In the past, the storage cycle was every two or three years.

Now, buyers are signing five-year contracts.
The storage cycle logic is: when demand comes, manufacturers expand production, leading to oversupply and price crashes, then cut capacity for the next round; the mechanism itself has not disappeared.
However, there is now a variable that disrupts the rhythm.
Samsung, SK Hynix, and Micron have shifted 93% of their capacity to AI data centers, leaving only a small amount for consumer-side capacity.
Google, Microsoft, Amazon, and Meta's capital expenditures this year total $725 billion, a 77% increase from last year.
Microsoft's CFO specifically said that $25 billion of that was due to increased storage prices.
Phison's CEO was more direct: by 2026, all NAND capacity for the year has already been fully booked.
Demand locked in by multi-year contracts is on a different scale than quarterly procurement.
The cycle still exists, but AI has permanently raised the bottom, with each trough higher than the last.
In the past, storage cycles had an endpoint; currently, the endpoint of this cycle is not visible.