To The Moon

Buy the Dip, or Get Off the Train? My Honest Take on the AI Selloff

Let’s not pretend this pullback came out of nowhere. Chip and memory names got hit hard over the past few weeks — Micron ($MU) dropped double digits in a single session ahead of earnings, SanDisk shed similar ground, SK Hynix fell sharply in Seoul, and the Nasdaq had its roughest stretch since the AI rally began. Headlines are calling it “the AI bubble finally popping.” I don’t think that’s the right frame, and here’s why.

Two different stories are getting mixed together

Story one is valuation. Some AI-adjacent names ran up so far, so fast, that any pause in the narrative was going to trigger profit-taking. That’s normal market mechanics, not a verdict on AI itself.

Story two is fundamentals, and this is where I think the panic overshoots. Look at what Micron actually disclosed around its last earnings: 16 Strategic Customer Agreements — hyperscalers and automotive suppliers locked into five-year, take-or-pay contracts through 2030, backed by billions in cash deposits and letters of credit. Those aren’t soft “letters of intent,” they’re binding commitments with pricing floors set above Micron’s best historical margins. That’s not the profile of a company riding a hype cycle that’s about to evaporate — that’s a company that’s structurally de-risked its next several years of earnings.

So why the selloff?

A few real reasons, worth taking seriously rather than dismissing:

• Nervousness ahead of earnings prints (SanDisk, Micron) that has historically caused overreactions in both directions

• Rotation out of richly-valued tech into defensive sectors ahead of Fed commentary

• Genuine, fair questions about whether AI capex is outrunning realized enterprise ROI — an MIT study earlier this year found the large majority of companies still haven’t seen returns on their AI spend

• Export-control and geopolitical noise around semiconductors, which adds a layer of policy risk that’s hard to model

Where I actually land

I’m not “buying the dip” indiscriminately, and I’m not “running for the exit” either — I think both framings in the poll are a false binary. My take:

1. Memory (MU, and to a lesser extent SanDisk) looks like the most defensible pullback to me. The demand side (AI training and inference both eat DRAM/NAND at scale) is contractually locked in, not just narrative. Jensen Huang’s comment that the industry could still expand 5–10x before a supply glut is a real tailwind if even partially right.

2. Pure-play “AI story” stocks with no earnings to point to are the ones I’d actually be cautious on. Those are the names most exposed if sentiment — not fundamentals — keeps driving price.

3. Macro risk is the wildcard. Fed policy and any escalation in chip export restrictions could move this whole sector more than company-specific news does in the next quarter. That’s the piece nobody can hedge with a spreadsheet.

Bottom line: this doesn’t look like 2000 to me. It looks like a re-rating of expectations inside a genuine buildout, with memory chips specifically showing more contractual backbone than the market is currently pricing in. I’m treating the weakness in $Micron Technology(MU)$  as an opportunity to build a position gradually rather than all at once, and staying away from the more speculative, story-driven names until they show actual revenue to match the multiple.

Curious where the rest of the community lands — are you seeing this as capitulation, or just a healthy reset after a 40%+ run?