Daily Stock Analysis • Monday, November 9, 2026 • Intraday prices ~11am ET / 16:00 UTC • By @dailyanalysts

Nvidia's $4.5T CEO Scandal: Why the Stock Underperforms Into Nov 17 — and What to Own Instead

中文版 Chinese Edition →

The call: NVDA — TACTICAL UNDERPERFORM / trading short $180–$188, target $155–$160, kill on daily close >$205, 1–3 weeks, SPECULATIVE-HIGH.

Why now: Board opened independent probe Nov 8 after Huang admitted affair with alleged Chinese spy; premarket -5% while SOXX +7% — an 11-point split no megacap survives without an overhang.

The disagreement: Wall Street kept every Buy ($315–$515) and called it sentiment; we say Washington turns it into revenue — H200 China license revoked, Blackwell-to-China odds to zero, SEA loophole closed.

The level that changes everything: $170 intraday low — lose it into Nov 17 earnings and $164 52-week low opens; reclaim $205 and the scandal is priced and I cover.

What just happened is a governance crisis, not gossip

Nvidia's board opened an independent investigation Sunday after Jensen Huang admitted an affair with a woman U.S. intelligence says is a Chinese spy — and Congress wants him under oath in 9 days. That sentence explains why NVDA fell 3.8% to $179.62 Monday morning while AMD jumped 11.4% to $500.42, Broadcom jumped 9% to $320.23, Meta jumped 5.2% and the SOXX jumped 7.3%.

The facts, from the CNBC account of Huang's email and the board statement: Huang told staff Sunday "I was unfaithful to my wife... There is no excuse," and the board said it "is initiating an independent investigation" into whether Huang was the target of a foreign intelligence effort, adding it has "no reason to believe" secrets were compromised — but will examine all information. The woman, Alice Huang, is affiliated with Megaspeed, an Nvidia cloud customer previously investigated (Singapore-U.S., October 2025) for helping restricted Chinese firms acquire Nvidia AI chips.

Timeline that matters: Oct 29 New York Times first reports targeting; Nov 5 three House Republican chairs demand a closed-door briefing Nov 18; Nov 8 admission + probe; Nov 17 Q3 FY27 earnings after close; Nov 18 Hill briefing; late-November Nvidia conference in Washington. Earnings and interrogation within 24 hours — that sequencing is the trade.

Trump called Huang "an honorable guy" and China called the claims "smears," but former SEC chair Gary Gensler put the market point bluntly on CNBC: the premarket drop shows investors consider it material. A $4.5 trillion company (per CNN quote page) does not get a personal-news discount — it gets a risk-premium repricing.

Nvidia the business is still a monster — which is why this hurts

Q2 FY27 (reported Aug 26) was $96.2B revenue, +106% y/y, with Data Center $89.0B, +117% y/y, and Q3 guide at $108B — the greatest AI print in history, and the market already owns it.

Business in one paragraph for the finance, not the tech: Nvidia sells the full AI factory — Blackwell and Blackwell Ultra GPUs shipping at scale, Vera Rubin now in full production per the earnings call, CUDA software locking in hyperscalers and sovereigns, gross margin ~74.7%, TTM P/E ~23x on trailing $7.91 EPS for 83% revenue growth and 204% 3-year EPS growth. 52-week range $164.27 (Mar 30) to $243.37 (Oct 6). That is cheap for the growth — until you remember the growth assumes China optionality and frictionless Washington.

My opinion: the moat did not crack this weekend; the multiple's permission slip did. When the CEO is the company — founder, salesman-in-chief, Washington envoy arguing for broader China exports — any question about his judgment becomes a question about export licenses, and export licenses are revenue.

Washington turns headlines into lost revenue

The most underpriced sentence Monday came from Baird: expect the U.S. to repel the H200 China license, kill any Blackwell-to-China hope, and close the Southeast Asia transshipment loophole.

Read the full analyst roundup and the pattern is clear — everyone holds their Buy, everyone warns the overhang lasts:

Second-order effect most miss: newly empowered Democrats after the midterms now have fresh ammunition for oversight (see CNBC on the Democratic angle), while Republicans who demanded the briefing want tighter controls. Both parties now have a reason to squeeze Nvidia on China. Bipartisan agreement on only one thing — less China revenue — is the worst Washington outcome for a bull.

Quantify it: Q2 guidance already assumed zero Data Center compute revenue from China. Revoking H200 formalizes zero and removes the call option the $500 targets implicitly carry. Closing the SEA reseller path hits the grey-market buffer analysts quietly model as 5–8% of Data Center sales. That is not a quarter-miss — it is a 2027–28 TAM haircut worth multiple turns of P/E.

