To The Moon

On 5 November the Philadelphia Semiconductor Index closed down 12.25% — its worst session since March 2020. SanDisk fell 22.26%. Micron 17.46%. AMD 17.26%. Nvidia 9.26%. By the close, roughly $2.5 trillion of US tech market value had evaporated in a single day.

Two days earlier, Apple had said something that should have mattered more than any of it: gross margin drops from 47.9% to 45–46% this quarter, because memory costs are eating the P&L.

Put those two facts side by side and the market's reaction looks backwards. Apple's cost line is Micron's revenue line. If Apple is paying more for DRAM, Micron is collecting more for it. Yet both were sold, on the same week, for opposite reasons.

The trigger was policy, not demand

Democrats took both chambers of Congress. Texas's governor-elect called for an immediate pause on data centre development. That sentence does not change how many GPUs are on order. It changes who might be allowed to plug them in.

Within hours the same shock crossed the Pacific. KOSPI opened down 2.24%, broke 8%, tripped a circuit breaker and halted trading for twenty minutes. SK Hynix fell 13%. Samsung Electronics fell 10%. The Nikkei lost more than 3,600 points and closed below 60,000 for the first time since May.

Korea is the purest memory index on earth. When you need cash in a hurry, that is where you sell — and that is a liquidity fact, not an opinion about memory demand.

Apple just published the memory makers’ income statement

Fiscal Q4, ended 26 September: revenue $114.48 billion, up 11.7%. EPS $2.10, up 14%. Gross margin 47.9% — with roughly one point of that coming from tariff refunds rather than operations. iPhone $57.58 billion. Services $31.57 billion, an all-time high. Full-year revenue $479 billion.

Then the guide. Gross margin 45–46%. Services growth 8–9% against a 10.7% consensus. The stock fell 4% after hours, then halved the loss before the open.

The mechanics sit in three numbers. Purchase commitments up $33 billion in the quarter, per COO Sabih Khan. Inventory at $14.54 billion. And list prices up $100 on some iPhone 17 models, with MacBooks raised too.

Translation: Apple is pre-buying memory at today's prices because it expects tomorrow's to be higher, and passing part of the bill to the customer. That is not a company in trouble. That is a company paying a tax it can partially on-charge.

Worth sitting with one contrast. Apple's 47.9% gross margin is less than 60% of SK Hynix's 80% operating margin last quarter. The same wafer, routed to a server instead of a phone, earns several times more. Capacity follows margin. It always does.

The other leg of the same trade

SK Hynix's September quarter: revenue 104.62 trillion won, operating profit 83.33 trillion won, an 80% operating margin. Long-term agreements with 14 customers covering more than half of 2027 and 2028 capacity. DRAM and HBM effectively sold out through 2029.

TrendForce's Q4 contract forecast splits the market cleanly. Conventional DRAM +10–15% quarter on quarter, NAND +15–20%. Server RDIMM +18–23%. Enterprise QLC SSD +23–28%. PC DRAM only +3–8%. Mobile LPDDR and UFS 0–5%.

Read the spread, not the average. Price increases are concentrated where AI capex lands and thinning where consumers live. Apple sits on the thin end and is still losing two margin points — which tells you how much pressure is coming from the fat end.

And here is the bear case nobody sold

DRAM contract prices, quarter over quarter, through 2026: +90–95%, then +58–63%, then +13–18%, then +10–15%. Four quarters, one direction, and that direction is down.

Volume is going the same way. Counterpoint has Q3 global smartphone shipments down 20% year on year — the weakest third quarter since 2012. IDC has PC shipments down 3%.

Rising prices on falling volumes is not a healthy cycle. It is what the final act looks like. That is a real reason to be careful with memory names. It is not a reason to care about a press conference in Texas.

What the market actually sold

Beta. At -12% on an index in one session, the sellers were not expressing a view on HBM pricing. They were de-risking, meeting margin calls, and cutting exposure to the single most crowded trade on the board.

The tell is the other side of the tape. The same night, Wind's China tech leaders index rose 2.88% and the Golden Dragon index added 0.43% — NetEase +4.37%, JD +3.28%, Alibaba +2.67%. Same capital, same session, opposite direction. That is rotation. Liquidation looks different.

Bull Case

1. $33 billion of incremental purchase commitments is forward demand written into a contract, not a survey.

