Databricks Makes Lakebase Postgres Generally Available on Google Cloud

Compute separated from storage

Databricks says the Lakebase architecture separates compute from storage, which removes the resource conflicts and infrastructure overhead of traditional databases. The database also shares the same storage layer as a company's analytical data, so there is no need to build and maintain pipelines between the two.

The company frames this as a response to a long-standing split between transactional databases and analytics platforms. Moving operational data into the lakehouse has meant building ETL pipelines, duplicating storage and accepting data lag. Databricks argues that this overhead becomes a bigger liability now that AI agents act on data in real time.

Lakebase is described as a key component of the Lake Transactional/Analytical Processing (LTAP) architecture, which Databricks says lets organizations unify their data foundation, decrease operational overhead and lower infrastructure costs.

Features for building and shipping faster

Databricks lists several capabilities in the Google Cloud release:

• Serverless autoscaling and scale to zero: compute adjusts to traffic and shuts off when idle, so customers do not pay for unused capacity.
• Instant branching: isolated, zero-copy branches of production data can be created in seconds to test changes safely.
• Point-in-time recovery: a database can be restored to a specific moment to recover from bugs or accidental deletions.
• Standard Postgres: existing tools, drivers and extensions work, including a built-in search engine that is pgvector-compatible for AI-driven search.
• Unified governance with Unity Catalog: consistent access control, lineage and auditing across operational, analytical and AI data.

KaBuM! is quoted as an early user. Pablo Henrique de Freitas, Data and AI Manager at the company, said KaBuM! selected Lakebase Postgres as the default memory layer for AI agents because it delivers the low-latency reads and writes its conversational AI requires.

Availability, security and performance claims

On reliability, Databricks says multi-zone high-availability endpoints with automatic failover are backed by a 99.99% SLA, while single-zone endpoints carry a 99.9% SLA. Enterprise customers can also use Google Cloud Private Service Connect, currently in Beta, to keep database traffic off the public internet and connect applications in their own VPC directly to Lakebase.

Users and service principals authenticate with Databricks OAuth or native Postgres credentials, with access governed centrally in Unity Catalog.

For search, Databricks says Lakebase Search combines vector and full-text search in Postgres for 2x throughput than pgvector at 4x lower cost, without a second system or sync pipelines. For ingestion, a feature called Direct writes pushes enriched lakehouse data into Lakebase Postgres without slowing down an app. According to the company, testing shows loading 1TB of data takes less than five minutes.

Regions and getting started

At general availability, Lakebase Postgres on Google Cloud is available in five regions: asia-southeast1, europe-west3, us-central1, us-east4 and us-west1.

Customers can create a first Lakebase project from a Databricks workspace on Google Cloud, connect with existing Postgres tools and start building. For teams already working on Google Cloud, the release means a managed Postgres option that sits under the same governance model as their Databricks analytics and AI data. The performance and cost figures are Databricks' own claims, and the announcement does not include pricing details.