Databricks used its Data + AI Summit to unveil LTAP, a new architecture it says is designed to bring transactional and analytical processing onto a single copy of data in the lake. The company also announced new enterprise features for Lakebase, the Postgres-based foundation underpinning the approach.

LTAP stands for Lake Transactional/Analytical Processing. Databricks describes it as a way to combine operational, analytical, streaming, and transactional workloads without the separate pipelines, replicas, and ETL processes that have long connected traditional data systems. The company says the model is intended to give enterprises one governed data foundation that can be read, analyzed, and acted on in near real time.

The pitch reflects a broader shift in data infrastructure as organizations build more AI-powered applications and agents. Databricks argued that older architectures were designed for a world where transactional databases and analytics platforms were kept apart, with data copied between them through change-data-capture pipelines. According to the company, that approach is increasingly difficult to manage as software development accelerates and AI systems require fast access to current data.

Databricks also positioned LTAP as a response to earlier attempts to merge operational and analytical workloads. It said prior hybrid approaches either forced both workloads into one engine, creating performance compromises, or preserved the pipeline while hiding it from users. LTAP, by contrast, keeps the workloads separate while unifying them at the storage layer.

At the center of the architecture is Lakebase, which Databricks said already serves thousands of customers and supports 12 million database launches per day. The company said users include Block, Ensemble, Superhuman, and Zillow. Lakebase was launched last year and is built as a serverless Postgres service on open object storage.

With the latest update, Databricks said Lakebase gains cross-cloud and cross-region disaster recovery, along with branching and snapshot tools modeled after Git. The company said those features are meant to make it easier for teams to test changes against production-like data and recover from issues. Databricks also said it is adding autonomous database operations that can monitor system health, flag slowdowns, suggest indexes, and help with recovery tasks.

Databricks says LTAP is built on open standards and uses Unity Catalog for governance, so operational and analytical data share the same identity, permissions, and audit framework. The company said the architecture uses open table formats such as Delta and Iceberg, allowing different engines to work from the same underlying data.

The company framed the launch as a move to remove long-standing tradeoffs between consistency, scale, and complexity. It said transactional workloads will continue to run in Postgres with ACID guarantees, while analytics can scale independently across the Lakehouse without copying data into separate systems.

Grant Veazey, CTO of Ensemble, said the company’s early work with Databricks helped it build a governed foundation for more than two petabytes of revenue cycle data. He said Lakebase and LTAP could give its AI systems the real-time access needed for live operations.

Databricks said LTAP will be available soon as part of Lakebase. The company did not give a specific release date.