I'm not saying you're wrong but it seems incredibly difficult. The problem is, how do you choose who to fund to go and do maths, how on earth you'd track their progress if any, and how do you achieve a return on your investment - even if 'you' means a sovereign nation and you have very general and long-term goals. If you can come up with something distinct from a university mathematics department to achieve this, it would be an achievement. I think we should count ourselves lucky that these departments have come to exist, and they provide such unreasonably useful output.
It might well be possible to use modern communications to distribute the work, improve efficiency and reduce overheads, basically turn all the world's maths departments into one big one - but I bet you still need something like the PhD thesis and viva.
I think what surprises me is the discrepancy between reading about mathematicians who worked with younger students out of seemingly altruistic motives (Kolmogorov and Erdos come to mind) and then a contemporary math department that might only accept 1% of applicants out of a pool of qualified people because of perhaps funding concerns or some interest in maintaining exclusivity. On top of that, there is a long history of mathematicians who worked on problems without any monetary compensation and perhaps without being associated with any university. Decartes and Fermat were both lawyers and even recently Yitang Zhang didn't seem to be paid to do research, and I don't count adjunct lecturer as a paid research job.
Given these, I don't see why an organization can't get together and simply agree to communicate, work together, somehow find a way to minimally provide access to journals for its members and dutifully train new students and grant them degrees. A Phd may or may not be optimal, perhaps something less stringent. A lot of people I know personally didn't get in to the field because they wanted to get rich, some sacrifice may be required but at least there is no denying a sort of "volunteerism" in mathematics.
This is more or less how the Royal Society started.
Academia should now be considered a conglomerate of business corporations, dedicated primarily to increasing profits and income and cutting costs.
Like other corps there's a trend towards increasing perceptions of value by marketing mediocre products and experiences to death, rather than providing actual social value with products and relationships that are truly inspiringly awesome.
So I don't think the current model is sustainable. It won't collapse next week, but I'd be surprised if we're still here twenty years from now.
As for PhDs - in my own area of interest the tenured people are pretty mediocre (believe me - I've worked through their papers) and most of the innovation and invention happens in industry.
So, there's no obvious benefit to spending the money. As it happens I earn enough to do research part time - no special resources needed in this field - and I'm getting a lot of satisfaction from making my own way for my own reasons.
It might well be possible to use modern communications to distribute the work, improve efficiency and reduce overheads, basically turn all the world's maths departments into one big one - but I bet you still need something like the PhD thesis and viva.