Competitors smelled blood the same morning

Google launched two specialized TPUs Monday — and both are aimed directly at Nvidia's most profitable jobs.

Per Quartz/Bloomberg on the launch: the TPU 8tx does confidential computing (chip-level encryption, paired with an AMD CPU — normally data must be decrypted in memory to be worked on) for governments and regulated industries; the 8i Max packs two TPUs for long-context work (drug discovery, math, coding). Both are reconfigurations of existing silicon, available next year, unveiled by chip chief Amin Vahdat. Four TPU variants in one year after a prior cadence of one every two years. Every TPU unit is already committed, per Google's earnings — supply, not demand, is the constraint.

Why this timing matters: Google designs with Broadcom and builds at TSMC, sells via Google Cloud and Oracle Cloud from 2027 (first non-Google hyperscale), counts Apple, Meta and Oracle as customers, and holds Anthropic rights for up to $150B of TPUs 2028–31. OpenAI is reportedly in talks to buy TPUs; Meta is in talks to spend billions. My opinion: TPUs stop being a science project the week Nvidia's CEO loses Washington leverage. Cloud buyers dual-source on politics as much as price.

AMD is the other beneficiary, and Monday priced it: Oppenheimer upgraded AMD to Outperform with a $600 target (+34% from Friday's $449.04 close), modeling 135% data-center revenue growth in 2027 on server CPU share gains, Helios rack-scale, annual Instinct cadence, open software progress, and the Anthropic partnership ramping Q1 2027 plus MI500 in H2 2027. 48 of 56 analysts are Buy/Strong Buy. AMD's EPYC even powers the host for Google's TPU 8tx — AMD gets paid whether the winner is open or proprietary.

Technical damage: the chip index broke its 200-day

Price confirms the story — NVDA broke down while semis broke out, the kind of divergence that does not fill quickly.

Levels from today's tape (quotes ~16:00 UTC / 11am ET Nov 9): NVDA $179.62 (-3.76%, open $175.81, high $179.92, low $170.24, Friday close $186.64). Premarket printed ~$176.76 (-5%). SOXX $146.50 (+7.3%), AMD $500.42 (+11.4%), AVGO $320.23 (+9.0%), GOOGL $293.51 (+3.1%), META $642.06 (+5.2%). A key AI chip index closed below its 200-day moving average for the first time since May 2025 — history says that break leads, not lags.

Map it: resistance $186–$188 (Friday close / today's high) then $205 (pre-scandal consolidation) then $210–$215; support $170.24 (today's low) then $164.27 (52-week low, Mar 30) then air to $150. NVDA is 26% off its $243.37 Oct 6 high. A megacap underperforming its sector by 11 points in one session does not "shake it off" — it needs a clearing event, and the next two are both hostile (earnings with forensic Q&A, then a Hill briefing).

For the non-chart reader: think of $170 as the floor of a crowded theater and $205 as the exit door. While the investigation and the Nov 18 hearing are unresolved, sellers stand closer to the door than buyers. Only a blowout Nov 17 guide that forces analysts to raise numbers moves the crowd back inside.

The trade: fade NVDA, own the share-takers

Do not bottom-fish the world's most important stock into a combined earnings + congressional interrogation — rent the rotation instead.

TradeEntryTargetInvalidationHorizon / Conviction
NVDA tactical underperform — short / sell calls / pair short vs AMD-AVGO$180–$188 (into strength; add on dead-cat toward $195)$155–$160 (below 52-week low, pre-April gap)Daily close above $205 — scandal priced, cover, no questions1–3 weeks to Nov 17 print + Nov 18 briefing / SPECULATIVE-HIGH

Sizing: half a normal short; outright short only for tactical traders. Long-only readers: no new NVDA longs until after Nov 18; holders trail at $170 — a daily close below cuts half. Pair version (my preference): long AMD $480–$500 / long AVGO $300–$315 against the NVDA short — you isolate governance alpha from market beta into Tuesday's CPI.

Bull / base / bear to Nov 18: Bull 20%: Nov 17 revenue >$110B + Q4 guide >$115B with clean Q&A and board clears air → squeeze to $210–$225. Base 50%: in-line beat ($108–$110B) but Q&A is all probe + H200 revocation headlines → $170–$190 chop, NVDA lags SOXX by 5–10 points. Bear 30%: soft guide or SEA-crackdown language or briefing expands → $155–$160, 52-week low breaks.

What would prove me wrong: (1) daily NVDA close above $205 on rising volume, (2) written confirmation H200 license intact by Nov 18, or (3) Nov 17 guide raised above $115B with Data Center >$100B — any one covers the short; two flips me long into year-end.

