Mega-cap tech is getting punished for AI spending, but secondary chip deals are booming

Verdict: The divergence is real, but the trade isn't armed yet — wait

The thesis has real substance: Broadcom's quarter ended May 3, 2026 grew revenue 14.9% sequentially to $22.19B with free cash flow up 28.1% to $10.26B, and Palantir's June-quarter free cash flow more than doubled to $2.09B — the picks-and-shovels demand story is visible in the numbers, and the July 1 reporting on the OpenAI and sovereign-AI deals supports it. The single strongest point against is timing and freshness of evidence: the most recent 12 months of the backtest lost 1.3% with a 30.8% win rate, insiders were net open-market sellers across all three names in the June 30, 2026 period, and no robust parameter setup was established, so the setup's best returns came from a window that is over. Right now there is nothing to enter — AVGO's 20-day rate of change is -17.7% and PLTR's long still needs a negative MAGS reading to arm. Verdict: wait, keep both legs on watch with the -5% stop and +10% target pre-set. What would flip it: PLTR arming on a negative MAGS 20-day rate-of-change print, or AVGO recovering its rate of change above zero, ideally after a supportive early-September Broadcom report.

Trade now: AVGO fails its entry gate; PLTR waits on MAGS

There is nothing to enter today. The strategy goes long AVGO and PLTR only when each stock's 20-day rate of change is above zero **while** the Magnificent Seven ETF's 20-day rate of change is below zero. AVGO fails on its own leg: its 20-day rate of change is **-17.7%** with the stock at **$346.08**, well short of the required at-or-above-zero reading — that leg is not close to triggering. PLTR passes its stock-side condition with a 20-day rate of change of **+16.2%** at **$182.53**, but the entry still requires the MAGS 20-day rate of change to turn negative, and that reading is not in the live feed right now, so the position remains on hold rather than live. Once an entry does trigger, the risk math is fixed by the rule set: a stop at **-5%** and a take-profit at **+10%**, a 2:1 reward-to-risk ratio, with positions capped at 25% of the book and sized to roughly 2.5% risk per name. Exits also fire if either stock closes below its 20-day EMA — currently about **$374.6** for AVGO (already breached at $346.08, so any AVGO position taken now would immediately face that exit pressure) and **$172.4** for PLTR, about 5.5% below the last close. The completed 24-month backtest supports the setup as designed: **+97.3%** return over 18 trades with a **38.9%** win rate and a **23.1%** maximum drawdown. The low win rate paired with the high return is exactly what a 2:1 payoff structure looks like — losers are cut fast, winners run to the profit target. Note the exits in that test were filled on daily bars, not intrabar, so treat drawdown and win rate as coarse. Concretely, "wait" means: hold off on AVGO until its 20-day rate of change recovers above zero from -17.7%, keep PLTR on the watchlist pending a negative MAGS 20-day rate-of-change reading, and pre-set the -5% stop and +10% target for the day either leg goes live.

The picks-and-shovels divergence is showing up in the numbers, not just the headlines

The idea argues that while mega-cap tech is being punished for AI spending, the companies actually building AI infrastructure are locking in revenue — and the latest fundamentals back that split. Broadcom's quarter ended May 3, 2026 grew revenue 14.9% sequentially to $22.19B, lifted net income 26.7% to $9.31B, and pushed operating margin to 48.6% from 44.3% one quarter earlier. Free cash flow jumped 28.1% to $10.26B — enough to cover a $2.54 trailing-12-month dividend that has grown roughly 10% a year. Palantir tells the same story on a smaller canvas: revenue up 18.6% to $1.94B, net income up 22.0% to $1.06B, and free cash flow more than doubling to $2.09B with an 84.7% gross margin. These are the 'guaranteed revenue' mechanics the thesis describes. The cited news corroborates the demand side. Per the July 1 Yahoo Finance piece, Wall Street sees Broadcom gaining up to 40% after the OpenAI deal, and the same day Palantir jumped 9% on a NVIDIA sovereign-AI partnership. Meanwhile the June 30 Yahoo report notes the Magnificent Seven shed $2.3 trillion in June as AI spending came under scrutiny — exactly the divergence the strategy is designed to exploit: go long the infrastructure names when their 20-day momentum is positive while the mega-cap ETF's momentum is negative. The completed backtest gives the setup a realized track record. Over the 24-month window on daily bars, the AVGO leg produced a 97.3% total return across 18 trades — albeit with a 38.9% win rate and a 23.1% maximum drawdown, a pattern typical of trend-following: many small stops, a few large winners. The equity curve shows the strategy riding the late-2024 and mid-2025 AI runs to gains of 70–100% before giving some back in drawdowns. Note that exits were filled on daily bars rather than intrabar data, so stop and take-profit fills are approximate and the drawdown figure should be treated as coarse. The rule design also matches the risk thesis. The 5% stop and 10% take-profit brackets cap single-trade damage while the 20-day EMA exit pulls the strategy out when momentum breaks — meaning you are structurally protected from holding the picks-and-shovels names into a broad AI unwind. Broadcom's balance sheet reinforces the durability case: debt-to-equity fell 10.6% to 0.71 in the latest quarter, continuing a multi-year…

Scores

• Conviction score breakdown: 58
• Thesis support: 75
• Trade readiness: 35
• Risk quality: 45
• Backtest evidence: 55
• Fundamentals trend: 80

Watch items

• PLTR — MAGS 20-day rate of change
• AVGO — 20-day rate of change
• PLTR — Close vs 20-day EMA
• AVGO — Close vs 20-day EMA
• PLTR — Insider net open-market activity (June 2026 filing)
• AVGO — Dividend ex-date
• AVGO — ROC (20) above 0
• AVGO — ROC (20) below 0
• AVGO — Price below EMA (20)
• NVDA — ROC (20) above 0
• NVDA — ROC (20) below 0