Intel's Worst Day in 50 Years: PC Winter Arrives, Server Fire Isn't Enough

1. What happened: a beat buried by a winter warning

Intel beat Q3 and lost 29% anyway — because the guide told the truth about PCs. Q3 revenue was $16.51B, up 21% year-over-year, with adjusted EPS of $0.52 vs $0.40 expected. Data-center and AI (DCAI) revenue jumped 78% to $7.3B — Intel's strongest data-center growth in 15 years. That should have been a celebration.

Instead CFO David Zinsner said the PC slowdown continues "at least through the first half of next year," guiding Q4 to $15–16B revenue and $0.25 adjusted EPS vs $17.1B and $0.43 expected. Client Computing (CCPG) fell 5% to $8.1B and missed. Morningstar now expects PC units down 12% in 2026 (down 20% in the second half) and another 10% in 2027. The stock, up ~150% year-to-date into the print and down ~23% already in October, finally capitulated — on pace for its worst day since 1974.

My opinion: the market is right to punish the guide, not wrong to have loved the quarter. This is a two-company story — a server company on fire stapled to a PC company in winter — and the PC side still decides the quarter.

2. The core argument: servers can't carry PCs yet

DCAI's 78% growth is the most bullish Intel number in a decade, and it still isn't enough. Do the arithmetic: DCAI did $7.3B in Q3; CCPG did $8.1B. In a normal quarter PCs are 50–60% of revenue. A 5% PC decline on a bigger base plus surging memory costs overwhelms even explosive server growth, because every PC Intel sells now carries a memory tax it doesn't capture.

Counterpoint Research pegged Q2 global PC shipments down 4% on soaring memory and component costs. Intel confirmed it: high memory prices are suppressing unit demand, and Intel "doesn't expect the crunch to abate soon." Morningstar agrees. That is the load-bearing fact — if memory prices roll over, the whole bear case cracks. Until then, Intel sells fewer boxes at thinner margins while its hottest product (Xeon) is supply-constrained on substrates and fab capacity.

Then came the one-off that isn't quite a one-off: an explosion at Fab 28 in Kiryat Gat, Israel killed two workers and shut the plant for an extended period. Cost: ~$600M in lost revenue over two quarters (~$300M in Q4) plus 200bp of Q4 gross-margin pressure, concentrated in 10nm Xeon supply — exactly the product you least want to short. Management called it "tragic"; traders heard "tight into 2027."

3. Finance: the quarter was good, the balance sheet paid for it

Intel's P&L is healing; its share count is the scar. Gross margin hit 43.0% GAAP (44.1% non-GAAP), up ~4–5 points year-over-year. Operating margin hit 15.3% GAAP (19.6% non-GAAP). Operating cash flow was $5.1B in the quarter. DCAI operating income tripled to $3.04B; Foundry losses narrowed to -$1.85B from -$2.32B.

But Intel completed a $23B public equity offering in August, plus the U.S. government's 10% stake, Nvidia and SoftBank investments, and $13B of new term debt in nine months. Shares outstanding are up from ~4.51B to ~5.31B diluted — roughly 18% dilution in a year. Cash looks fortress-like ($21.8B cash + $32.2B short-term investments), yet retained earnings fell from $49.0B to $37.0B. You are buying a better company with many more claims on it.

Valuation: at $65, Intel trades well below Morningstar's fresh $85 fair value (cut from $105, no moat, Very High uncertainty). That sounds cheap until you note forward EPS is unanchored — GAAP Q4 guide is $0.10 — and the dividend yield (~1.1%) doesn't compensate for 2.3 beta. My fair value: $75–85, centered on $80 — in line with Morningstar — but only if DCAI holds 30%+ growth through H1 2027.

4. Competitive position: the only American foundry, still without the customer

Intel's strategic value has never been higher; its commercial proof has never been thinner. It is the only U.S. maker of advanced logic, in high-volume 18A production in Arizona, with 14A PDK 0.9 released, High-NA EUV work with ASML, and Oracle multi-year Xeon collaboration plus Meta Muse Spark support. CEO Lip-Bu Tan promises "a couple of customers ready to announce" for Foundry in the next two quarters and teases accelerating the Ohio build-out — possibly with SK Hynix memory on site.

None of it is signed. "No major external customers for its new chips" remains the sentence that matters. Meanwhile AMD fell ~10% to $509 and Dell ~11% to $423 on Intel's PC warning — the whole x86 chain repriced — while Nvidia held up far better at $206 (-3.8%). The market's verdict: this is an x86-PC problem, not an AI problem. AWS growing 39% to $46B (see Amazon below) proves AI demand is intact; Intel just monetizes the wrong half of it today.

