There are also cost issues even when the book is already public domain. Yes, there have been efforts to scan a lot of old books, but often that has just been making fairly low-resolution scans with little or no clean-up process, and assembling those scans into huge and ungainly PDFs. That is OK for old prose works where a reader may only need to consult a single page. But appropriately digitizing something like a book of elaborately typeset poetry, or an art book (i.e. a large-format book with reproductions of paintings) requires a great deal of human effort, and that is expensive. It is no surprise that the print format still reigns in those genres.
I’m with the grandparent here – the problem isn’t the amount of work required to republish a book in a digital-first format like EPUB. It’s the fact that sharing books electronically in any format is a copyright violation, so the scans that exist already aren’t publicly accessible in DRM-free formats. That being the case, recreating those books in digital-first formats requires even more effort (because you have to find and scan the book first), for almost no reward (because you can’t publish the results).