LTAP aims to unify these workloads on a single copy of data stored in the data lake. According to Databricks, the architecture eliminates ETL, replicas, and data movement by design . Transactional data becomes available for analytics instantly, without transformation or pipeline maintenance.

The foundation of LTAP is Lakebase, Databricks’ serverless Postgres service built on open object storage. Lakebase already serves thousands of customers and handles 12 million database launches per day across the platform . Under the LTAP model, Lakebase stores data directly in Unity Catalog using open formats—Delta Lake and Apache Iceberg—so that governed transactional data is immediately queryable for analytical workloads .

The company describes several key properties for the architecture: unified governance with a single source of truth, independent scaling for transactional and analytical workloads, full ACID semantics for Postgres workloads, and no hidden pipelines or connectors to maintain .

New Lakebase capabilities

Alongside the LTAP announcement, Databricks revealed several enhancements to Lakebase itself:

• Cross-cloud, cross-region disaster recovery for production Postgres databases.
• Git-style branching and snapshots, enabling safe experimentation directly on production data.
• Autonomous database operations that monitor for slowdowns, detect anomalies, and trigger automatic recovery .

These features signal Databricks’ intent to make serverless Postgres a first-class operational database for applications and AI agents, not just a convenience layer for analytics.

Lakehouse//RT: real-time analytics on governed data

The second major infrastructure announcement was Lakehouse//RT, a real-time lakehouse powered by a new compute engine called Reyden (short for “Reynold’s Dream Engine,” named after co-founder Reynold Xin) . Databricks says Reyden delivers millisecond query latency at tens of thousands of concurrent users and agents, running directly on governed Delta Lake and Apache Iceberg tables .

The implication is significant: enterprises no longer need to set up separate serving infrastructure—such as caching layers, materialized views, or external query engines—to achieve real-time performance. Sigma Computing joined as a launch partner, connecting directly to Lakehouse//RT for embedded analytics .

Databricks co-founder Reynold Xin described the launch as “probably the single largest introduction we have done since the launch of Lakehouse” .

The agentic push: Genie One, Agent Bricks, and broader AI tooling

Databricks used the summit to position its platform as the foundation for enterprise AI agents. The announcements included:

• Genie One: An AI agentic coworker designed to help business teams automate work grounded in real business data, with enhancements for agentic context .
• Genie Agents and Agent Bricks: Tools for building and deploying AI agents that can observe, reason, and act across enterprise data. Agent Bricks are pre-built components that auto-optimize agents using governed data .
• Unity AI Gateway: Multi-AI governance and cost control for organizations managing multiple AI model providers .

The broader narrative, as captured by industry analysts, is that LTAP and Lakehouse//RT are the data-serving layers underneath an agentic enterprise architecture. By placing operational data in open formats on governed storage, Databricks believes AI agents can access, reason over, and act on production databases without moving or copying data .

Azure and Microsoft integrations

Databricks deepened its Azure ecosystem integration with several jointly announced capabilities:

• Copilot Studio and GitHub Copilot: Users can author Copilot Studio agents that reason over an entire Azure Databricks workspace through a single MCP (Model-Centric-Protocol) connection .
• Genie for Microsoft Teams and M365 Copilot (Beta): Users can tag Genie in a Teams thread and receive context-aware answers directly from their Azure Databricks environment .
• Azure Databricks Excel Add-in (Public Preview) and native Excel ingestion (GA) enable driverless analytics from spreadsheets .
• Databricks CustomerLake: The first lakehouse-native customer data platform, unveiled for Azure Databricks .

These integrations suggest a strategy to embed Databricks’ governance and AI capabilities into the collaboration tools where business decisions happen, rather than requiring users to switch to a separate analytics interface.

What this means for enterprise data architecture

Collectively, the summit announcements represent a coherent platform bet: that the next generation of enterprise applications will be agentic, real-time, and governed. LTAP removes the transactional-analytical divide, Lakehouse//RT removes the latency compromise for analytical queries, and the Genie family provides the agent orchestration layer.

If successful, this architecture could reduce the number of moving parts in a typical enterprise data stack—fewer databases, fewer pipelines, fewer serving layers—while providing AI agents with the governed, real-time context they need to act autonomously on business data.

Databricks is not alone in pursuing this convergence, but with Lakebase already at 12 million daily database launches and a 30,000-attendee summit reinforcing its ecosystem, the LTAP announcement marks a significant milestone in the lakehouse architecture’s evolution from analytics platform to operational data backbone .

