To me it really seem like you are way off in the economical aspects of your proposition. Consider:
> As for why Facebook isn't tossing Raspberry Pis at everyone: that doesn't fit its business model (...)
There is nothing that would stop them to give these boxes in exchange of access to the data produced by the users. They would still have access to the data, and could still sell ads based on it. You fail to understand that they can find ways to increase their revenue per user (making ads more relevant or bringing some kind of value proposition where they can get users to volunteer more data), what they can't do is reduce their operational costs. They are doing it as much as they can - look at all the thing about Open Compute, which is nothing more than trying to find a way to reduce the costs of operating data centers - but building multi-million datacenters around the world is cheaper than giving away hardware with "irrelevant cost".
> An origin server seeds cache, but once you've put content on the caching tier (...)
Who is the caching tier? Your neighbor? Akamai? Amazon? Google? ISPs? Your church? What is the incentive for them to keep this data for you and others, and how will they finance the cost of storage, electricity, etc?
> Cost is a minor concern relative to control over data.
You want to make cost operational costs irrelevant, but it is wishful thinking. The whole reason I was asking OP how much he would be willing to pay for this service is precisely to know how much of "Cost is a minor concern relative to control over data."
> Getting the software up to snuff is the missing part.
This is backwards. Software doesn't write itself magically. There needs to have some kind of economical incentive for someone to go over there and develop this software, and people need to accept the costs of maintaining the system operational. Diaspora got started exactly because there was some demand around 2010, but it did not have enough momentum.
http://www.youtube.com/watch?v=XkfBWAJ7kbI
As for why Facebook isn't tossing Raspberry Pis at everyone: that doesn't fit its business model, which is based on 1) sucking in metric gobtons of personal data, 2) correlating it among accounts, and 3) spamming the users with ads. Quite simply, it's not in Facebook's interests to make alternative distributed models possible.
The $5-$15/mo bit matters not in that it's the point to beat, but that that's all Facebook is making from this. If that margin's cut into (say, by giving the most valued users attractive options other than FB), then Facebook faces a long and slow value slide, accelerating as more users (usually starting with the most valuable) hop off the train. Dittos for Google, Yahoo, and other online properties. Amazon and Apple are in a better position to survive this, and might even benefit (in which case, sorting out how to help support the system would be in their interest and network resources could be devoted to that end).
On storage: MicroSD fits on a small device, and the general point is that the gateway you use to access the Internet is already roughly powerful enough to host your content. I strongly dispute your 80 GB of data per summer per user value, though that's probably going to be valid in the not too distant future.
It's also worth considering that MicroSD isn't the cheapest data option. A 1TB SATA drive runs for $55 on Amazon right now. Spinning rust ain't the fastest, but remember, it's _origin_. This is also not just your social networking and cloud, but local storage (you need somewhere to come up with your 80GB to start witht). Remember that people already have extensive existing hardware, and that it's the data more than hardware that's valuable.
For the typical user, a 1TB disk will likely be sufficient throughout its lifetime (3-6 years or so), at which time you'll replace it with whatever's current at the time. Hard drive cost per GB has been falling at a very consistent rate for decades, falling by an order of magnitude every three years:
http://ns1758.ca/winch/winchest.html
http://ns1758.ca/winch/cost-hard-drives-komorowski.jpg
http://www.deepspar.com/wp-data-loss.html
http://www.deepspar.com/images/Storage_Cost.jpg
I could be wrong about this, but at some point you reach data saturation. There's no way in hell I can create GB of text in a year (a few MB is more like it). Audio and video data add up, but there's only so much of that worth saving. Dittos for logging and financial data.
DVD-quality video is 4.7 - 8.7 GB single-sided. Blu ray is 25 GB. That's 40 1080p full-length movies per TB of storage capacity, and three years from now you'll have room for 400 movies. Blu ray video storage is 7 hours at 32 Mbit/s, 3h 30m at 64 Mbit/s (ultra high definition TV).
There's only so much detail your eye can take in.
On caching: I didn't say the less you store yourself, but the less you serve. An origin server seeds cache, but once you've put content on the caching tier, it distributes across the cache and is served from it. That's how CDNs work, and the origin infrastructure of social networks can be frighteningly primitive (or not).
But you're missing three key factors:
1. The barrier isn't hardware. I cited my facts to emphasize this point, you've fallen into that tarpit. The point is: hardware's cheap and not the barrier. It hasn't been for most of a decade.
2. The hardware cost itself is largely irrelevant. Sufficiently capable systems exist and are likely already owned. An old server, desktop, laptop, or within a year or three, tablet or phone, with an appropriate server build on it, will be sufficient. This is a system that can be cobbled on extant hardware at a bare minimum, more advanced kit if desired.
3. It's the software itself, installation, configuration, management, and most importantly, adoption that are the real barriers. Once this gets going it should snowball and steamroll over much of the existing online world. Cost is a minor concern relative to control over data. We're post-Snowden, and everyone knows that both spooks and black-hats are watching, listening, prodding, and prying. Systems which decentralize data and put more of it under access control and encryption guard against that on both fronts.
That's the win.
Getting the software up to snuff is the missing part.