What Luca said makes perfect sense. Online data is readily available to all production applications and processes. Backups, on the other hand, are usually impossible to get to (even without GDPR, most companies secure access to their backups very tightly, because it is the easiest way to steal company's data).
One interesting fact here is that erasure of data from production automatically turns "backup" into "archive" (as the very difference between backup and archive is whether the protected data is still present in production). While it sounds like a subtle difference, you have to keep in mind that there are existing legal requirements associated with the data archival when a company is legally required to perform one. For example, to be compliant you must ensure data immutability - that is, at any point of time the archived documents retrieved from the archiving system must be the same as originally saved in the system. In other words, you can't simply delete a person from an archived spreadsheet, because as soon you do this to "meet" GDPR, you will instantly fail at another equally important legal requirement. Catch-22? Although, as I have already noted - purpose-built, certified data archival systems ensure immutability by simply not supporting modification of data (WORM tapes, hardware-based write lock for disk storage, etc.)