For years, using real-time data products on the lakehouse meant making tough trade-offs. Teams had to manage governance in Databricks and then set up another system, such as ClickHouse, Druid, or Pinot, on top of it. This process involved copying, re-ingesting, and re-governing data just to support applications that needed milliseconds of response time.

However, the true cost was engineering overhead and serious governance gaps caused by the lack of Unity Catalog integration. Teams had to build complex pipelines to circumvent these limitations. During Data + AI Summit 2026, Databricks announced a new real-time query engine, named Lakehouse//RT, which runs directly on the lakehouse and is powered by a new compute engine called Reyden. It could be the end of the fragmented architecture of external serving databases like ClickHouse, Druid, or Pinot that forces teams to mirror data, maintain fragile sync pipelines, and operate outside of unified governance.

What Is Lakehouse//RT?

Lakehouse//RT is Databricks' real-time data warehouse for operational analytics, BI app serving, and observability workloads. We can query directly into Delta Lake or Apache Iceberg tables, without copying data into a separate serving layer, and Reyden delivers the speed.

Reyden is a “clean room” build rather than an evolution of Photon. It's tightly designed for the concurrency and latency requirements that Photon was never intended for. Initial preview metrics demonstrate sub-100ms latency while handling 12,000 queries per second, with some customers experiencing latencies as low as 10ms on smaller datasets. This provides up to 16x faster performance when compared to traditional real-time stacks without moving any data.

All queries run within Unity Catalog. There is no need for extra permissions, proprietary formats,  or sync pipelines. Point it at any existing, governed table and start querying live data in minutes.

Why it Matters: The Serving Layer's Hidden Tax

The main advantages are better performance and the removal of three major costs:

• Data Duplication: Doubles costs for storage, synchronization, and monitoring.
• Governance Fragmentation: Introduces compliance risks, as Unity Catalog's security and audit controls are lost when data is moved outside the lakehouse.
• Engineering Overhead: Increases maintenance complexity due to fragile CDC pipelines, difficult schema management, and the need for dedicated support resources.

Lakehouse//RT eliminates all three of these compounding costs with a single stroke of the architectural pen.

Key Use Cases

• Fraud detection & threat intelligence: A 5x faster response time for live threat lookup queries on the data platform, enabling native lakehouse execution as a production option.
• Observability and performance monitoring: All real-time performance metrics can now be served on the same governed foundation as other analytics workloads, with platforms serving hundreds of queries per second.
• Agentic AI workflows: Autonomous agents are not concerned with “stale data”. They get live, contextual data without a separate serving infrastructure with Lakehouse//RT.

Getting Started: Prerequisites

• All queries must be executed under the governance of Unity Catalog.
• Use of the delta or Iceberg format is required; no restructuring of existing tables is required.

Lakehouse//RT is currently in public beta and available in selected regions. It is best to test it alongside your current production setup before moving live traffic over.

Conclusion

In addition to LTAP (Lakehouse Transactional and Analytical Processing, which unifies transactional and analytical workloads) and Lakeflow (a suite for data ingestion and orchestration that operates natively within Unity Catalog), Lakehouse//RT is the latest part of Databricks' effort to eliminate infrastructure silos. It eliminates the need for fragmented, "side stack" architectures by enabling real-time, low-latency queries directly on the lakehouse. It provides a single, governed foundation, enabling AI agents to read, write, and act without accessing a separate stack at each step. As businesses strive for agility, this leap in performance and simplicity represents a major milestone in building a unified data and AI ecosystem. For organizations looking to modernize, the transition to Lakehouse//RT is a clear way to achieve faster, safer, and more autonomous AI-driven workflows.

How v4c.ai Can Help

For enterprises looking to simplify their architecture and rely less on side stacks, v4c.ai offers a strategic solution to modernize data platforms and accelerate AI initiatives. v4c.ai brings technical know-how to the Databricks ecosystem, helping organizations achieve scalable business outcomes by solving technical challenges.

Our dedicated teams can help you set up the Lakehouse and move your workflow smoothly, so that you can manage and access your data easily. With Lakehouse//RT, we support high-performance ML pipelines and real-time analytics, allowing AI agents to use live, contextual data without worrying about proprietary formats or sync pipelines.

References:

• Databricks Lakehouse//RT
• Region-wise Availability of Lakehouse//RT