arXiv, the physics preprint server, is in the process of moving to a new form of business model, one based on voluntary contribution based on usage. Following is a suggestion for a slight tweak that I think will make the transition smoother; but don't wait, please sign up right away as the financial aspects likely need no change.
The cost of maintaining arXiv, including further developing the service to fully meet the needs of the physics community, is less than $7 per submission or 1.3 cents per download. The current approach is to ask heavy users (readers) of the service to contribute to the funding of arXiv, and 22 institutions have made this commitment so far.
My suggestion, one which would not be that different financially but might be a model that is easier to sustain into the future, is to follow a similar institution-based approach, but base the charges on approximate submissions rather than usage. Regular contributors could pay a flat fee, perhaps calculated to reflect an average of about $10 per submission (to allow for subsidy for authors from developing countries and/or to build an endowment for arXiv's future). Institutions that choose not to contribute on a regular basis could then pay one-off fees reflecting the higher administration cost associated with the one-off payment system, e.g. perhaps about $50 per submission.
To transition from the current plan, likely all that is needed is a little wordsmithing in the current document, i.e. change: "The top 100 institutions based on the previous year's download activity" to "The top 100 institutions bases on the previous year's submission activity". The list of institutions in each category are likely about the same, as it is the active physics researchers who are more likely to be doing the most reading and the most contributing.
This approach is more likely to succeed in the longer term, as this way the institution is clearly purchasing a needed service; and, if necessary, the cost of paying per submission is easily within the reach of the individual author, if necessary.
Noting a bit of inspiration from Joe Esposito on Scholarly Kitchen.
Note also that $7 per submission appears to be covering preservation costs rather than this year's submissions. Clarification on this point from arXiv would be appreciated.