Chinese AI firms are closing the gap with Nvidia's chips, which chips away at the story propping up the most crowded trade in the market. If Nvidia's moat narrows, the expensive US AI complex is exposed while cheap Chinese tech offers the same AI theme at

Verdict: A coherent China-vs-Nvidia rotation — but the entry machinery has never fired

The rotation thesis is more interesting than its tradability. The strongest evidence for it is on the China leg: KWEB's trailing dividend per share more than doubled to $2.096, up 104% year over year, concrete proof that Chinese tech platforms are returning real cash to shareholders, while Nvidia's own filings (quarter ended July 26, 2026) show free cash flow down 56% to $21.4B and net margin down 13.2% to 62.0% even as revenue grew 17.9% sequentially to $96.2B. The strongest point against is the smart-money tape: the ownership filing for the period ended June 30, 2026 shows roughly $375.3M of net open-market insider selling across 23 NVDA reporters — not current positioning, but a signal that insiders do not behave as if the moat were eroding as fast as the thesis implies. Mechanically, nothing is actionable: the entry rules produced zero trades across 1,220 daily bars in every window evaluated, the price conditions sit tens of dollars below every last close, and no robust parameter setup was established after the sensitivity search exceeded its time budget. The verdict flips if KWEB's 12-day rate of change turns positive while price recaptures its 20-day Bollinger midpoint near $26.77, or if the December ex-dividend window confirms that the doubled payout holds. Until then, this is a watch-list idea worth following, not a trade worth taking. **Conviction breakdown** - Thesis support: 65 — cash-flow deceleration at Nvidia plus KWEB's 104% payout growth give the rotation a numbers-backed core. - Trade readiness: 25 — zero historical entries, unreachable price conditions, and no recommended parameter setup. - Risk quality: 45 — FXI's dividend fell 29% year over year and near-zero measured correlations between two China ETFs may not survive a real stress episode. - Trigger proximity: 15 — every entry condition is far; only the Bollinger midpoint condition is near (KWEB $0.72 below). - Fundamentals trend: 45 — Nvidia's revenue still grew 17.9% sequentially, but free cash flow and net margin both fell sharply last quarter, and issuer fundamentals for FXI and KWEB are not yet available to verify the 'cheap China' claim.

Trade now

Nothing is actionable today. This is a watch-list idea, not a buy. The strategy's entry rules were checked against live daily bars and none fired, so the right move is to wait. The trade is a relative-value rotation — long Chinese tech (KWEB, FXI) against the crowded Nvidia-led AI complex — and it is designed to enter only on a pullback with momentum turning up, not at current prices. The gap to entry is wide on every leg. The price conditions require KWEB, FXI, or NVDA to close at or below roughly $1.01 — tens of dollars below every last close (KWEB $26.05, FXI $35.88, NVDA $230.36) — so as written those conditions are effectively out of reach. The momentum condition needs the 12-day rate of change to cross above zero; it currently reads -2.7% on KWEB, +0.6% on FXI, and +5.9% on NVDA. One supportive condition is already in place: all three tickers closed above their 20-day Bollinger midpoint (KWEB is the nearest miss, $0.72 below it). If an entry does fire, the risk framework is already defined: a hard stop at a 2.4% loss and a take-profit at a 4.8% gain — a 2-to-1 reward-to-risk — with a fallback exit at the first resistance level above entry and a time stop after 75 trading days. Position size is capped at 25% of the book, with about 2.4% of capital risked per position. One honest caveat on setup quality: after the rules failed to trigger across 1,220 evaluated daily bars, the author requested a bounded parameter search, but the sensitivity evaluation ran past its time budget and no robust nearby-parameter setup was established. Concretely, "wait" means: no orders until either the price conditions are recompiled to realistic levels, or KWEB's 12-day rate of change turns positive while price recaptures its Bollinger midpoint near $26.77.

China Tech At A Discount: The Value Leg Of The Rotation Has Room To Run

The idea is a relative-value rotation: stay long the AI theme, but in the cheap corner. The data supports the 'cheap corner' half more strongly than the headline suggests. KWEB's trailing dividend per share has more than doubled — $2.096 over the last twelve months versus $1.025 the prior year, a 104% increase in annual payout — a concrete signal that Chinese tech platforms are returning materially more cash to shareholders, not just promising growth. FXI adds a broad-China exposure layer, though its own payout trajectory is rockier. On the other side of the rotation, the idea argues the US AI complex is crowded and expensive. Nvidia's latest quarter (period ended 2026-07-26) still shows scale — revenue of $96.2B, up 17.9% from the prior quarter, with gross margin holding at 75.0% and net income of $59.7B — but the marginal numbers are softening exactly where a moat-erosion thesis would show up first: net margin fell 13.2% quarter-over-quarter to 62.0%, and free cash flow dropped 56% to $21.4B from $48.6B. Operating margin rose only 1.0% to 66.2%. The dominant franchise is still elite — operating margin sits at the 98.9th percentile of 854 Information Technology peers, and revenue growth at the 85.9th percentile — but the direction of change is consistent with the Reuters piece (2026-09-07) on China's AI dragons pressuring Nvidia's moat. The balance-sheet data adds nuance the…

Scores

• Conviction score breakdown: 39
• Thesis support: 65
• Trade readiness: 25
• Risk quality: 45
• Trigger proximity: 15
• Fundamentals trend: 45

Watch items

• KWEB — ROC (12)
• KWEB — Close vs Bollinger (20) midpoint
• KWEB — Next ex-dividend date
• FXI — ROC (12)
• FXI — Dividend annual growth
• NVDA — Insider net open-market activity
• NVDA — Free cash flow