Weirdly, sometimes the author has to pay the publisher to get published. And then the publisher again gets money from people buying the paper! All in the name of getting published in a 'respectable' journal.
And yes, alternatives exist. But companies are selling the bits for more than the cost of production and millions of people are buying it, despite cheaper competition. This seems to be sustainable. So the original argument about economics and what markets will bear is false.
The point of bringing up video, movies, and books is to show that it's not that inevitable and markets don't have to work that way. Netflix subscriptions pay for a lot more than download bandwidth. Similarly with Kindle.
So it depends on the specifics of the market. Currently, academic publishers are not meeting that community's needs very well (understatement), so it's ripe for disruption. Competing with Amazon, iTunes, and Spotify is a lot harder.
My guess is that eventually a better service will come along, colleges will be willing to fund it (at far cheaper than today's prices for journals), and many people will stop using SciHub. But, as with the music industry, that may take a while; it doesn't happen automatically.
Scientific papers are far easier. The demand even for the most popular ones is likely a lot less than the demand for a semi popular song. It's is easy to self host them. It's also easy to categorize and tag them. And that makes them easy to index. I am going to venture a guess that all scientific papers ever written (stuff that would be published in a journal, not necessarily all the accompanying data), could probably fit on your phone's storage. A kick ass engine to search it could be an app you download. Oh and for the most part there are no licenses to negotiate. You just publish.
So while the only way to get a new scientific paper in front of a large audience of your peers was to print 10k copies of it in a journal and mail the journals, publishers provided value. Now they do not.
Oh and Spotify is not a counter example. If you could have access to app scientific papers in the world ever published as soon as they are published, but it cost you $10/month we wouldn't be talking about this. It's still too expensive, but in relative terms cheap enough that nobody would give a crap. Subscribing to multiple publications to get a small subset is silly. Imagine if instead of Spotify you had to subscribe to each label's streaming service. Would you?