> They can't afford anything "per user" or any significant number of users.
That was my point. In this thought exercise, we ought to separate the amount paid to users by country, because it will differ as much as a few orders of magnitude per country, just as the per user revenue most probably does. That is roughly the scale of GDP per capita per country between 1st world and 3rd world countries, which would be a good proxy for ad markets. I agree with rest of your argument, if per person revenue in the US was $150, would that be enough to make users care? It is unknown.
> If you think there's some huge untapped source of revenue here that might make the idea work, show it to me
I will concede that I am leveling an argument that entirely hinges on a hypothetical. But I don't think it is fantastical. The core of the problem is information asymmetry. These companies have huge amounts of high resolution data on us, but we don't have anything about the breakdown and how they operationalize that data. The argument that "if there was more ways to milk money out of it, we would have seen that in revenue" has several drawbacks in my opinion for several reasons.
a) The data doesn't have to be used in ways that translate to quarterly revenue but still great value to the company; this includes back channel uses (elections, famously brexit, that was won by data, and it is hard to put a revenue tag to its effects), it also includes expenditures such as acquisitions of competitors (famously instagram and whatsapp, I don't doubt for a second facebook made use of their data to assess the viability of these acquisitions), or temporarily unused data for game theoretic reasons ("I don't want to expose the fact that I can utilize racial data, health data, covert speech recognition data on my users as long as I can, but if market conditions/lobbying efforts are fruitful, I will translate that to revenue").
b) Even directly revenue increasing uses of data is subject to R&D and mining day in day out. So new combinations of existing data can be figured out, or new products might potentiate existing data to be more valuable without preceding revenue footprints. Think about self driving cars; how traffic operated has been public data, individual sensor technologies aren't that novel, but self-driving car is joining all this together in the hope of converting that data into revenue. When it does, it will be a phase shift that tremendously increases the value of that data, and we will have not been wiser to it if we only watched the revenue. I'm not suggesting FB is necessarily after such an exponential tech, but it is to keep in mind the potentiality stored with data is not readily observable, so “show me it can translate into revenue” criteria can be failed for a long while despite the fact that it eventually can.