Databricks Launches Serverless Postgres 'Lakebase'
Databricks has launched Lakebase, a new serverless Postgres database built on open formats to avoid vendor lock-in. It separates compute and storage, allowing it to scale to zero when idle and branch production data for development in seconds, positioning it as a flexible alternative for data-intensive applications.
Databricks' move into the operational database space was significantly accelerated by its acquisition of Neon, a company specializing in serverless Postgres, for a reported $1 billion in May 2025. This acquisition brought in the foundational technology and expertise for separating compute and storage, a core architectural principle of Lakebase. The development of Lakebase has been underway since at least June 2025, with a public preview announced then and general availability on AWS and beta on Azure as of February 2026. The architecture of Lakebase is designed to address the bottlenecks of traditional operational databases where a single heavy query can impact all live operations because compute and storage are tightly coupled. By separating these, Lakebase can scale compute resources to handle traffic spikes and then scale down to zero when idle, optimizing costs. This model allows for independent scaling of compute and storage resources based on workload demands. A key feature for developers is "instant database branching," which creates zero-copy clones of production data in seconds. This allows for risk-free testing and development against a current snapshot of the production database without impacting live applications, a concept similar to branching in Git. This is complemented by point-in-time recovery, enabling database restoration to a specific millisecond. Under the hood, Lakebase integrates with the broader Databricks Lakehouse Platform, including Unity Catalog for unified governance and access control across both operational and analytical data. This aims to eliminate data silos and the need for fragile ETL pipelines that traditionally move data from operational databases to analytical platforms. The service supports Postgres 17 and includes extensions like pgvector for AI-driven search applications.
