Nvidia chips are delayed but Meta is making their own — rotate into software

Verdict: a real divergence, but the trade still hasn't earned its entry

The divergence thesis has real fundamental support: Nvidia's quarter ended 2026-07-26 showed free cash flow down about 56% to $21.4B and operating cash flow down 52% to $24.1B even as revenue rose 17.9% to $96.2B, while Meta still generated $31.9B in operating cash flow and grew revenue 8% to $60.8B — the hardware-strain-versus-software-resilience profile the idea argues for. But the mechanical entry never fired in 1,235 daily bars across 60 months (or in the 24- and 12-month checks), and the parameter review returned no robust setup, so there is no historical evidence the pair actually trades as described. Ownership filings through the period ended 2026-06-30 cut both ways: about $565.4M of net open-market insider selling at NVDA versus about $33.9M at META supports the rotation, but insiders selling at Meta too weakens the 'safe software bet' framing. Meta's own numbers carry the biggest fundamental risk — net income down about 41% to $15.8B and free cash flow down 87% to $1.7B in the latest quarter — so further margin compression would undercut the long leg. Today the correct action is wait: META's side is nearly armed (RSI 86.8, above its 10-day Donchian band of about $607), while NVDA's 14-day RSI of 52.0 must fall to at or below 45 alongside a fresh 10-day low (Donchian level about $222, nearest support about $220.5) before anything is valid.

Trade now: the long-META/short-NVDA setup is a watch-list trade, not a live one

Do nothing today. The idea's entry requires META making a 10-day high and NVDA making a 10-day low at the same time, filtered by momentum: META's 14-day RSI above 55 and NVDA's 14-day RSI below 45. Half the picture is already in place — META trades at about $652.8 with RSI at 86.8, comfortably above the 55 threshold, and its 10-day Donchian band sits at about $607, so META is near the top of its short-term range. But NVDA is the blocker: it closed at about $224.2 with RSI at 52.0, which needs to fall to at or below 45 — roughly 7 points of momentum — before the short leg's condition can even be evaluated. NVDA also sits just 4.9% below its range high, the opposite of the 10-day low the setup wants. The compiled rules have not opened a position on real daily bars over the evaluated windows, so this is a waiting setup by design, not a signal we missed. On risk, the plan is already defined in the rule set: a hard stop at a 7% maximum loss on the combined spread (with an earlier de-risk trigger at about a 2.5% unrealized loss), a 10% take-profit on the spread, and a floor on the META leg at the 38.2% Fibonacci retracement level if it breaks down. Position sizing caps each leg at 25% of the book, which fits the equal-dollar long/short construction. 'Wait' means concretely: check NVDA's 14-day RSI and its 10-day low each close, alongside META's 10-day high. If NVDA breaks to a fresh 10-day low while META holds above its own 10-day high and RSI stays above 55, the pair arms — enter the long META/short NVDA spread the same day. If META's RSI falls back below 55 or NVDA's RSI rebounds above 45 before the pair aligns, reset the watch; no entry is valid on a partial match. The thesis behind the setup — per the idea, Nvidia's chip delays pushing rotation toward Meta's custom-silicon efficiency — is still intact in the filings: NVDA's latest quarterly free cash flow fell about 56% to $21.4B and debt-to-equity jumped to about 14%, while META grew revenue nearly 8% to $60.8B even as its net margin compressed to about 26%.

The divergence the thesis needs is visible in the filings

The rotation thesis has a real fundamental anchor. Meta's latest full fiscal year (period ending 2025-12-31) shows $201.0B of revenue, up 166.1% year over year, an 82.0% gross margin, a 41.4% operating margin, and $46.1B of free cash flow — placing it in the 95th percentile or better among Communication Services peers on growth, operating margin, and free cash flow. That is the 'software layer wins when hardware wobbles' profile the idea is arguing for. Per the Yahoo Finance piece (July 7, 2026), Nvidia delayed its next AI system to 2028 over a circuit board problem, and per the MarketWatch piece (July 9, 2026), Meta's rebound came precisely from agentic AI coding and custom chips easing spending fears — the catalyst pair the thesis hinges on. Meta's most recent fundamentals also argue it can absorb an aggressive AI buildout. Its own quarterly trajectory shows the strain of capex — free cash flow fell 86.8% sequentially to $1.7B in the quarter ended 2026-06-30, and debt-to-equity rose to 0.32 — but revenue still grew 8.0% to $60.8B in that quarter, and operating cash flow held at $31.9B. A company generating that much cash while funding its own silicon can pursue the custom-chip route without depending on any single vendor's roadmap — exactly the point of the cited custom-chip milestone. On the short side, Nvidia's numbers remain enormous but the marginal direction is softening. The quarter ended 2026-07-26 showed revenue up 17.9% sequentially to $96.2B, but free cash flow fell 56.0% to $21.4B, operating cash flow dropped 52.2% to $24.1B, and net margin compressed from 71.5% to 62.0%. Debt-to-equity jumped from 0.04 to 0.14 in one quarter. If hardware delays stall the…

Scores

• Conviction score breakdown: 40
• Thesis support: 60
• Trade readiness: 35
• Risk quality: 40
• Trigger proximity: 20
• Fundamentals trend: 45

Watch items

• NVDA — RSI (14)
• NVDA — 10-day Donchian low
• NVDA — Nearest resistance
• META — RSI (14)
• META — 10-day Donchian high
• META — Stop level on combined spread
• META — Net income trend
• NVDA — Insider ownership filings