The hosting is not the hard part. Arxiv.org and SciHub do that on a shoestring budget. The hard part is changing the mindset of the scientists. Reputation is a

The hosting is not the hard part. Arxiv.org and SciHub do that on a shoestring budget. The hard part is changing the mindset of the scientists. Reputation is a very strong motivation factor for scientists to publish in prestigious journals. And journals (prestigious or not) like to charge a lot of money just because they can. Fortunately things are changing. Most people in the machine learning community posts their papers directly on arxiv.org.