AI hype losing steam as tech giants stumble — short the chip rally

Verdict: a watch-list short with the thesis half right — wait for the RSI gate

The idea's core claim — that AI-hardware sentiment is cracking after Marvell's 8% drop and Alphabet's Gemini delays — has real fuel in Super Micro's cash burn: FY2026 revenue of $39.1B grew 77.8% year over year, yet free cash flow was negative $7.0B, the worst among 791 IT peers, with a gross margin of 10.8% in just the 8.9th percentile. But the thesis stumbles on its own anchor: Broadcom's quarter ended May 3, 2026 showed revenue up 14.9% sequentially to $22.2B, free cash flow of $10.3B (the 99.2nd peer percentile), a 42.0% net margin, and debt-to-equity down 10.6% — shorting the sector through its strongest cash generator means shorting improving fundamentals. The insider-selling disclosures cited (about $283.3M net at Broadcom across 23 reporters and $8.0M at Super Micro) are from the period ending June 30, 2026 with the deadline already passed, so they are months-old filings, not a fresh signal. Execution is the bigger problem: the three-condition entry never triggered once across 1,229 daily bars in five years, and no robust parameter setup was established because the sensitivity evaluation ran out of time. Right now only Broadcom is close, with an RSI of 46.6 versus the below-45 gate and price already below the $373.31 Bollinger level; Super Micro (59.8) and SOXX (63.0) remain far from triggering. The verdict is a clean wait: watch daily RSIs, act only when the gate fires.

Trade now: three setups, none triggered — here is what is missing

This is a waiting setup, not an active signal. The strategy shorts a sell-the-rally breakdown: a daily close at or below the rank-1 resistance level, at or above the lower Bollinger Band (20, 2), with the 14-day RSI below 45. On live data as of the latest daily close, none of the three tickers has all three conditions aligned — that is the setup working as designed, not a warning sign. Where each leg stands: **SOXX** closed at $532, above its lower Bollinger level of $523.46, but its RSI of 63.0 sits 18.0 points above the below-45 requirement — the furthest from trigger of the three. **SMCI** closed at $38.93, above its Bollinger floor of $37.09, with an RSI of 59.8, about 14.8 points too high. **AVGO** closed at $364.38 with an RSI of 46.6, only 1.6 points above the threshold, and price already below the $373.31 Bollinger level — the closest to firing, but still not there. Risk is defined mechanically: each position carries a fixed stop at a 2.4% adverse move and a take-profit at a 4.7% favorable move, roughly a 2:1 reward-to-risk, plus a hard exit if price closes back above the rank-2 resistance level and a 45-bar time stop. Position risk is capped at about 2.37% per trade with a 25% maximum position size. Until an entry prints, "wait" means exactly that: no position, watch the RSI readings daily. One caveat on the research: no robust parameter setup was established — the parameter-sensitivity evaluation ran out of its time budget — so the published thresholds are the ones you see here, taken as-is.

The Bull Case for the Short: Sentiment Has Turned While the Cash Flows Tell a Shakier Story Than Headlines Suggest

The idea argues that the AI spending boom is cracking, and the cited tape supports the mood shift. Per the Yahoo Finance piece from July 16, 2026, Marvell dropped 8% on AI capex slowdown fears and dragged Broadcom, AMD, and Intel down with it; the day before, Dell fell 14% with HPE and Super Micro sliding as AI hardware names gave back gains. MarketWatch reported Alphabet's stock falling on Gemini AI model delays. A dominant hyperscaler stumbling on its own AI roadmap is exactly the kind of catalyst the thesis says converts "buy at any price" into "sell the rally." The ownership picture adds weight. In the most recent filed period ending June 30, 2026, both tickers show net open-market insider selling: about $283.3M net sold across 23 institutional reporters in Broadcom, and $8M net sold in Super Micro. Insiders with the best view of the order book choosing to sell into the peak does not confirm a top, but it is the posture you would expect if the thesis is right. The most decision-useful evidence is in Super Micro's quality of growth. Yes, FY2026 revenue (period ended June 30, 2026) hit $39.1B, up 77.8% year over year and in the 86th percentile of Information Technology peers — but free cash flow was negative $7.0B, the absolute worst among 791 IT peers, and gross margin of 10.8% sits in just the 8.9th percentile. That is a business buying its own growth with working capital. When a sector's parabolic story rests on names that convert little of it to cash, the short-side thesis has a concrete vulnerability to trade against. One scoping note: the compiled short rules did not trigger across 1,229 evaluated daily bars over the last 60 months — the three-condition entry (close at or below the rank-1 resistance level, at or above the lower Bollinger Band, with 14-day RSI below 45) never aligned. This is a watch-list setup, not an active signal. The research author…

Scores

• Conviction score breakdown: 47
• Thesis support: 55
• Trade readiness: 40
• Risk quality: 55
• Trigger proximity: 45
• Fundamentals trend: 40

Watch items

• AVGO — RSI (14)
• SOXX — RSI (14)
• SMCI — RSI (14)
• SOXX — Close vs lower Bollinger Band (20, 2)
• SMCI — Free cash flow (quarterly)
• AVGO — Insider net open-market activity (13F cycle)
• AVGO — Quarterly revenue momentum