Databricks has launched Lakehouse//RT, a real-time analytics layer built on its new Reyden compute engine that delivers sub-second query response times for high-concurrency analytical workloads. The breakthrough capability, announced at the Data + AI Summit 2026, fundamentally changes what enterprise AI applications can accomplish by making it possible to query and act on the most current data available — even when that data is being generated at high velocity across distributed systems.
Traditional data lake architectures have been optimized for batch processing and periodic updates, which worked well for business intelligence use cases where yesterday’s data was sufficient for today’s decisions. But as enterprises deploy AI agents and real-time decision systems, the need for truly current data has become critical. A customer service AI agent that references outdated inventory or pricing data provides worse outcomes than one that can access live operational data, even if the underlying model is more capable.
The Reyden compute engine at the heart of Lakehouse//RT uses a novel approach to query planning and execution that reduces latency while scaling to support many concurrent users. Rather than relying on pre-computed aggregations or materialized views that go stale between updates, Reyden can efficiently compute results over live data streams and historical data together — delivering the freshness of streaming with the analytical depth of a full data lake.
Sigma Computing, named Databricks’ 2026 ISV Business Intelligence Partner of the Year, announced it has integrated directly with Lakehouse//RT as a launch partner. This integration enables Sigma’s spreadsheet-style interface to execute sub-second queries exploring billions of rows of data — making real-time big data analytics accessible to business users who are comfortable with spreadsheets but don’t have data engineering expertise.
For enterprise customers, Lakehouse//RT enables a new class of applications that were previously too latency-sensitive to run on data lake infrastructure. Financial services firms can run real-time fraud detection models that query the full transaction history. Retailers can serve personalized product recommendations based on live inventory and pricing. Logistics companies can optimize routing decisions based on current traffic, weather, and carrier availability — all from a unified data platform without the complexity of maintaining separate real-time and batch processing infrastructure.
Databricks is positioning Lakehouse//RT as part of what it calls an LTAP — Lake Transactional/Analytical Platform — a new category that merges the capabilities of transactional databases, data lakes, and real-time streaming systems into a single governed platform. If successful, this consolidation could significantly simplify the enterprise data architecture landscape, reducing both the cost and complexity of building AI-ready data infrastructure.
