Memory pricing has returned, in normalised terms, to where it stood in 2007. The calculation comes from a researcher tracking the long series, showing that the shortage driven by AI demand undid two decades of continuous price decline in a matter of months.
Memory is the best-documented case of steady cheapening in computing. Returning to 2007 doesn't mean the technology regressed; it means demand grew at a rate installed fabrication capacity can't match, and won't match soon, because a new fab takes years.
The effect has already left the abstract. Industry reporting puts roughly a billion dollars of silicon destined for the next iPhone generation sitting idle awaiting packaging because of the memory shortage. When the bottleneck reaches the highest-volume product of the largest phone maker, it has stopped being a niche problem.
On the supply side, manufacturers are moving. Samsung introduced three next-generation memory technologies aimed at AI data centres, and Sandisk and SK hynix published a specification with up to sixteen stacked layers. Real responses, on slow timelines.
For anyone buying equipment, the practical reading is that a budget built on last year's prices no longer closes, and the gap won't revert because a supplier decides to be reasonable. It is why laptop makers began accepting second-source memory, as reported the same week.