Built for open, intelligent data storage

Databricks lakehouse storage lets you choose your storage location and open table format — Delta Lake or Apache Iceberg™ — while keeping a single copy of source data accessible by any engine. It combines this with unified governance through a single catalog for data discovery and AI asset management, plus AI-driven models that autonomously optimize data for speed and low cost. The result is full ownership and portability of your data without vendor lock-in.

It's built for data and AI teams that need one storage layer to support ETL pipelines, analytics and BI, and data sharing. Managed tables act as both batch tables and a streaming source and sink, so streaming ingest, batch backfill and interactive queries integrate directly with Spark Structured Streaming. Teams that also need to share live data with other platforms, clouds or regions without copying it can do so directly from Delta Lake or Apache Iceberg tables.

Delta Lake and Apache Iceberg™ let you operate a multicloud lakehouse architecture that provides data warehousing performance at data lake economics, delivering up to 6x better price/performance for SQL workloads than traditional cloud data warehouses. This lets you run analytics and BI workloads directly on your data lake instead of moving data into a separate warehouse. See Databricks SQL for more on running these workloads.

Managed tables give you best-in-class read and write performance for Delta Lake and Apache Iceberg™ tables out of the box, with storage optimizations not available in any other lakehouse. Foreign tables instead let you access tables managed by external catalogs, such as Glue, HMS or Snowflake Horizon, while still applying Unity Catalog's fine-grained access controls. Both approaches keep your data portable and let you avoid vendor lock-in.

You can try lakehouse storage for free by signing up directly from the product page. If you'd rather see it in action first, you can watch a demo covering Delta Lake 3.0, UniForm and Liquid Clustering. Additional tutorials, including a CDC pipeline with Delta, are available in the Delta Lake demo library.

Predictive Optimization uses AI-driven models to tune tables based on your data and usage patterns automatically, so tables stay optimized without manual maintenance. Liquid Clustering is an out-of-the-box, self-tuning data layout that scales with your data and requires no partitions. Customers have reported concrete results from these features: Anker Innovations saved 50% in annual storage costs while speeding up queries more than 2x, and Shell improved time series query performance up to 10x using Liquid Clustering.