AI stocks suffer their worst slide since the DeepSeek moment: Goldman's momentum pair wipes out its yearly gains as SOX underperforms by 13% in nine days, a 20-year record, pressing the 200-day line

AI momentum trading is living through its worst battering since the DeepSeek shock: Goldman's high-beta momentum pair's year-to-date gain — nearly 70% at the June peak — has been entirely surrendered; the Philadelphia Semiconductor Index (SOX) has underperformed the S&P 500 by about 13% over the nine sessions since October 12, the worst such window in more than two decades, which Goldman's trading desk calls "the worst performance since the DeepSeek day"^[1].

The full data picture: gains zeroed, SOX at the 200-day

Goldman's statistics sketch the rout: the unconstrained momentum strategy is suffering its worst drawdown since the global financial crisis (second only to the pandemic shock); SOX has been pushed to its 200-day moving average — the first touch since June 2025 — and whether that line holds is the first marker for whether the correction deepens^[1]. The selling spread to portfolio construction: per Goldman Prime, hedge funds' fundamental long-short net leverage fell 3.1 points this month to 49.4%, the 1st percentile of the past year; the global AI long-short ratio dropped to the 6th percentile; UBS data show systematic strategies' short leverage at the 99th percentile of five years — but long leverage is also at the 94th, meaning this is not a one-way turn bearish but a position reshuffle under high leverage on both sides^[1].

Style rotation: low-vol and quality take the baton

As high-beta momentum ebbs, the low-volatility factor is back in favor and the earnings-quality factor is strong too (exposure at five-year highs) — Goldman's quality index has risen about 20% over three months. The migration path is clear: from chasing elasticity toward defensiveness and earnings certainty^[1]. BofA's Hartnett points risk aversion's persistence into mid-November (cooling AI trade plus midterm pressure); Bloomberg's Ball names the switching risk line: rising oil is replacing credit worries as the main threat to the S&P 500's structural rotation^[2].

The tape's mirror: energy highs, Apple overtakes

Friday's US session offered a microcosm of the rotation: the Dow fell 1.23% to 52,215.16, the S&P 500 0.52% to 7,623.68 and the Nasdaq a slim 0.21%; consumer staples led the decline at -1.83% (Coca-Cola -2.76%, P&G lower) while energy rose 2.51% to record highs and tech added 0.35%^[2]. The most symbolic moment: Apple's market value briefly exceeded Nvidia's — in a cycle led down by AI hardware, the neglected "old tech" standard-bearer completed the overtake, itself a footnote to momentum reversal^[2]. The oil tape supplies the macro pressure: Brent settled near $117, WTI rose 4.2% on the day to $103.63, and the 10-year Treasury yield climbed back 5 basis points to 5.00%, giving up most of the prior day's rally^[2].

What to watch

First, SOX's battle at the 200-day — a technical break would trigger more forced de-risking by systematic strategies. Second, the Fed's language under oil-driven inflation next week. Third, follow-up bond deals and earnings wording from AI heavyweights like Meta — the capex narrative lit this fire. Fourth, policy noise ahead of the November midterms, which together with the AI narrative decides whether momentum factors stabilize before year-end^[1][2].

Historical coordinates: three endings for a momentum collapse

Placed in historical coordinates, this momentum collapse has three possible endings. The first is a "mechanical unwind of a crowded trade": like the reversal after the GameStop affair in early 2021 — once positioning clears, the factor stabilizes within weeks and fundamentals resume the wheel, provided no macro shock relays the baton. The second is a "narrative inflection": like March 2000's internet top, where the momentum collapse is merely chapter one of a valuation reset, followed by a long grind of earnings downgrades — if AI capex financing keeps tightening (this week's credit spreads are the evidence), that scenario's probability is rising. The third is a "policy-rescue turn": like Q4 2018, when the Fed's dovish pivot triggered momentum stabilization — making next week's FOMC more important than usual. The current evidence mix (SOX at but not through its 200-day, spreads wider but absolute levels still low, earnings season overall beating) sits between the first and second; the key variables separating them are the giants' bond-funding costs and the pace of AI revenue delivery.

An overlooked detail: long leverage is also extreme

Most readings focus on short leverage at the 99th percentile as the "bearish signal," but UBS's data show long leverage at the 94th — the combination actually means "extreme disagreement," not one-way bearishness. With both sides positioned at historical leverage, any price break gets amplified: an upside break squeezes shorts, a downside break forces long liquidation. In other words, this positioning structure has locked in a floor for volatility over the next two weeks, while direction depends on catalysts — earnings aftershocks, oil, or the Fed, any one ready to light the fuse.