Humanity captures approximately 5 billion to 5.3 billion photos every single day. But this number increases to 14 billion images a day through sharing and reposting across messaging and social media. In data terms, this is 50 petabytes of photos, every, single, day!
Adding video to the equation completely blows the doors off the numbers. Video accounts for roughly 82% of all global internet data traffic. the world now creates, captures, or copies roughly 400,000 petabytes (400 exabytes) of data every single day.
To handle this volume, we need to build, finish, and fully pack 2 to 3 brand-new, massive hyperscale data center buildings every 24 hours.
That is the Sci-Fi apocalyptic view–– the pouring of concrete from helicopters onto the homes in rural neighborhoods. Fortunately is doesn’t actually look like this for one simple reason… we delete things. Because we cannot build 1,000 data centers a year, the industry relies on a massive filter. Out of those 400,000 petabytes generated today, 98% to 98.5% is instantly dropped, compressed, or deleted within seconds or hours of creation (temporary video streaming caches, transient network packets, automated logs, and extreme social media video compression).
There are roughly 1,200 to 1,300 active hyperscale facilities running globally and the tech industry builds and activates roughly 120 to 150 brand-new hyperscale data center buildings per year worldwide. A single modern hyperscale building typically spans between 200,000 and 500,000 square feet- or roughly 8 football fields, and many of the established centers are empty in anticipation of them being used.
To handle our actual retained data, we don’t need thousands of new buildings. We need to build about 130 new buildings a year, with each building averaging about 400,000 square feet. The real trick isn’t the square footage; it’s packing those square feet with denser, high-capacity hard drives and securing the massive electrical grid connections required to power them, and fortunately for Moore’s Law, hard-drive technology is still getting faster, smaller and cheaper.
But then there is the power and infrastructure issue– more crucial than the datacenter side of things, but that will be a future post. Until then–slow down on the picture taking. Do we really need to take or share a picture of a parking lot from a train window?