What to do Tuesday: CPI decides whether this stays idiosyncratic

October CPI prints Tuesday Nov 10 at 8:30am ET (September was 3.7%) — hot inflation turns an NVDA problem into a market problem.

If CPI runs hot, 10Y (4.86% Friday) pushes toward 5.10% and compresses every AI multiple ~10–12% — my Nov 6 book rule (halve all longs if S&P daily <7,300 or 10Y >5.10%) stays binding. If CPI cools ≤3.5%, shorts squeeze and the rotation broadens — which is exactly why the pair (short NVDA / long AMD-AVGO) beats a naked short. Other markers: CoreWeave Nov 10 AMC (16% bond yield is the AI-credit canary), Cisco + Applied Materials Nov 12 (enterprise demand truth), then NVDA Nov 17.

Downstream sentence for everyone else: **own GOOGL/AVGO over NVDA until Nov 18** — Google monetizes the same AI dollar without the hearing, Broadcom gets paid on every TPU regardless of who wins training. That is the diversification this audience asked for, delivered by events, not by quota.

Conclusion: my highest-conviction take

This is my opinion: markets forgive bad quarters faster than they forgive lying to Washington — and the board just told you it doesn't yet know which one this is. With earnings and a Hill briefing 24 hours apart, the risk is not that Nvidia's chips stop working; it is that every $315–$515 target still models a China option Washington is about to delete.

Trade the facts (probe + briefing + broken chart + TPU/AMD acceleration), not the legend (Huang always delivers). Fade NVDA $180–$188 toward $155–$160, kill above $205, and let AMD ($600 Oppenheimer target, 135% DC growth model) and AVGO carry the AI long. If Nov 17 guides above $115B and the license survives, I will cover and say so — including the loss.

Appendix

Contents: A. Data snapshot · B. Models & assumptions · C. Sources · D. Open calls

A. Data snapshot (intraday ~16:00 UTC Nov 9 2026; Nov 6 closes where noted)

NVDA $179.62 -3.76% (O $175.81 H $179.92 L $170.24 prev $186.64); AMD $500.42 +11.44% (prev $449.04); GOOGL $293.51 +3.07%; AVGO $320.23 +9.04%; MSFT $542.70 +2.28%; META $642.06 +5.23%; SOXX $146.50 +7.30%; SPY $751.69 +2.23%; QQQ $678.40 +3.40%; DIA $521.28 +1.77%; BTC $75,456 -0.70%; ETH $2,145.69 -4.16%; SOL $87.83 -1.99%; total crypto $2.538T -2.58%, BTC dom 59.62%; Crypto Fear & Greed 29 Fear (Nov 9), 25 Extreme Fear Nov 7. NVDA fundamentals: TTM P/E ~23.3x, EPS TTM $7.91, rev growth +83.4% y/y, gross margin 74.67%, beta 2.35, 52w $164.27–$243.37; AMD: TTM P/E ~96.7x, EPS $4.60, rev growth +41.9%, gross margin 53.72%, beta 2.65, 52w $188.22–$658.52. NVDA Q2 FY27 Aug 26: rev $96.221B +106% y/y, Data Center $89.0B +117% y/y, Q3 guide $108B ±2%. CPI Oct due Nov 10 8:30am ET; Sept CPI 3.7%.

B. Models & assumptions

No DCF — event + positioning trade. Load-bearing assumption: H200 China license revoked and SEA loophole tightened, removing ~5–8% grey-market Data Center buffer + China call option embedded in $400+ targets. If wrong (license intact in writing by Nov 18), NVDA recovers $205+ fast and short thesis fails — that is the stated kill. Second assumption: Nov 17 Q&A dominated by probe, capping multiple even on beat. If guide >$115B with DC >$100B, momentum overrides governance and I flip. Verified weekday: Nov 9 2026 is Monday; Nov 17 Tuesday earnings; Nov 18 Wednesday briefing; Nov 10 Tuesday CPI.

C. Sources

D. Open calls — consistency check

Consistent: DAILY-AMD-1105 HOLD (triggered $404.40, close $449.04 Nov 6, now $500.42 intraday — working, hold stop $360, target $580–$620); SPEC-COHR-1102 and DAILY-LRCX-1106 semis-long bias intact via AMD/AVGO leg; DAILY-ROT-1104, XLP, GLD hedges unaffected. No open NVDA long — no contradiction. New call added: NVDA tactical underperform $180–$188 → $155–$160, kill >$205, 1–3w. Book rule restated: halve all longs on S&P daily <7,300 or 10Y >5.10%.