (GA), Change Data Feed to Delta (Public Preview), Lakehouse//RT on live Postgres (Beta) (docs.databricks.com)
Governance	Snowflake platform access control plus Postgres-level controls	Unity Catalog governs the Postgres tier and the lakehouse from one policy plane
AI and ML	pgvector extension for vector data	Model Serving, Agent Bricks, and Databricks AI Search on the same governed tables; low-latency store for AI applications
Cloud availability	AWS and Azure (docs.snowflake.com)	Databricks platform, multicloud by design (databricks.com)

How Databricks approaches Postgres for the lakehouse

Databricks positions Lakebase as operational Postgres that lives inside the lakehouse rather than beside it. Neon, now part of Databricks, provides the serverless Postgres engine: compute scales with demand and suspends when idle, and branching lets teams create isolated copies for development or testing. The data sits in open formats in the lake, so the same records are available to Databricks SQL, Spark, and AI without a separate analytics database. Unity Catalog applies one set of access controls, audit history, and lineage across the Postgres tier and the rest of the lakehouse. Synced Tables, the Change Data Feed, and Lakehouse//RT move or expose data between Postgres and the lakehouse where a workload needs it.
Key docs:

• Lakebase and OLTP on Databricks
• LTAP overview
• Synced Tables
• Register Lakebase in Unity Catalog

FAQs

Is Databricks Lakebase built on Postgres?

Yes. Lakebase is a fully managed Postgres database built on Neon, the serverless Postgres engine Databricks acquired in 2025. Because it runs standard Postgres, existing Postgres drivers, clients, and tools connect to it directly.

How is Lakebase different from Snowflake Postgres architecturally?

Both are managed, standard-Postgres databases from a data-platform vendor. Lakebase keeps transactional data in the lakehouse in open formats and governs it with Unity Catalog. Snowflake Postgres runs Postgres on dedicated virtual machines managed by Snowflake and connects to the Snowflake platform through data mirroring built on logical replication and pg_lake.

How does Lakebase move data between Postgres and analytics?

Lakebase keeps transactional data in the lakehouse, so analytics engines read the same governed data. Synced Tables (GA) and a Change Data Feed to Delta (Public Preview) shape or land data for a specific engine, and Lakehouse//RT (Beta) queries live Postgres data.

What governance does Lakebase provide?

A Lakebase database can be registered in Unity Catalog (generally available), applying access control, auditing, and lineage across both the Postgres tier and the lakehouse from one policy plane.