Databricks Announces Lakebase Database
Databricks announced Lakebase, a new serverless OLTP database compatible with PostgreSQL and designed for AI workloads. The new offering integrates with the company's lakehouse architecture to provide ACID transactions directly on scalable data lakes. The move positions Databricks to compete with traditional database vendors by unifying transactional and analytical workloads.
- The technology is partially built on acquisitions of the PostgreSQL company Neon and was enhanced by the purchase of Mooncake to improve integration with lakehouse data. - Lakebase supports up to 8TB per instance and runs on the latest Postgres 17, which includes the pgvector extension for AI-powered similarity search. - The architecture separates compute from storage, enabling features like serverless autoscaling with a "scale to zero" option to avoid costs when idle and instant, zero-copy data branching for development and testing. - Billing for the autoscaling version is usage-based and calculated in Databricks Units (DBUs) based on capacity unit hours consumed, with storage billed separately. - The service is now generally available on AWS and in public preview on Azure, with full Azure support and Google Cloud availability planned for later in the year; SOC2 and HIPAA certifications are anticipated in early 2026. - Since its initial preview launch in June 2025, Databricks reports that Lakebase adoption has grown at more than double the rate of its own data warehousing product. - It integrates with Databricks Unity Catalog, providing a unified governance layer for access control, auditing, and data lineage across both transactional and analytical data. - Databricks CEO Ali Ghodsi positioned the offering as a direct challenge to the $100-billion-plus market for traditional OLTP databases like Oracle Database and Microsoft SQL Server.
