Is a data mart denormalized?

The word data-mart, depending on the author, can mean a part of a larger data warehouse covering a subject matter area, or a small derived data warehouse that draws all of its data from a larger master data warehouse through a separate ETL process, or just a small data warehouse that is considered not big enough to justify the name “data warehouse”.

In either case, the architectures vary. There are completely normalized data warehouses / data marts, there are such that have a dimensional (star schema) architecture with normalized fact tables and denormalized dimensions, and combinations: data warehouses that (in addition to the ETL/staging layer) have a normalized “business layer”, and a dimensionally designed “presentation layer”, which is fed from the business layer.