Here is a sentence you have read before: 90% of the world's data was created in the last two years.
It is in slide decks, keynotes and newspaper columns. It was also in an IBM press release in October 2011, alongside the claim that the world was creating 2.5 quintillion bytes of data a day. In May 2013 the Norwegian research institute SINTEF repeated it, and the line has kept circulating ever since.
The claim is really about speed.
"90% in the last two years" is not a statement about how much data exists. It is a statement about how fast it grows. For 90% of everything to be less than two years old, the total has to grow about tenfold every two years — roughly tripling every year, year after year.
That can be checked, and today's figures don't support it.
IDC estimates the world generated around 181 zettabytes of data in 2025, and projects around 394 zettabytes in 2028. That is growth of roughly 30% a year. Enormous — but a long way short of tripling annually.
As a rough illustration: if data grows steadily at about 30% a year, the last two years account for something like 40% of everything ever produced, not 90%. The true share is hard to pin down — IDC counts data that is captured, copied and consumed, not just newly created, and most of it is never stored — but no reasonable version of the arithmetic gets you to 90%.
Why it survives.
The sentence has no date in it. "The last two years" moves forward every time it is quoted, so it never looks out of date — even when the growth rate it depends on is nowhere near what it requires.
What we can actually measure.
The volumes themselves are real and tracked. 181 zettabytes a year works out to roughly half a zettabyte a day — about 500 million terabytes.
Why now? The drivers are genuine:
1. Everyone is a producer. Billions of people with smartphones means billions of cameras, microphones, GPS trackers and keyboards generating content continuously.
2. Machines make more data than humans. Over 40% of internet traffic is now machines talking to machines — IoT sensors, server logs, automated systems, AI training.
3. AI itself is a data factory. Every query, every generated image, every voice assistant interaction creates new data that gets logged, stored, and fed back into training.
None of that needs the 90% line to be impressive. The pile is genuinely enormous, and still growing fast. It just isn't growing fast enough to make a fifteen-year-old sentence true.
Sources checked 23 September 2026·4 sources ↓