That prediction has proven false since first made in the Eighties, a quarter century ago, when PCs started becoming fairly common in households.
Not as common as we RPGers would like! ...er, wait, you mean Personal Computers, don't you. darn.
I also agree that the subscrption model will continue to be used for as long as it continues to serve so well by companies of a certain size and larger who have the ability to maintain the sort of release schedule that can satisfy the model. That, however, isn't really a prediction so much as an observation, since it has pretty much become standard practice for both industries.
As to the one actually prediction you made of 2 years for 5E? Maybe. They like to release in June to get the full summer bump ending with the big conventions, so maybe a bit less than that. I think we'll see admission of them working on 5E in earnest by next Gencon at the latest and a June 2013 release at the latest as well. I think they will also move away from the big core release followed by the supplement, supplement, etc, etc, etc, model that doesn't seem to work for them very well anymore.
I think the new 5e discussions Mearls has posted hint that they'll have a very basic system as the core, then future releases will be supplemental rules like skills, feats, psionics, ship combat, mounted combat, aerial combat, monster races, underwater fighting, horror mechanics, new items, new monsters and all manner of things that can be added onto the core or ignored.
Add to that the usual adventures, adventure paths and settings, each with new areas published via the subscriptions, and you've got 5e.
The point being that they'll offer much less with the core mechanics next time around, maybe.
Money-wise, there are other companies out there, including Goodman Games and White Wolf that seem to make a buck or two. Enough that books are sold regularly in all sorts of comic stores and book chains, at least, if not online. Any stats from their sales? I find this interesting.
Frankly, I'm impressed Paizo makes as much as you say. I'm glad, but those numbers are huge, imho. Granted, this isn't personal income it's the budget of a company, but still. Divide the sum you mentioned by the number of staff names on a module and that should be most of the company's salaries (approximately) or so (not including monies sunk into production, obviously, and I can't think of the reasonable amount for that, as I don't work in finance/economics).