Although less than ideal, cold backups are a perfectly legitimate way to backup a database. The assumption is that the database has been shut down, hence the "cold" nomenclature versus a "hot" or "warm"(1) backup where the database is online.
The down side of a cold backup is you're copying the entire file structure of the database regardless of how full the files are. So if you have a 1TB database that only has 10MB of data in it, you're going to be copying 1TB worth of data files. The hot backup approach works a different way in that only those pages in the database that actually have data are backed up. There's also some level of compression that hot backups support. Generally speaking, hot backups are faster and smaller than cold backups.
In either case, you're going to end up with a consistent backup.
(1) Warm backups are kind of special case depending on the platform. Some database platforms support "quiescing" of the database while it is online, meaning write activity to the data files is halted. You can still query the database, and at the same time you can effectively do a cold backup of the database files. Why would one follow this approach? Some storage subsystems (like EMC) support snapshot-ing a volume over to another volume, effectively mirroring the first volume onto the second almost instantaneously. You can use this to mirror your production data over to a development environment or to a standby system. You need to put the database in a consistent state for the duration of this snapshot process, which is usually only a few seconds. So a "hot" system plus a "cold" backup equals a "warm" backup.