
The famous NSA Utah Data Center was publicly estimated from blueprints at roughly 3 to 12 exabytes back in 2013. The real number is classified and it may have grown, but even 12 exabytes would only hold a fraction of one month of global internet traffic. At 522 exabytes per month, 12 exabytes is about 2.3 percent of one month, or less than a day’s worth of global traffic. At Sandvine’s higher 33 exabytes per day estimate, 12 exabytes is only about 9 hours of traffic.
That is why I think the more realistic model is not ‘store every word, call, text, and packet forever.’ It is broad metadata collection, traffic filtering, selectors, targeting rules, keyword or intelligence triggers, and then storing selected content where it meets a requirement.
Metadata is compact and extremely revealing. Universal raw-content storage is a much bigger claim, and I would need evidence for that.
English

















