Databricks introduced LTAP (Lake Transactional/Analytical Processing), a new architecture that merges OLAP and OLTP workloads on a single data copy in the lake. The platform eliminates traditional ETL pipelines, replicas, and data movement by design, addressing a fundamental bottleneck in enterprise data infrastructure that has persisted for decades.

LTAP combines Lakebase (serverless Postgres on object storage) with the Lakehouse under unified governance. Unlike previous HTAP solutions that compromised performance through shared engines, LTAP maintains strict workload isolation while enabling independent scaling. All data lives in open formats like Delta and Iceberg, accessible through standard Postgres interfaces without proprietary lock-in.

The announcement responds to AI agents multiplying application development by roughly 50x, creating demand for real-time data access across operational and analytical systems. Traditional CDC pipelines proved brittle under human-scale development; they collapse entirely under agent-driven workloads requiring near-instantaneous read-reason-act cycles.

Lakebase already serves thousands of customers including Block and Zillow, handling 12 million database launches daily. New capabilities include cross-cloud disaster recovery, git-style branching for safe experimentation, and autonomous database operations that let agents monitor performance and propose optimizations. LTAP represents a fundamental shift from pipeline-based integration to storage-layer unification.