Another record quarter from MU, down 2%: why estimates are not expectations

edit: uhh I know it closed up 3%. my point is that the reaction to epic results was pretty much nothing.

I'm a professional analyst, started covering semis this year, though I have over a decade of experience in equities.

I've seen a number of posts/comments about how companies could crush estimates so handily and not pop on the news, so I figured I'd post a quick explainer with mechanics, since I just went through the numbers.

Micron printed $54.2B revenue vs \~$51.1B consensus, $33.42 EPS vs \~$31.61, 87% gross margin, and guided next quarter to $61.5B vs \~$57B expected. These estimates were in line to my expectations as of June, though I had since updated them prior to the Q4 release.

The stock is down \~2% today. If that confuses you, you may be equating estimates and expectations.

Here's the mechanics.

**Guidance:** At Q3 results back in June, Micron guided Q4. Revenue guidance for a memory company is basically: (DRAM bit growth × DRAM price growth) + (NAND bit growth × NAND price growth). That's the whole model. Two product lines, two levers each.

Bit growth is partially knowable. Node migrations are planned quarters in advance, wafer starts are known, and management guides bits conservatively by habit. Average selling price (ASP) is the unknowable half. Contract prices are negotiated with big customers - not entirely known at guide time, but stickier than spot. Spot is a daily auction and is basically unknowable three months out. So guidance is: conservative bits times a price guess that's half contract, half hope.

Then the information leaks, all of it in public, before Micron reports:

July 29 SK Hynix reports Q2. Record everything. Both DRAM and NAND prices up big quarter-over-quarter, 76% operating margin.

July 30 Samsung reports Q2. Record quarter for the memory business, and tells you supply stays tight into 2027.

July 31 Kioxia reports. NAND ASP up \~70% quarter-over-quarter, bit growth in the low single digits.

Plus the monthly contract-price prints and the weekly spot screens running through August. By the time Micron reported on September 30, the price half of the equation was mostly public information. The only genuinely unrevealed parts were Micron-specific: its own bit shipments, its product mix, HBM pricing, and what management says on the call.

So what did Micron "beat"? It beat consensus estimates. But consensus estimates are a stale snapshot - a bunch of models built off June guidance. Market expectations are a live number that updates every time a peer reports, every monthly print, every spot move. The market had already bought the pricing story. the stock is up \~3x this year. Beating June's numbers on September 30 is beating where the herd was standing three months ago, not where anyone actually was yesterday.

My overall thoughts: Heavily long. Micron is being valued like a cyclical stock. And it is. But it's a lot less cyclical than ever before. The current share price reflects a 70% reduction to selling prices within 5 years, an assumption consistent with past memory commodity cycle drawdowns. But it's also a nearly impossible outcome at this point.

Yesterday, mgmt disclosed strategic customer agreements on 35% of revenue through 2030. These are take or pay contracts. Continuing to target 50% through 31. The pricing on these is not flat, but rising. To state the obvious, at 35% of revenue locked in with floors, RAM prices would have to fall to 0 to get anywhere close to a 70% reduction to selling prices on a realized basis.

I model a 40% reduction to selling prices by 2031, and still come to a valuation of \~$1,900/sh. But y'kno, disclaimer, I'm *a* financial analyst, not *your* financial analyst.