The AI rout is spreading from Silicon Valley to Asian factories and copper mines — short the AI chip leaders

Verdict: The bearish AI thesis is unproven against record fundamentals — stand aside

Our stance over the next 3 months is wait: the bearish idea has genuine sentiment confirmation, but the supplied fundamentals contradict the demand-deterioration story, and the idea's own trading rules are untriggered and could not be backtested. The strongest point for the thesis is cross-asset confirmation — copper and tin falling alongside AI equities (per the Bloomberg piece of November 3) plus roughly $569M of net open-market insider selling at NVDA in the provisional September 30 ownership reports. The strongest point against is that NVDA's latest quarter showed revenue up 17.9% sequentially to $96.2B with a gross margin near 75%, and AMD grew net income 66% to $2.3B, so the reported weakness is a working-capital and sentiment story, not demonstrated demand destruction. A second scope issue: historical performance of the rule set was not measured (NVDA raw and adjusted prices use different bases), so there is no evidence the trigger conditions have ever reliably worked. The verdict flips to actionable either way when NVDA resolves its price tension: a close above the 20-day EMA near $217 with RSI recovering above 45 weakens the short case, while a close below support at $208.85 strengthens it.

Trade now: wait — the momentum entry condition is not met

The plain answer today is: wait. The idea argues the AI rout that began in October has further to run in leading AI chip stocks — a bearish call on AMD, NVDA and SMCI. But the attached rule set is a long-only NVDA signal, and its entry conditions were not met as of the November 3 market state. NVDA closed at $208.43, and the 14-day RSI — a 0-100 momentum gauge — reads about 37, well below the 45 the entry requires. Three of four entry conditions are satisfied or near-satisfied (trend strength above 20, price below the 20-day EMA), but the momentum filter is decisively far, so the setup is in waiting mode, not execution mode. That is consistent with the thesis rather than contradicting it: a stock this weak has not repaired its momentum, and the long rule set would need conditions to improve before firing. Separate the two questions. The business thesis (bearish on AI chip equities) has some support in the supplied data: insiders were net open-market sellers across all three names in the September 30 provisional ownership reports, including roughly $569M of net selling in NVDA. But the fundamentals do not show a deteriorating business — NVDA's latest reported quarter (ended July 26, 2026) showed revenue of $96.2B, up 17.9% quarter over quarter, with a gross margin near 75% and $21.4B of free cash flow. AMD also grew revenue 12.5% to $11.5B with net income up 66% to $2.3B. This is a sentiment-and-positioning call, not a broken-fundamentals call, which is exactly why the bearish case is more fragile. Scope note, stated once: the historical performance of this rule set was not measured (NVDA raw and adjusted price data use different bases), and no robust parameter setup was established, so we cannot say the entry conditions have been a historically reliable long trigger. Combined with the mismatch between the bearish thesis and the long-only rule set, that lowers conviction in any mechanical entry — the pivotal thing to watch is whether NVDA's momentum and trend repair (RSI back above 45, price recovering above the 20-day EMA near $217) or breaks down further through support at $208.85.

Why the bear case has real support this time

Takeaway: the bear thesis is not contradicted by the sentiment evidence — metals selling off alongside AI equities, the loss of Nvidia's market-cap crown, and net insider selling at all three names support the idea that this is a broad de-rating rather than a one-day dip. Note first: this idea's rules could not be backtested — historical performance was not measured because NVDA raw and adjusted prices use different bases — so we judge only the fundamental and flow evidence. ### Sentiment and materials are confirming each other Per the Bloomberg wrap of November 2, the Nasdaq slide clouded the Asian open, and the November 3 Bloomberg piece reports copper and tin falling on the same AI-selloff tape. The thesis's mechanism — raw-material demand signaling what equity buyers haven't yet priced — is directionally consistent with two named, dated sources. Reuters' November 2 story frames the October AI dive as a test of the broader market's resilience, i.e., the drawdown is weeks old and unresolved, not a completed correction. ### Nvidia's cash flow momentum has already cracked NVDA's Q2 (period ended July 26, 2026) shows revenue up 17.9% sequentially to $96.2B, but operating cash flow fell 52.2% to $24.1B and free cash flow — cash left after capital spending — fell 56.0% to $21.4B. Debt-to-equity jumped 236% quarter over quarter to 0.15 and net margin compressed 9.4 points to 62.0%. A bear does not need an earnings miss: if working-capital strain and leverage build while the market is already de-rating AI on demand (per the cited metals signal), the multiple compresses on sentiment alone. ### Insider and institutional posture is uniformly negative Provisional ownership data for the period ending September 30, 2026 shows net open-market insider selling at all three companies: roughly $569.5M at NVDA, $185.5M at AMD, and $8.0M at SMCI. When the people closest to the numbers are net sellers into a downtrend, the bear case gains a confirmation signal the thesis itself did not even cite. ### SMCI shows how quickly AI-adjacent economics can invert SMCI posted a negative $6.7B operating cash flow quarter in Q3 (ended March 31, 2026) before recovering to +$747M in Q4, with gross margin collapsing from roughly 17.5% in Q4 FY2025 to as low as 6.3% mid-year before bouncing to 17.5%. Its debt-to-equity went from 0.03 in December 2025 to 0.28 by June 2026. This is the failure mechanism the thesis implies: if AI demand cools, the lowest-margin, most working-capital-intensive names in the chain absorb the damage first and hardest. ### Rotation has a concrete price signal The idea argues Nvidia ceded the 'most valuable company' spot to Apple — a visible reallocation toward perceived safety. Combined with the metals weakness, that is exactly the cross-asset confirmation the thesis requires to argue the selloff 'usually has further to run' rather than being a dip. For that to hold, the demand deterioration the copper and tin markets are pricing must eventually show up in NVDA and AMD order books.

What could break the short thesis

Takeaway: the fundamentals in the supplied data actively contradict the demand-deterioration story — revenue, margins, and cash generation are accelerating at AMD and NVDA, the reported weakness is concentrated in sentiment and one company (SMCI), and the strategy behind the trade was never validated. ### Demand signals in the filings point up, not down AMD grew revenue 12.5% sequentially to $11.5B in Q2 2026 (period ended…

Scores

• Conviction score breakdown: 45
• Thesis support: 45
• Trade readiness: 20
• Risk quality: 45
• Fundamentals trend: 70

Watch items

• NVDA — RSI (14)
• NVDA — Close vs 20-day EMA
• NVDA — Close vs nearest support
• NVDA — Insider net open-market activity
• AMD — Close vs nearest resistance
• SMCI — Close vs nearest support