An archive node stores the same data, in an information-theoretic sense, as a standard "full" node. The archive node just has all the state deltas unpacked out

An archive node stores the same data, in an information-theoretic sense, as a standard "full" node. The archive node just has all the state deltas unpacked out to disk with all intermediate states readily accessible, which is useful for specific use cases like block explorers and analytics services. You can confirm this by syncing an archive node off of a standard "full" node on an isolated network. That 9TB (or 2TB) is only 400GB when it's stored without all of the states unpacked to disk (i.e. non-archive node).