It's not just a collection, it's the collection. It contains almost every scientific book ever printed, for one thing.
Frankly, it's a massive boon to researchers. It's like a top-tier research university library at your fingertips, and usually more convenient than the real thing.
A lot of people would gladly pay. I'm a paying subscriber to Anna's Archive, which vastly improves the experience of that site. (It's borderline unusable without a subscription.)
Thing is, the Elsevier/Springer model makes it incredibly difficult to pay them. With single papers or book chapters in the $30-40 range, an afternoon's research can easily cost $600. (Note that the authors and reviewers don't get royalties on this, and the Editor-in-Chief of any given journal usually only makes a small stipend!)
There are services like DeepDyve, but they're intentionally gimped and difficult to use, because their user interface is 100% built around preventing you from downloading or screenshotting the papers you "rent"!
If the publishers set up a $100/month all-open-access program, and if the experience were at least halfway decent, I'd bet that a lot of people sign up. And that's not cheap!
Funny that the world where almost all human knowledge and art is free and accessible for everyone exists in parallel to one where articles about which McDonalds meal are you
are paywalled, and funny which world civilized nations have chosen in order to protect The Suite Life of Zack & Cody and all the artists whose livelihoods depend on reruns of iCarly.
Frankly, it's a massive boon to researchers. It's like a top-tier research university library at your fingertips, and usually more convenient than the real thing.