5. Why Amazon matters to the Intel thesis

Amazon's print is Intel's alibi — and its indictment. AWS grew 39% to $45.98B, its fastest in nearly five years, beating $45.3B expected. That confirms agentic-AI CPU demand (Muse, Dots) that favors Xeon. But Amazon missed EPS ($1.89 vs $1.96), missed ads and retail, burned -$42.4B trailing free cash flow, spent $77.3B in quarterly capex, held 2026 capex at $220B, and warned it will "spend meaningfully more in 2027." AMZN fell to $239 (-1.9%, off worse after-hours lows).

Read-across: hyperscalers will keep buying Xeon (good for DCAI) while refusing to ease memory supply or PC demand (bad for CCPG). Our WATCH-SOFT-1029 gate — enter software only if AWS ≥38% — just triggered to the upside at 39%. That numbers Intel's pain as cyclical (memory + units), not secular (AI demand). It also keeps SPEC-AVGO-1 and MACRO-LITE-1 alive: custom silicon and memory, not merchant PC CPUs, capture the spend.

6. Technicals: how a 50-year worst day trades

Capitulation is a process, not a print. INTC hit $64.11 intraday, down from a $92.07 prior close, on what CNBC called the worst day in 50+ years. The 52-week range is $32.89–$142.35. The August $23B raise and the October slide from ~$120 (Jefferies' old target, now cut to $85) to $65 left no nearby support — just air, then the $50–58 zone where the summer breakout began, then the $32–40 base.

Analysts are chasing, not leading: BofA and Roth cut to Neutral, Jefferies cut to $85 Hold, Melius to $325 (on its own scale convention) — a "substantial reset." Expect more cuts Monday. In my experience, the first down-50%-volume day rarely marks the low; the retest a week later does. Do not buy green on Monday. Let $52–58 come to you.

7. Bull, base, bear — with triggers

Base (50%): PC winter lasts through H1 2027, DCAI grows 30–40%, stock bases $55–75. Trigger: Q1 guide flat, memory prices stay high, no foundry customer. Action: patient scale at $52–58, sell $80–85.

Bull (25%): Memory rolls over + a foundry customer is named + Xeon supply unlocks. Revenue re-accelerates to $17B+ in H1 2027, stock reclaims $95–110. Trigger: SK Hynix Ohio confirmation, 14A design win, or DRAM/NAND spot prices falling 15%+. Invalidation of WATCH becomes a buy.

Bear (25%): PC units fall faster than -10% in 2027, Israel drag extends past Q1, another raise or guide-down. Stock tests $35–45. Trigger: Dell/HP cuts orders, Q1 guide below $14.5B, or gross margin below 35%. Daily close under $48 confirms — walk away.

8. What would prove me wrong

I am wrong if Intel closes two consecutive weeks above $85 on rising DCAI backlog while PC units stabilize. Specifically: (1) November channel checks show PC units flat-to-up QoQ, (2) Intel names a funded external 18A/14A customer before year-end, or (3) Q1 2027 guide comes in above $16.5B with gross margin above 40%. Any two flip this WATCH to a BUY. Conversely, a weekly close below $48, or Dell cutting full-year PC outlook, confirms the bear — I will say so plainly.

9. What to do — including our open calls

Today is a risk-off day demanding book discipline, not heroics. QQQ -3.1% to $693, SPY -1.6% to $757, fear-greed 30 (Fear, down from 51 Neutral yesterday), BTC $77,490 (-5.2%) on the record $6.3B Binance hack (FBI/Binance working to recover ~$3B frozen; North Korea blamed). Hormuz tanker attacks have XOM (+1.4% to $185.80) and CVX (+1.7% to $235.28) at all-time highs — the rotation out of tech into energy is today's tape.

• INTC: no chase. Wait $52–58. If you must trade the bounce, use 1/4 size with a $60 stop — but that is gambling, not investing, and I don't recommend it.
• SPEC-MU-1028 (our memory long): SNDK's 85.1% margin + $12.4B midpoint guide is the confirmation we required — upgrade the thesis, but respect price: MU at $884 is now below our $930–975 half-size zone after a -5.3% sympathy drop. Hold existing half; add second half only on MU reclaiming $930 with volume. Kill stays daily <$860.
• SPEC-GOOGL-1029: holding — GOOGL $282.31 in the $280–295 washout zone (-0.5% today, resilient). No change.
• SPEC-BTC-1027: live $77,490 has pierced our $77,600 kill by $110 on hack-driven liquidations ($1.57B) plus tech risk-off. My rule: a daily close below $77,600 confirms the stop; intraday is not a close. Reduce to 1/4 if it closes there; full exit on a second close below. Honest flag — this one is stressed.
• Fresh trigger: AMZN AWS 39% satisfies WATCH-SOFT-1029's ≥38% gate. Next session: pick ONE software name with accelerating subscription revenue for a focused entry. Not today — let the INTC/AMZN dust settle.