LTAP aims to unify these workloads on a single copy of data stored in the data lake. According to Databricks, the architecture eliminates ETL, replicas, and data movement by design . Transactional data becomes available for analytics instantly, without transformation or pipeline maintenance.

The foundation of LTAP is Lakebase, Databricks’ serverless Postgres service built on open object storage. Lakebase already serves thousands of customers and handles 12 million database launches per day across the platform . Under the LTAP model, Lakebase stores data directly in Unity Catalog using open formats—Delta Lake and Apache Iceberg—so that governed transactional data is immediately queryable for analytical workloads .

The company describes several key properties for the architecture: unified governance with a single source of truth, independent scaling for transactional and analytical workloads, full ACID semantics for Postgres workloads, and no hidden pipelines or connectors to maintain .

New Lakebase capabilities

Alongside the LTAP announcement, Databricks revealed several enhancements to Lakebase itself:

• Cross-cloud, cross-region disaster recovery for production Postgres databases.
• Git-style branching and snapshots, enabling safe experimentation directly on production data.
• Autonomous database operations that monitor for slowdowns, detect anomalies, and trigger automatic recovery .

These features signal Databricks’ intent to make serverless Postgres a first-class operational database for applications and AI agents, not just a convenience layer for analytics.

Lakehouse//RT: real-time analytics on governed data

The second major infrastructure announcement was Lakehouse//RT, a real-time lakehouse powered by a new compute engine called Reyden (short for “Reynold’s Dream Engine,” named after co-founder Reynold Xin) . Databricks says Reyden delivers millisecond query latency at tens of thousands of concurrent users and agents, running directly on governed Delta Lake and Apache Iceberg tables .

The implication is significant: enterprises no longer need to set up separate serving infrastructure—such as caching layers, materialized views, or external query engines—to achieve real-time performance. Sigma Computing joined as a launch partner, connecting directly to Lakehouse//RT for embedded analytics .

Databricks co-founder Reynold Xin described the launch as “probably the single largest introduction we have done since the launch of Lakehouse” .

The agentic push: Genie One, Agent Bricks, and broader AI tooling

Databricks used the summit to position its platform as the foundation for enterprise AI agents. The announcements included:

• Genie One: An AI agentic coworker designed to help business teams automate work grounded in real business data, with enhancements for agentic context .
• Genie Agents and Agent Bricks: Tools for building and deploying AI agents that can observe, reason, and act across enterprise data. Agent Bricks are pre-built components that auto-optimize agents using governed data .
• Unity AI Gateway: Multi-AI governance and cost control for organizations managing multiple AI model providers .

The broader narrative, as captured by industry analysts, is that LTAP and Lakehouse//RT are the data-serving layers underneath an agentic enterprise architecture. By placing operational data in open formats on governed storage, Databricks believes AI agents can access, reason over, and act on production databases without moving or copying data .

Azure and Microsoft integrations

Databricks deepened its Azure ecosystem integration with several jointly announced capabilities:

• Copilot Studio and GitHub Copilot: Users can author Copilot Studio agents that reason over an entire Azure Databricks workspace through a single MCP (Model-Centric-Protocol) connection .
• Genie for Microsoft Teams and M365 Copilot (Beta): Users can tag Genie in a Teams thread and receive context-aware answers directly from their Azure Databricks environment .
• Azure Databricks Excel Add-in (Public Preview) and native Excel ingestion (GA) enable driverless analytics from spreadsheets .
• Databricks CustomerLake: The first lakehouse-native customer data platform, unveiled for Azure Databricks .

These integrations suggest a strategy to embed Databricks’ governance and AI capabilities into the collaboration tools where business decisions happen, rather than requiring users to switch to a separate analytics interface.

What this means for enterprise data architecture

Collectively, the summit announcements represent a coherent platform bet: that the next generation of enterprise applications will be agentic, real-time, and governed. LTAP removes the transactional-analytical divide, Lakehouse//RT removes the latency compromise for analytical queries, and the Genie family provides the agent orchestration layer.

If successful, this architecture could reduce the number of moving parts in a typical enterprise data stack—fewer databases, fewer pipelines, fewer serving layers—while providing AI agents with the governed, real-time context they need to act autonomously on business data.

Databricks is not alone in pursuing this convergence, but with Lakebase already at 12 million daily database launches and a 30,000-attendee summit reinforcing its ecosystem, the LTAP announcement marks a significant milestone in the lakehouse architecture’s evolution from analytics platform to operational data backbone .