Databricks introduced online autoscaling for Lakebase Postgres and detailed its storage architecture.

In August 2026, Databricks introduced autoscaling for Lakebase Postgres, allowing databases to adjust compute capacity without restarts or downtime [1]. The Terraform provider followed with export support for Lakebase autoscaling resources in release v1.130.0 [2]. Databricks also showcased Lakebase at VLDB 2026 [5], detailing an architecture that decouples compute from storage by pairing cloud object storage with a write-ahead log [8]. This design addressed input-output bottlenecks in agentic workloads [8] and supported sub-second feature serving in Databricks Feature Store [12].

Databricks brought the Electric team on board to embed WASM Postgres and PGlite into AI agent sandboxes, establishing patterns where local client databases synchronize back to centralized Lakebase storage [15]. Databricks and partners positioned the managed Postgres service for real-time operational workloads, such as automated fraud workflows [7] and transactional storage behind Genie AI applications [17].

Within the community, developers focused on lifecycle tooling, sharing workflows for database branching [9] and automated schema deployments using Databricks Asset Bundles [10]. Users also addressed integration constraints, such as Lakebase synced tables failing to recognize change data feed metadata generated by materialized views [13], while others evaluated how Lakebase fits into SaaS architectures for traditional transactional workloads [11].