Not quite. Nvidia benefitted from platform dominance and software lock-in. Micron and peers are riding a structural upcycle driven by AI servers needing far more DRAM and HBM per rack. The gains are broad across the memory sector rather than company-specific.

How long can the imbalance last?

Likely 2 to 3 years. HBM capacity is tight, capex remains disciplined after prior busts, and AI demand keeps rising. Supply will expand, but not fast enough to quickly normalise pricing.

Is it too late?

Late for easy upside, not late for returns. Much optimism is priced in, so upside depends on sustained pricing power rather than multiple expansion. Risk is cyclical reversal once capacity catches up.

Bottom line: this is a durable but cyclical memory supercycle. Still investable with a medium-term horizon, but timing, sizing, and volatility tolerance matter more than chasing momentum.