And, you'd want/need redundancy. One on-site back up for quick restoration and one off-site for surviving physical disaster. So, you'd need at least 3 times that. In HDD prices, that is roughly 2.5 million per set-up, or 7.5 million total for all three. And in SSD prices, well it's about 3x that. 7.5 million per set up and 22.5million for all three.
An alternate option is a distributed back-up. They could have people volunteer to store and host like 10 gigs each, and just hand out each 10 gig chunk to 10 different people. That would take alot of work to set up, but it would be alot safer. And there are already programs/systems like that to model after. 10 gigs is just an example, might be more successful or even more possible in chunks of 1-2 terabytes. Basically one full hard drive per volunteer.
Lol, had to add that after doing the math for 10 gigs to ten people and realising that was 1000 people per terabyte, so would take 150 million volunteers. Even at 2 petabytes each, assuming we still wanted 10x redundancy in that model, it would be like 750 thousand volunteers or something like that. Maybe there is no sustainable volunteer driven model, lol.