2. Server memory pricing is still accelerating (+18–23% RDIMM, +23–28% enterprise SSD) even as consumer decelerates.

3. Apple raising hardware prices by $100 validates pass-through across the whole hardware stack.

4. A state-level pause is not federal AI policy. Two chambers changing hands does not cancel a capex plan.

Bear Case

1. DRAM pricing decelerated from +90–95% to +10–15% in three quarters, and memory equities discount pricing roughly two quarters ahead.

2. Smartphone volumes -20% and PC -3% mean the volume leg is shrinking underneath the price leg.

3. Policy contagion is the underrated risk: one state pause is noise, three is a schedule slip, and a schedule slip pushes server orders out a quarter.

4. The upside is contracted. Capacity sold out to 2029 means good news already sits in the backlog, not in future pricing power.

5. Apple absorbing two points of margin before passing it on is how unit demand breaks — hardware first, Services attach later.

Strategic Slam

• $Micron(MU)$ fell 17.46% into a quarter where its own contract prices are rising 10–15%. That is a beta event, not a revision. What matters is not the 5 November low — it is the December guide. If management signals deceleration, the multiple compresses regardless of the tape.

• $Apple(AAPL)$ has now published the margin reset. 45–46% is the bar, not a bad quarter. The re-rating risk sits in Services (8–9% versus 10.7% expected), not in iPhones.

• $NVIDIA(NVDA)$ lost 9.26% on a policy headline with an untouched order book. Its customers are the ones whose data centre plans just became politically contested. That is a 2027 question wearing a 2026 costume.

• Consumer hardware — PC and handset — is squeezed from both ends and I would not add here.

• Rule for this week: do not average into a policy headline. Wait for a session that holds the 5 November low and closes green. If that does not arrive before Nvidia prints, the market is telling you the deceleration matters more than the election.

Catalyst Calendar

• 12 Nov — Apple dividend record date

• 16 Nov — Apple dividend payable, $0.27 per share

• Mid-to-late Nov — Nvidia Q3 FY27 results (expected; confirm on the IR calendar)

• Early Dec — Micron fiscal Q1 FY27 results (expected)

• Ongoing — state-level data centre moratorium headlines, Texas first

Verdict

Two things are true at once. The memory cycle is decelerating — a 2027 problem with a two-quarter lead time. The 5 November crash was a policy shock — a this-week problem with no earnings content in it at all.

The market traded the second as if it were the first, and sold the companies that are still getting price increases.

My view: the policy panic is overdone and the deceleration is under-priced. Those two do not cancel out — they sequence. Take the bounce in memory, but size it so a December guide confirming the slowdown does not take you out of the game.

Key Takeaways

1. SOX -12.25% was the worst day since March 2020, and the trigger was an election, not an order book.

2. Apple's 45–46% guide is Micron's revenue guidance under a different logo: $33 billion of new purchase commitments, $14.54 billion of inventory, $100 price rises.

3. SK Hynix's 80% operating margin is the widest this industry has printed at this scale, and more than half of 2027–28 capacity is already contracted.

4. DRAM contract prices decelerated for three straight quarters before Thursday. That was the bear case, and nobody sold it.

5. A KOSPI circuit breaker while China ADRs rallied in the same session tells you this was forced selling, not a change of thesis.

What would change my mind: a Texas-style pause spreading to two more states, or Micron guiding Q1 gross margin down. Which of those hits first — memory pricing, or the policy consensus? And if you are holding through the December print, where is your line in the sand?

Tickers: $Micron Technology(MU)$ $Apple(AAPL)$ $NVIDIA(NVDA)$ $Advanced Micro Devices(AMD)$ $SanDisk Corp.(SNDK)$ $Broadcom(AVGO)$

Sources: Apple Newsroom (FY26 Q4 release, 2 Nov 2026); Reuters coverage of

the post-earnings call; TrendForce contract price forecasts (24–30 Sep 2026); SK Hynix Q3 FY26 results; Counterpoint Research Q3 2026 smartphone data; IDC Q3 2026 PC data; exchange closing data for 5–6 Nov 2026.

Disclaimer: This is my own research and opinion, not investment advice or a recommendation to buy or sell any security. Figures are as reported at the time of writing. Do your own due diligence.