Micron just printed $23.9 billion in revenue against a $20 billion consensus, guided next quarter to $33.5 billion with gross margins approaching 81%, and the stock barely moved after hours. That reaction tells you something about where expectations already sit. But I think most of the coverage is anchored on the wrong variable. The consensus narrative is that HBM pricing power is the story, and that capacity expansion from Samsung or SK Hynix is the thing to watch for a reversal.
I started there too. But the more I dug into the wafer economics, the less that framing held up. HBM production consumes roughly three times the wafer capacity of standard DRAM per gigabyte. Every wafer Samsung and SK Hynix redirect toward HBM4 is a wafer that doesn't produce DDR5 or LPDDR. Samsung's DDR5 module pricing is already up 60% since September. Gartner is projecting a 47% DRAM price increase across the board in 2026. The real structural story isn't that memory makers have pricing power in HBM. It's that HBM's wafer intensity is taxing the entire memory stack simultaneously, and mid-market enterprise buyers without hyperscaler-scale contracts are getting rationed.
The break condition most people cite (Samsung or SK Hynix pulling forward HBM supply) is already happening. SK Hynix started its M15X fab four months early. Samsung is tripling HBM output. Prices are still rising. New fabs take 18-24 months to reach volume, and HBM's addressable market is growing from $35 billion to $100 billion by 2028. The supply response that actually breaks this isn't a capacity pull-forward. It's a demand shock: hyperscaler capex revisions downward, or an architectural shift that reduces memory per accelerator. Neither is visible right now, so the memory bonanza will continue!