A reason to preserve the physical book that has been digitized is that it is the authentic and original version that can be used as a reference in the future. If there is ever a controversy about  the digital version, the original can be examined. A seed bank such as the Svalbard Global Seed Vault is seen as an authoritative and safe version of crops we are growing. Saving physical copies of digitized books might at least be seen in a similar light as an authoritative and safe copy that may be called upon in the future.

Assume 200-million books, a reasonable estimate of books ever. Also assume an average of 300 pages per book, 2KB per page. Then assume physical storage space of 8”x8”x2” for a terabyte of data. Result:

So every book in the world could be stored in 8 cubic feet of 1TB disks. Why not increase that to, say, one small datacenter, to include cooling, power, etc. (Although really, why power them? Just put them on a shelf!)

Now, set up 10 of these datacenters world-wide. Checksum one disk per-datacenter-per-day and compare with a master list. If the disk fails, clone one of the good copies.

So in the space used by a mid-sized company’s HR department you can keep the entire printed history of the world safe and secure. 

Scale the space by a factor of 100 if you want reasonable-resolution scans (instead of just searchable, indexable text) of the originals. In a few years the storage density will surely increase by that much and you’re back to eight cubic feet per library.

Keeping the physical books is a waste of space. Preserving fetishes, not knowledge.