AI money is rotating from chips into platforms — long Meta against the chip basket

Verdict: a plausible rotation thesis with no live signal — wait for the pullback to confirm

The idea's rotation narrative — chips cracking while platforms hold firm — has a genuine anchor: Meta rallied over 3% on its new Wonder model family per the October 12 MarketWatch report even as the Broadcom CEO's 'gamblers' remark kicked off a two-day chip selloff per the October 13 Bloomberg coverage. But the platform leg of the trade looks shaky in the latest quarter: Meta's free cash flow fell 86.8% quarter-over-quarter to $1.7B and operating margin dropped nearly 10 points to 30.9%, while NVDA's free cash flow fell 56% to $21.4B and its debt-to-equity jumped from 0.04 to 0.15. Provisional ownership filings for the September 30, 2026 period show net open-market insider selling of roughly $569.5M across 127 holders at NVDA and $57.4M across 108 holders at META — filing data as of that reporting period, not a current tally — which cuts directly against the rotation thesis. The trade itself is a watch-list setup: the entry rules produced zero triggers across 1,235 daily bars over 60, 24, and 12 months, and the author has requested a bounded expanded search while no robust parameter setup has been established. NVDA and SMH are close to their pullback conditions (price within $3.72 and $4.43 of their EMA triggers, ADX about 1.1 and 0.5 points short), but META needs roughly a 10% pullback before its sequence can even start. The right move is to watch, not act, and reassess when either NVDA or SMH prints a close below its EMA followed by an RSI cross back above 45 with trend strength above 18.

Trade now

Nothing triggers today. The strategy is a pullback-reclaim long on META, NVDA, and SMH: price below the 50-day EMA, the 14-day RSI at or below 45 and then crossing back above it, and trend strength (ADX, period 14) above 18. Only one of those conditions is broadly in place — RSI is at or below 45 on NVDA (41.3) and SMH (36.1) — while META sits at 61.5, well above the threshold and more than $72 above its own EMA trigger of $670.04. NVDA and SMH are the live watch names. NVDA closed at $226.27 versus an EMA trigger of $222.55 — about $3.72 away — and its ADX of 16.9 needs roughly 1.1 points to clear 18. SMH closed at $590.99 versus a trigger of $586.56, about $4.43 away, with ADX at 17.5. META, after its rally of over 3% under the Wonder model family per the idea, is the furthest from its own setup: it needs a pullback of roughly 10% to get under the $670 EMA before a reclaim could fire. When an entry does fire, the exits are mechanical: a fixed take-profit at +5.0% against a fixed stop at −2.5%, plus a resistance-level target and a support-level stop per symbol. That works out to roughly a 2:1 reward-to-risk. Position risk per trade is sized at about 2.5% of equity with a 25% cap on any single position. What "wait" means concretely: do not pre-position. Watch for an NVDA close below $222.55 (or an SMH close below $586.56) followed by an RSI cross back above 45 with ADX above 18 — the entries are sequential, not same-day, so patience is the trade right now. A search for a more permissive parameter setup exceeded its evaluation budget, so no adjusted thresholds are recommended; the published levels above are the ones to act on.

The platforms still hold the profit pools — and the entry waits for the washout

## The rotation thesis has a real anchor in the numbers The idea argues that the AI trade is rotating from hardware suppliers into platforms with real products and profits, and the fundamental evidence for that framing is genuine. Per the MarketWatch report of October 12, Meta rallied more than 3% on its new Wonder model family, returning to the AI leaderboard — while the Broadcom CEO's 'gamblers' remark triggered a two-day chip selloff, per the October 13 Bloomberg coverage of a deepening chipmaker slump. Meta's most recent full fiscal year shows what 'real products and profits' looks like: fiscal 2025 revenue of $201.0B (up from $164.5B in 2024), full-year operating margin above 41%, and $46.1B in free cash flow. On a peer basis, Meta's Q4 2025 operating margin sat in the 97th percentile of its Communication Services group, and revenue growth of 23.8% year-over-year landed in the 86th percentile. That is not a company whose franchise depends on a single capex cycle. ## NVDA's profitability is extreme by any standard The platform leg of the divergence is even more pronounced. NVIDIA closed its fiscal year ended January 25, 2026 with $215.9B in revenue — up 73.2% year-over-year — and a 65.0% operating margin that ranks in the 99th percentile of its sector. Net income for that fiscal year reached $120.1B, and full-year free cash flow was $96.7B. Even in the most recent reported quarter (ended July 26, 2026), revenue grew to $96.2B from $81.6B the prior quarter and net income rose to $59.7B, with the operating margin actually ticking up to 66.2%. The thesis's claim that profit pools in the AI theme are concentrated in businesses with pricing power is hard to dispute on these figures; the question the selloff poses is about multiple, not franchise. ## The pullback structure is what makes the trade disciplined The entry design matches the thesis's own logic. The rules wait for a genuine washout — price below the 50-day moving average, RSI at or below 45, then a momentum recovery with trend strength above 18 — across META, NVDA, and SMH, before going long. That is a 'buy the dip after capitulation, not during it' structure, which is exactly the right posture when a single executive's remark can move the whole sector. A 2.5% stop against a 5% take-profit, a 60-bar time exit, and a 25% maximum position size bound the risk of being wrong on timing. Important honesty note: the rules did not trigger anywhere…

Scores

• Conviction score breakdown: 45
• Thesis support: 55
• Trade readiness: 45
• Risk quality: 55
• Trigger proximity: 30
• Fundamentals trend: 40

Watch items

• NVDA — Close vs EMA (50)
• NVDA — RSI (14)
• NVDA — ADX (14)
• SMH — Close vs EMA (50)
• SMH — ADX (14)
• META — Close vs EMA (50)
• META — RSI (14)
• NVDA — Insider net open-market value (2026-09-30 provisional filing)
• META — Insider net open-market value (2026-09-30 provisional filing)