MU
NVDA
SKHY
The AI boom found Nvidia’s weak spot
For most of the AI frenzy, Nvidia looked like the kid who found the cheat code: demand exploded, supply stayed tight, and margins kept stretching. Now memory prices are barging into the room like an uninvited plus-one and stealing the champagne.
Nvidia says its gross margin should slip from 75% in the second quarter to about 74% in the third quarter and then 71%–72% in the fourth quarter. That’s not a tiny wobble — that’s the first meaningful margin downshift of the current AI cycle.
Why this matters more than it sounds
The culprit is high-bandwidth memory, the pricey stuff Nvidia has to buy for every accelerator it ships. And guess who’s benefiting from that inflation? Suppliers like Micron and SK Hynix, who suddenly have the sort of pricing power everyone in tech pretends not to dream about.
The twist is deliciously annoying for Nvidia: the same AI buildout pumping up demand is also making its inputs more expensive. So even though the company says demand is growing around 100% next year, supply constraints limit revenue growth to roughly 70% — and now margin pressure is creeping in too.
The fix is coming... later
Nvidia can raise prices to pass through some of the pain, but not right away. Management says the higher pricing should start helping in fiscal 2028, which means investors are being asked to swallow a margin dip now for relief later.
That’s the real story here: the AI boom isn’t just about who can ship the most chips. It’s also about who gets squeezed when the ingredients get more expensive. Big picture: revenue growth is still wild, but in the next phase of the AI trade, margins may matter just as much as the hype.