Millions of users, thousands of paying users

Warning: this whole article is an expresion of desire, until further notice.
I am working on a project that will hopefully become a company within the next couple of years.
I want this project to reach millions of people. Partly because of ambition, of course – but also because if this project truly solves the problem I want to solve, then it is highly likely that many people will find it useful.
The economic goal of this project is to create something that a) will be able to pay for its own infrastructure plus a few salaries (including mine); and b) add as much possible value to everybody who uses it.
My intuition is that, if I’m not aiming to maximize profits, an overwhelming majority of my users/customers won’t ever pay for the services they use. In other words, a few users will bear the cost of financing the maintenance and improvement of the service.
Why do I predict such an imbalance, and why do I consider it as a positive thing? Because a tail distribution will very probably describe the amount of money made (or saved) by the users – which means that very few users will make/save a ton of money using the app. Hence, these users can afford to fork over a fraction of this gain/saving to my application, hence paying for everybody’s operating costs. Because this fraction will be comparatively small, it would take a seasoned spin doctor to argument that the paying users are being exploited.
Notice that these users at the head of the money curve are not necessarily the ones deriving the most value from the application. This statement requires the narrow and even immoral definition that whoever makes the most money with something gets the best use of something. To understand why this is so, consider the case of users that have comparatively little money, but who still find the tool to be profoundly useful – these are not a bit less important than the paying users. The only thing that distinguishes them is their access to wealth. But how can we detect these paying users?
The heaviest users, in terms of operating costs, will be highly correlated with the ones having a large monetary gain/save, so it is easy to detect them and persuade them to pay. An important exception would be non-profits, which could be subsidized to a certain extent. So, to implement this model, you simply charge the users which consume most of your CPU, bandwidth and storage space. In other words, let the head pay.
I think this arrangement is possible because of the abundance brought by modern software, hardware and telecommunications. A well-made, well-run web application can generate orders of magnitude more value than what it costs. This means that, using economics jargon, the cost of the service is an easily satisfiable border condition, and most of the value is distributed to the consumers – hopefully generating positive externalities to non-customers as well.
I sincerely hope to be able to launch this project, make it solve many people’s problems, and treat every user alike, whether they pay for the service or not. It would be a beautiful embodiment of that socialist principle of from each according to their ability, to each according to their needsAnd it feels way more exciting than displaying ads or having a freemium model.