[pwdisswordfish5]

I share your childlike optimism. But another thing (that neither of the sibling comments mention) is write speed. Having a ton of capacity to store massive amounts of data at rest is not the same thing as being able to easily make near-instantaneous (or even slow) copies. Getting all that data on there is going to be a problem, and so it would only be economical for data sources in high demand, not unlike the way that optical media get the pits and lands stamped in at the factory based on an expensive master. "All of YouTube" might be (probably is) in high enough demand, but it would also require cooperation from the gatekeeper of that content, who at this time has adverse incentives because it's making (many times) more money selling ad impressions for basically every view that it's able to. Even aside from that, what other large datasets are in the same demand category such that the work could or would be subsidized in that way?