to aggregate, slice and dice data in more sophisticated ways. (Compare with OLAP)
to have one version of the truth (in order to improve data quality)
Save resources on production database systems: since the data is copied from the source database systems and queried in the data warehouse, this relieves stress on the production database system and gives them better performance.
The value of data can be augmented with data from various sources in the organization.
The data is stored in a common format on a single databaes which makes it easier for data users to access them.
Data can be historized and archived (for example for regulatory reasons). Reports can be dynamically re-created for any dates, thus reports need not be archived.
The physical data model can be Optimized for
read access and
sequential disk scans.
Table and column names can be given more meaningful names than in the source system.
Changing (or upgrading) one source system does (should...) not affect end users.