Memory chip markets operate on a structural timing mismatch that has recurred for decades. A new fab takes years to permit, build, and bring into yield-qualified production. By the time that capacity comes online, demand conditions have often shifted, either overshooting into glut or undershooting into scarcity. That lag drives the violent DRAM and NAND price swings that have defined the sector's economics and its reputation for punishing investors. The industry accepted the pattern as structural law. On Thursday, Micron CEO Sanjay Mehrotra said AI has "totally changed" that equation.
The cycle's persistence was never primarily a management failure. Capital decisions made at the start of a construction timeline are locked in well before the cycle peaks or troughs. Consumer electronics, the prior primary demand driver, followed predictable saturation curves. PC refresh rates plateaued and smartphone upgrade cycles lengthened. Fabricators sized capacity to a growth trajectory that the market could no longer sustain, and the correction followed on schedule.
What Mehrotra is arguing
Mehrotra's assertion is that AI represents a demand driver different in character from those prior cycles. He described AI as having "dramatically changed" the memory business in Thursday remarks, language that marks a departure from how the industry has historically framed its own cycle risk. The claim is structural, not cyclical: if AI infrastructure spending sustains memory demand past the peak that consumer devices could not, the oversupply-then-scarcity pattern loses one of its core preconditions.
Whether that holds is what fabricators' next round of capacity decisions will reveal. The supply-side constraint, multi-year build timelines that cannot respond quickly to demand signals, has not changed. What Mehrotra is arguing is that the demand side has.