For most of this week, the story on Wall Street was simple.

AI stocks got too expensive, Washington just got less predictable, and the whole trade started unwinding.

Chips led the carnage. Micron lost more than $100 billion in market value in a matter of days, right alongside Nvidia, CoreWeave, and the rest of the AI complex.

Then Micron’s own numbers from six weeks ago quietly resurfaced this week. Sold out memory. Record backlog. Pricing that’s still climbing, not falling.

That is not the balance sheet of a company that deserves to trade like a meme stock caught in a panic.

Today I want to walk through why Micron (MU) got dragged into a selloff that, on the numbers, doesn’t belong to it, and what would actually have to go wrong for the bears to be right.

Missed Friday’s Playbook on Palantir?

Cathie Wood argues that the current “wall of worry” around AI spending, debt, cash flow, and data-center construction is actually healthier than the blind optimism seen during the dot-com bubble. She still expects the bull market to continue into 2027, although she sees rising bond yields and the cost of capital as real risks. The bigger takeaway is that AI enthusiasm is being stress-tested, but Wood sees volatility as an opportunity to concentrate into her highest-conviction names.

Trying to wait for the “perfect” entry point usually backfires because there is always a reason to stay cautious. Instead, long-term investors should keep dollar-cost averaging with money they will not need soon, since falling prices allow each contribution to buy more shares. In plain English, a bear market may hurt in the short term, but consistent investing turns lower prices into an advantage rather than a reason to panic.

Arista’s Q3 revenue jumped 58.5% as hyperscalers expanded high-speed Ethernet networks for AI clusters, and management now expects AI fabric revenue to roughly double to $7 billion in 2027. The opportunity is huge because AI data centers need increasingly sophisticated networking, not just GPUs. The catch is valuation and concentration: Microsoft and Meta each account for more than 10% of revenue, while Arista has also committed heavily to future inventory and suppliers.

We gave GPT-6 Astra the same job twice: analyze 100 sales calls and draft the follow-ups, once in Ask Attio and once in ChatGPT.

Ask Attio finished in 4 minutes 49 seconds. ChatGPT took 26 minutes 11 seconds, about 5.4 times as long.

The most ambitious GTM work requires more than frontier intelligence. It requires architecture that can put that intelligence to work at scale.

Learn about Attio’s hybrid architecture.

Micron shed more than $100 billion in market cap this week.

That number got filed under one headline: the AI trade is cracking, and chips are leading it down.

Here’s the part that headline leaves out. MU didn’t report a bad quarter. It reported the best one in its history.

Revenue of $54.23 billion last quarter, up 379% year over year. Full fiscal 2026 revenue of $133.19 billion, nearly quadruple the $37.38 billion it did the year before.

Not a recovery story. A different company entirely.

So the real question isn’t whether AI memory demand is real. The numbers already answered that.

The real question is whether this week’s selloff is Wall Street pricing in a future Micron hasn’t shown us yet, or whether it’s panic that forgot to check the filing.

I keep coming back to one number. Micron’s long-term HBM customer commitments jumped to $32 billion, up from $22 billion as of June.

That’s not a forecast. That’s signed commitments, growing by $10 billion in a single quarter.

Remaining performance obligations, the total contracted revenue still owed to Micron, now sit near $150 billion.

CEO Sanjay Mehrotra said it plainly on the call. A large portion of 2027 HBM production is already sold out, at prices above where 2026 sold.

That last detail matters more than the revenue number.

Because a shortage that still commands rising prices a year out isn’t a short squeeze. It’s a structural supply gap.

Management guided fiscal Q1 2027 revenue to $61.5 billion, with adjusted gross margin near 86%, and called for sequential revenue growth every quarter this fiscal year.

That is a company telling you its best quarter is still ahead of it.

Here’s what complicates all of that.

Memory is the single most cyclical corner of the chip industry, and it has a long history of flipping from “sold out” to “oversupplied” faster than anyone on the earnings call expects.

Micron investors who lived through the 2018 to 2019 downturn know exactly what that transition feels like.

