Early in your career, yes. As you dig deeper they become available only in university libraries and then not even there.
Most material is available for free on the internet, but like most open source material, you have to compile it yourself and that takes time. Good books save you the time by including important dependencies, normalizing notation, etc.
Exactly this - most of the libraries in my area have little if no technical books. I know exactly why too - I've frequently come across libraries in the UK that are selling off their technical books, because they're not receiving much attention. I've had multiple occasions where I've managed to buy dozens of books for the value of just one of them.
It's not easy to even ask for a book. I've asked librarians for books before, and I even have a friend currently working in a fairly large library request a book for me (with evidence from me of interest in it), and the request was refused.
The real books worth reading are those that make you a greater expert than the majority of the population. Libraries don't benefit much from storing books that will only be read by one or two people.
I gave away technical books at my local library as they had a "free books section". Some of them were older editions of industry standard text books (Real-time rendering and PBRT first edition), but the library refused to take donations because they were "technical books". Granted a lot of them were really outdated 90s technical game programming books like "Windows 95 Game SDK Strategy Guide" and "The Black Art of Windows Game Programming".
What are examples of books you can't even find at university libraries?
I've been in software for 20 years and have not yet felt like I needed to read anything that wasn't pretty mainstream.
I imagine one's professional reading is more academic when working on things like hardware or compilers, but most of us (for better or worse) are just working on CRUD apps in a marketing-adjacent space or maintaining a corporate Java monolith.
You won’t find more than the introductory stuff for advanced subjects in math, physics and computation unless the university has someone working on that stuff. It’s been over 20 years since I was last enrolled as a student, but when I was, I had to ask for inter-library loaned books every other month because my deep interests differed from those of the faculty elders.
I don’t know what the CRUD equivalent would be, if there is one, but at the time random projections and large deviations were fringe math, and only the basic texts were available everywhere.
Most books on deep learning are too new or have too popular a style to be included in uni libraries. By the time they do appear, the examples no longer work because the tools have evolved.
The same goes for recent books on applied machine learning or recent tools or libraries intended for professional practitioners. I think libraries are reluctant to buy books they think will go out of date quickly.
Most material is available for free on the internet, but like most open source material, you have to compile it yourself and that takes time. Good books save you the time by including important dependencies, normalizing notation, etc.