Intel's 17% After-Hours Drop Is a Memory-Cost Story: Half of Revenue Sits in the Segment the AI Build Is Shrinking

Intel beat on every headline line: Q3 revenue of $16.51 billion, up 21% year on year, and adjusted EPS of $0.52 against the $0.40 consensus, on a 44.1% gross margin against the 42% it had guided. It then guided gross margin down 610 basis points to 38.0% on a $15.0–16.0 billion revenue range, 9% below the Street's $17.1 billion, with adjusted EPS of $0.25 against $0.43 expected. The shares fell as much as 17% in extended trading that evening, after a 3.9% drop in the regular session and on top of a roughly 23% slide over October.

The earnings beat was a margin beat, not a demand beat: revenue cleared consensus by 0.2%, and the $0.52 came from a 44.1% gross margin against a 42% guide. Add the new outlook to the three reported quarters and it puts 2026 revenue at roughly $61.7 billion, about 2% below the $63.2 billion consensus. The guide also absorbs about $600 million of lost revenue from the explosion at Intel's Israel fab, where two workers were killed, and management said server CPU supply will still cap shipments in the fourth quarter. Note the honesty in the yoy line: the guide is still growth against a weak year-ago base, which is why the stock fell on a guide-down rather than on a collapse.

Global PC shipments fell 22.4% year on year to 55.4 million units in the September quarter, the weakest third quarter since the first quarter of 2023. Gartner attributes the collapse to DRAM costs that surged fivefold year on year in the second quarter and were passed fully into PC pricing; Counterpoint Research put the decline at 21% to 57 million units on the same read. Every x86 PC maker sold far less: HP shipped 30.9% fewer units, Dell 25.4% fewer, Lenovo 22.4% fewer. Apple was the outlier, down just 7.1% while taking 1.7 points of share.

Intel's client and physical AI revenue fell only 5% to $8.1 billion in the same window. If its unit volume fell roughly in line with the market, that combination implies an average selling price and mix gain of about 20% — the segment held its revenue line with price, not with demand. HP made the same trade explicit: Personal Systems revenue rose 18.5% to $11.77 billion on a 40.8% increase in average selling prices while PC units fell 15.8%, and segment operating margin slipped to 4.6% from 5.4%. The inference is Intel's, not management's: unit volumes inside the client segment are not disclosed.

Cost inflation has now reached the CPU itself: October reporting put Intel's flagship notebook processors up as much as 15% to 20%, with Qualcomm up about 10% and AMD expected to follow. That matters because the price umbrella is self-limiting. Every point of CPU and memory inflation lifts the finished PC price, and the finished PC price is what is now breaking demand.

In the same quarter, Micron turned $54.2 billion of revenue into $47.0 billion of gross profit, an 86.7% margin, and $43.8 billion of operating income. SK hynix converted ₩104.6 trillion of sales into ₩83.3 trillion of operating profit, roughly 80%, said demand will outrun supply through 2030, and has locked long-term agreements with14 major customers covering more than half of its 2027 and 2028 capacity. Micron's chief executive said in October that a large part of 2027 output is already sold at materially higher prices. Downstream, HP earned $537 million of operating profit on $11.8 billion of Personal Systems revenue — 4.6% — while carrying the same memory bill.

At the Oct 29 close, Intel's roughly $485 billion market value was about 8.5 times trailing revenue and Micron's $1.11 trillion about 8.4 times, valuing a 74.6% operating margin and a 0.1% operating margin almost identically. That is the argument against reading this as a foundry story. The equity is a bet on an18A and 14A payoff, because the client engine that used to fund the build is the segment being repriced out of existence.

Intel's foundry segment did what the bulls wanted in Q3 — revenue of $6.7 billion, up 58% — and still lost $1,853 million. Consolidated operating income was $2,522 million on $16.51 billion of revenue, and the quarterly release never discloses how much of that foundry revenue came from outside Intel, so the $6.7 billion cannot be read as third-party foundry demand. That is the gap the 17% move priced.

For Intel the arithmetic is unforgiving: nearly half of revenue comes from a segment whose units are falling about 20% a year, and the company uses price increases to defend revenue in a market where units are collapsing. That defence is what protects the current quarter and what accelerates the next one. The only offset — data-center revenue up 78% year on year — is capped by supply rather than demand, so it cannot step in until Intel converts capacity it is financing with equity.

Over one to three years the branch points are simple. If PC units merely stabilise in 2027, the price umbrella holds, client margin survives, and the foundry build is funded by a recovering client base. If units keep falling, a $1.85 billion quarterly foundry loss lands on a shrinking base, and the $23 billion of equity already raised becomes a bridge rather than a runway. The most under-appreciated risk sits at the top of the chain: memory makers are selling scarcity into2028, so the same demand shock that would rescue PC units would also take the profit pool back.