All four listed problems have the fix action "remove" but each issue description concludes that there is no actual problem ("No problem here," "No problem," "which is acceptable," "ignore"). Since none of the named sentences require an edit, the article remains unchanged.
Chip Stocks Sell Off Even as Samsung, Micron and Arm Signal an AI Supercycle
A company just posted a quarter more profitable than both Nvidia and Apple. It told investors operating profit would rise 1,800%. Its stock fell 8%.
That's Samsung, on Tuesday. And the instinct is to call it irrational—a market too jaded to reward the best AI numbers on the board. It isn't.
This is a market doing precisely what it's supposed to do when a price has already priced perfection.
Last week we asked whether Nvidia's roughly 75% margin could hold as the AI infrastructure story matured. This week the question mutates into something harder: the margin is holding fine, so why is the tape bleeding? The answer isn't about the moat. It's about the multiple.
Start with what actually happened. Samsung's results were good—historically good—and the sector sold off anyway. Micron dropped about 5%, Intel 9%, AMD 6%, Marvell and the equipment names over 6%. The iShares Semiconductor ETF slumped 5%. Korea's Kospi fell in sympathy, and circuit breakers tripped.
CNBC's framing is the honest one: high expectations overshadowed the results. This is now a recurring phenomenon. It dinged CrowdStrike and Palo Alto Networks. It has repeatedly dinged Nvidia itself.
Look at Nvidia's own last print. EPS of $1.87 against a $1.77 estimate. Revenue of $81.6 billion, up 85% year over year, reportedly ahead of the $79.2 billion consensus. Guidance above the Street. A fresh $80 billion buyback authorization.
The stock dipped after the release. It was the fourth decline in Nvidia's last five earnings reports.
Sit with that. A company grew data-center revenue to $75.2 billion—nearly double the year-ago figure—and beat on every line, and the next session was red four times out of five. That is not the market disbelieving the demand story. It is the market discovering that "beat and raise" no longer clears the bar when the bar is a stock that ran vertically into the print.
That is not the market disbelieving the demand story. It is the market discovering that 'beat and raise' no longer clears the bar when the bar is a stock that ran vertically into the print.
Here is the mechanism, stated plainly. When a stock is up 229%, 273%, 581% in a year, the price already embeds an outcome. Results are measured not against last quarter but against the acceleration the price assumed. Merely excellent is a miss.
Micron is up 242% year to date, with a rolling one-year return over 700%. Sandisk is up 581% this year. AMD is up 273% over twelve months. Arm rocketed 224.4% in the first half of 2026 alone.
Against runs like those, Samsung's 1,800% profit guide isn't the ceiling. It's the floor investors had already paid for.
Now the more interesting fight, because it clarifies what this selloff is not about.
On one side, Robert Castellano argues Wall Street is misreading the whole thing. His case is built on demand indicators, and it's strong. Global HBM shipments are projected to climb from 1.2 billion gigabits in 2022 to 38 billion by 2027—a more-than-thirtyfold increase, still accelerating through the exact years investors claim demand is softening.
HBM is a clean read on AI compute because it's purchased almost exclusively for AI accelerators. If demand were collapsing, Castellano notes, the forecast would already be flattening. It isn't. His conclusion: the market erased over $1 trillion in semiconductor value on a slowdown fear the industry's own data doesn't support.
On the other side, Michael Burry shorted Micron near $1,052 on July 1st. Within days the stock fell nearly 20%. The thesis: memory is cyclical, the rally is a bubble, and a correction is coming.
Here's the thing. Burry's core observation is correct—memory prices peak and fall; that's what memory does. But his timing runs straight into Micron's numbers.
Fiscal Q2 revenue hit $23.86 billion, almost triple the year-ago quarter. EPS of $12.20 crushed the $9.19 consensus. Gross margins expanded to 72.6%, operating margins to 65.6%, return on invested capital to 58.6%. Trailing free cash flow of $26.2 billion, up 1,291%, against $20.3 billion in net cash.
That is not a company whose fundamentals are cracking. HBM remains supply-constrained across all three makers, and leading-edge capacity takes 12 to 18 months to build. The cycle doesn't pivot on a headline. So Burry is early, maybe by a lot.
But watch what Castellano and Burry are actually arguing about. Both are debating demand direction—one says it's accelerating, the other says it's about to roll over.
The tape is debating neither. The tape is debating arithmetic.
Look at Micron's own valuation and the whole thing snaps into focus. The stock trades at a trailing P/E of just 22x. That is not expensive—not for a company whose revenue grew 167% year over year.
Its forward P/E is 155x.
Read those two numbers together, because they contain the entire story. A 22x trailing multiple on current, real earnings. A 155x forward multiple that isn't a bet on growth—it's a bet on decline. As AInvest puts it, the forward figure is "the market's bet that the current earnings explosion will reverse dramatically."
That is the market pricing an earnings reversal it cannot yet see in the data. Not disbelieving AI demand—embedding a cyclical rollover into the denominator. The selloff isn't a rejection of the demand story. It's the tape refusing to pay 155x for earnings it assumes are near a peak.
Arm shows the same arithmetic from the growth side. Revenue was $4.92 billion in 2026. Management guides to $25 billion by 2031, driven almost entirely by its new AGI CPU—a 5x jump.
Take that guidance at face value. Management projects $9 in EPS in 2031. Against a $334 stock, that's a P/E over 36—five years out, assuming every aggressive target is hit.
You are paying 36x for earnings that don't exist yet and depend on a chip that has to win. The valuation has already borrowed years of the demand curve and spent it.
So the fundamentals thread we opened on July 3rd resolves cleanly. Nvidia posted a 72% net income margin last quarter. AMD posted 14%. The moat is intact; the margin is holding; HBM demand is setting records.
None of that is what corrected. What corrected was expectations.
This is the discipline to carry out of the week. A selloff on good-but-not-heroic news, in stocks up triple digits, is not bubble deflation—it's a re-rating of expectations back toward what earnings can plausibly deliver.
The DeepSeek chip headline and the SK Hynix listing added noise, and CNBC flagged the obvious recalibration after memory's historic run. But those are triggers, not causes.
The cause is that a stock cannot compound at 200-700% a year and also keep beating the bar that pace sets. Something has to give, and it's usually the multiple, not the business.
Which tells you what survives. Not the names with the best demand story—almost all of them share the same demand story. The ones that hold are the ones where the price hasn't already spent five years of the curve.
Micron at 22x trailing is arguing one thing. Micron at 155x forward is arguing the opposite. Own the version of that trade where the arithmetic still has room to be wrong in your favor.