There's also the fundamental limit to the bandwidth that a given consumer can realistically ingest. The richest inputs we have -- our sight -- has been estimate

There's also the fundamental limit to the bandwidth that a given consumer can realistically ingest. The richest inputs we have -- our sight -- has been estimated [1] at about 8.75 Mbps. The real driver for bandwidth (and thus local storage) these days is input latency. We're only just beginning to be able to react quickly enough to requests for different input (VR), so the historical workaround has been increasing the amount of data instantaneously available through things like HDTV. You're never fully consuming the bandwidth provided at any instant but you can increase your awareness by constantly looking at different parts of the screen. So again we're fundamentally limited by the speed of light and how quickly data can be retrieved. If we can get better at local caching and query prediction (streaming in data right before it's requested), then we could theoretically reduce massively the amount of site-local storage required. So for supply/demand reasons, I think software combined with HID improvements (better prediction through better sensors) has a significant potential to essentially put the storage industry into maintenance mode. [1] https://www.newscientist.com/article/dn9633-calculating-the-speed-of-sight https://www.newscientist.com/article/dn9633-calculating-the-...