Data lakes are cheap and flexible, but they are hard to manage safely. By default, a folder of files does not give you database-style transactions, strict schemas, or reliable updates.

Data warehouses solve many of those problems, but they are more rigid and usually more expensive. For a long time, teams used both: raw and ML data in the lake, trusted BI data in the warehouse, and many copy jobs between them.

A data lakehouse tries to combine the best parts of both. It stores data as open files on cheap lake storage, then adds table features such as transactions, schema checks, time travel, updates, and deletes. Formats like Delta Lake, Apache Iceberg, and Apache Hudi make this possible by turning folders of files into managed analytical tables.

This chapter explains why lakehouses exist, how table formats work, how they make object storage safer for analytics, and when to choose a lakehouse over a lake or warehouse.