Nobody Earns 86% on Memory for Long
INTERNAL MEMO, FOR THE COMMITTEE, DRAFTED BEFORE THE BELL
Subject: Tonight's Micron print, tomorrow's payrolls, and why the two are the same trade
Micron reports after the close tonight. The stock trades at roughly 25 times earnings, which sounds like a bargain until you remember what the earnings are. Last quarter the company booked $41.46 billion in revenue against $9.30 billion a year earlier, and $28.24 billion of GAAP net income on top of it. Consensus for the quarter about to be reported sits near $50.5 to $50.8 billion in revenue and around $31 a share in EPS, with a gross margin of 85.9%, up from 45.7% twelve months ago. Options are pricing a move of about 9%. Guidance for the next quarter, per the Street, is $56.6 billion and 86%.
Please read that gross margin figure again. Memory is a commodity. The product is a wafer of standardized bits that a handful of suppliers stamp out in enormous fabs, and every previous stretch when this industry looked like a great business ended in a swing from shortage to glut. An 86% gross margin on a commodity is a price signal. It says demand has outrun supply so completely that buyers will pay almost anything to keep the line moving. Markets that pay almost anything are also the markets that attract every dollar of capex within reach.
The tell is already on the tape
The Korean memory names have already told the story. Before tonight's report SK Hynix was roughly a third lower over three months and Samsung about a fifth lower, with a fresh leg down on Monday (SK Hynix around 5%, Samsung 4.6%, KOSPI down 2.7%). When the two largest suppliers in the world trade like that ahead of a record quarter, the market is telling you it can do the arithmetic on peak margins. The print will be spectacular. The question is what the stock does with a spectacular print, and the 2018 precedent, when memory shares topped out months before the earnings did, is not a comfortable one.
The rate leg
Now put the print inside the tape it will land on. The 30-year Treasury touched 5.62% on Tuesday, the highest since June 2002. The 10-year sat at 5.29%, a hair from its 2007 peak. A company earning $31 a quarter, capitalized at 25 times trailing, is a bet that the earnings stream is durable. Discount that stream at a 5.3% risk-free rate and the margin for error narrows, because a cyclical peak deserves a lower multiple than a compounder, and rates that high remove the option to argue otherwise.
Nvidia has read the room. It announced a $150 billion buyback and now trades at less than 17 times forward earnings, cheaper than the S&P 500. When the largest buyer of memory on earth buys its own stock at that price, it tells you what it thinks the cycle is worth at these yields. Micron has $30 billion of cash at last count, a record quarter and a capex plan, which is a different kind of firepower.
The demand leg
Two headlines from the last 36 hours belong in this memo even though they are not in anybody's model. The Wall Street Journal reported that OpenAI shelved its next frontier model after internal safety tests, following a weekend of disclosed agentic breaches. Reuters reported on a leaked Anthropic prospectus that flagged existential risk language and plans for roughly $518 billion of infrastructure obligations. Whatever you believe about safety, the buyers of high-bandwidth memory just got two reminders that their own product roadmaps can pause for reasons unrelated to capex budgets. A pause in training runs is a pause in bit demand. Sam Altman keynotes today, so the committee may get a live read before the print.
Before the print, the data
Micron is the second act. The first is a 7:30 a.m. gauntlet. August PCE is expected up 0.4% on the month for the headline and 0.3% for the core, with core year-on-year at 3.3% to 3.4%. One Fed policymaker said last week he wanted to see 0.2% or less. A 0.3% print leaves the October 27 to 28 meeting a coin flip, which is roughly where CME FedWatch already sits after John Williams leaned against an October move and Michael Barr said more hikes are likely. A 0.4% print gives the hawks their arithmetic. The third estimate of second-quarter GDP, expected around 1.5%, is background noise. ADP is background noise with a headline attached.
Tuesday told us the household is not waiting for the Fed. Conference Board confidence fell to 81.9, lowest since 2014. Year-ahead inflation expectations rose to 4.6%. JOLTS openings dropped 256,000 to 7.079 million. Friday's payroll consensus of about 84,000 is roughly half of August's 162,000, and the market has spent a month treating any strong number as a hawkish event and any weak one as a hold. Nobody has a clean view of which way the sign flips this week.
Recommendation
Do not build a position on the assumption that a great print resolves anything. Three outcomes are worth sizing.
If PCE runs hot and Micron beats, yields push higher, the multiple compresses, and the beat gets sold. That is the scenario that fits Korea's price action.
If PCE is soft and Micron beats, the memory complex gets a relief rally that probably lasts until Friday, and the rally then depends entirely on a payroll number that nobody can forecast.
If Micron guides flat to lower on margin, even by a whisker, an 86% gross margin has nowhere to go but down and the stock will make the 9% implied move look modest.
We are underwriting the cycle at exactly the moment the risk-free rate is telling us how much the cycle is worth. Keep the hedge on. Revisit after tomorrow's ISM.