Seventeen tech giants already sit −12% to −70% off their highs while the S&P 500 paints a calm tape over the wreckage. The easy money — the names already down 60% — is gone. The thesis is to short what still has room to fall, where the market hasn't finished pricing the deterioration.
In 2007 the index made new highs while the foundation rotted. A handful of people read the foundation instead of the index — and sized their conviction to the gap between price and reality.
Michael Burry didn't short the housing market because it had already fallen. He shorted it because the price hadn't caught up to the truth yet — the deterioration was visible in the data months before it was visible in the tape. That is the exact discipline here. The headline index is the 2007 home-price chart: serene. The names underneath are the subprime tranches: already cracking.
"Don't short the crash. Short the denial. The money is made in the distance between what the price says and what the cash flows already know."
Four structural pressures — each independently verified — that make a crowded, levered, semiconductor-concentrated market fragile to an unwind.
Record leverage is dry tinder: when crowded longs get a margin call, they sell what they can, not what they want to. Concentration is the accelerant: with semis at a record 18.8% of the index, a chip-led drawdown drags the whole tape. And the Fed posture is hikes-not-cuts — removing the put the bulls are counting on.
Already broken, but not yet capitulated. Each has a live catalyst pushing the next leg down. Drawdowns below are computed from verified 52-week highs as of 25–26 Jun 2026.
| Target | From high | Room left | Thesis & catalyst |
|---|---|---|---|
| ORCLOracle · $148.53 | −57% | −15 to −25% | AI capex burning cash; AI narrative migrating to Asian memory/semis. High-multiple cloud story losing the bid. |
| CRMSalesforce · $158.37 | −43%CORRECTED · brief said −57% | −15 to −25% | SaaS commoditization by AI agents. It has more room to fall than the original brief implied — strengthens the short, not weakens it. |
| PLTRPalantir · $112.22 | −46% | −15 to −20% | Valuation still extreme on any sober multiple; gov-spending tightening removes the growth premium. Sentiment, not fundamentals, holding the floor. |
| NOWServiceNow · $98.34 | −53% | −10 to −15% | Workflow automation being commoditized; post-Q1 margin concerns. Near 52-wk low — confirm break before pressing. |
What we deliberately avoid: COIN is already ~−60% from its July-2025 high of $444.75. The thesis already played out — the asymmetry is gone. Shorting a 60%-down name is paying full price for yesterday's insight.
This is not a market falling — it is capital migrating. Out of US software multiples, into Asian hardware that actually makes the AI buildout's irreplaceable component: memory.
The original brief estimated SK Hynix at +348%. The verified figure is over +1,000% in twelve months — the rotation is more violent than the thesis assumed. The structural shift isn't a risk to the short book; it's the engine behind it. The US software premium is the source of capital funding the Asian hardware bid.
Three forward paths with assigned probabilities. The book is sized so the base case profits, the bear case pays, and the bull case is survivable.
Margin-call cascade meets semi concentration. Grantham's tail. Tier A names cut another 15–30%; index −25%+. Puts pay multiples.
No crash, but the hidden bear keeps bleeding US software multiples as AI agents commoditize SaaS. Tier A grinds to targets. The core thesis.
Rate cuts + a fresh AI catalyst re-inflate multiples. Shorts squeeze. This is what the stop-loss and position caps exist to survive.
Conviction without structure is a casino. The structure is what turns a thesis into a survivable book.
The short book's job is to lose small and often, then win big and rarely. These are the hard, pre-committed rules.
Per the Autonomy Honesty Standard — every claim verified against live data, every correction surfaced, nothing flattered.
E5 Enclave Incorporated