A sold-out backlog is a snapshot of today’s contracts, not a guarantee those contracts survive if the customer on the other end changes its mind.

And the customers on the other end are hyperscalers, the same companies this week’s broader selloff was already questioning.

Not because Micron’s demand is fake. Because Micron’s demand is borrowed from someone else’s capex budget.

If Microsoft, Google, Amazon, or Meta so much as hint at slowing AI infrastructure spending, Micron’s backlog doesn’t vanish overnight.

But the market will price the fear of it vanishing long before it actually does.

That is the bet this week’s selloff is actually making.

Micron got caught in the same current that dragged the entire chip complex lower earlier this week, and for a few sessions it looked like the stock might simply keep going.

It bottomed intraday near $676 on Friday before buyers finally stepped back in, and MU closed the week at $690, up more than 5% on the day.

That’s still a long way down from the $1,255 high it touched earlier this year, and the 52-week range, $192.59 to $1,255, says everything about how violently this stock can move in both directions.

The $676 level held once.

If the broader chip unwind resumes next week, the real test is whether that floor holds again, or whether MU slides back toward the psychological $600 mark.

Hold above $690 into next week and the stock starts repairing the technical damage from Thursday’s rout. Lose $600 and the next place buyers have historically stepped in isn’t close.

The entire bull case runs through four companies’ spending plans, not Micron’s own.

Micron’s demand is borrowed from someone else’s budget, and that budget belongs to Microsoft, Amazon, Google, and Meta.

If all four reaffirm AI infrastructure spending in the next few weeks, the sold-out backlog stays exactly that. The wildcard is that a single cautious comment from just one of them gets treated by the market as a signal from all four, even while the other three keep spending at full speed.

Micron says 2027 HBM is selling above 2026 prices.

That claim gets tested the moment any customer successfully negotiates a discount.

Watch for whispers of price concessions before the next earnings call, that would be the first crack in the shortage story. The wildcard is competitor capacity: if SK Hynix or Samsung add supply faster than expected, the pricing math changes for everyone, Micron included.

Guidance calls for $61.5 billion in fiscal Q1 2027 revenue.

A clean beat and raise confirms the supercycle thesis and probably reprices the stock back toward its highs.

A miss, even a small one, gets read as the first sign the AI memory boom has a ceiling. The wildcard is Micron’s own habit of pre-announcing strong results ahead of the formal print, which would move this timeline up.

This is the one variable sitting entirely outside Micron’s control.

A newly divided Congress talking tough on AI infrastructure spending is very different from one that actually passes something that slows it.

Nothing in Micron’s backlog changes until policy turns from rhetoric into law. The wildcard is that markets don’t wait for that distinction, they price the headline risk now and sort out the details later.

I flagged this exact tension back on September 7, in Micron’s Backlog Has An Expiration Date.

The backlog didn’t expire. It grew by $10 billion in a single quarter.

The stock still sold off like the backlog was the problem.

This week’s selloff was about Washington and sentiment. Micron’s quarter was about contracts and cash.

Those are two different stories, and the market briefly collapsed them into one.

And honestly? That’s usually where the best setups show up, right at the point where a stock’s price stops reflecting its own business and starts reflecting whatever story it got filed under.

Micron isn’t CoreWeave, burning cash on a promise. It isn’t Palantir, trading on a narrative multiple.

It’s a company that just told you, in writing, with signed contracts, that its best quarter is still in front of it.

The market can keep selling the ticker. The backlog isn’t going anywhere.

With Lovable, you'll have a working version of your client's site ready for the first meeting. Plus:

Start building, $5 off

_{^{Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.}}

The hardest moment to hold a position is exactly when the market stops distinguishing between a company and its sector.

Micron didn’t get cheaper this week because its business got worse. It got cheaper because it shares an index with companies whose businesses actually might be.

That gap, between what a stock is doing and what its neighbors are doing, is where patient investors get paid.

It rarely feels that way while it’s happening. In the moment, every selloff feels like information.

Most of the time, it’s just correlation pretending to be analysis.

The checks come later, after the backlog either holds or it doesn’t.