AI Trade Flips From ‘Can’t Lose’ to ‘Pick Your Poison’
TREE NEWS reports: The AI investment narrative is fracturing under its own contradictions. What was once a one-way bet on hyperscaler capital spending has morphed into a lose-lose proposition: accelerate capex and free cash flow suffers; decelerate and the demand story unravels. Either path points to selling in AI-linked assets.
Over the past week, a cascade of negative catalysts across the AI ecosystem collided with the heavy long positioning built up during September’s “agent mania,” triggering a violent forced unwind. In the single worst session, 47 of the 124 US equity long-short pairs monitored by Nomura’s cross-asset strategy team led by Charlie McElligott swung by more than three standard deviations, with six exceeding five standard deviations—extreme outliers by any measure.
The shock did not stay confined to AI darlings. Through crowded-position chain liquidations, it spread to Treasuries, currencies, and commodities, setting off a cross-asset systematic deleveraging.
The Capex Prisoner’s Dilemma
McElligott describes the situation as “a choice between poisons.”
- Path one: capex keeps accelerating. As Meta’s latest earnings hinted, the market would fixate on the relentless drain on free cash flow and question financing sustainability. This path carries non-linear risk—requiring ever-larger debt issuance at wider spreads—and pressures the AI trade under an “unsustainable” narrative.
- Path two: capex materially decelerates. That would signal the market’s priced-in “best case”—spanning AI demand, technology adoption, profitability, and end-user productivity gains—is breaking down, delivering an even more brutal blow to AI assets.
The trap was set in September, when the “agent concept” rally prompted rapid rebuilding of long positions in AI core names, accompanied by sizable speculative call buying. That set the stage for a stampede exit.
Stop-Loss Triggers and Cross-Market Deleveraging
Crowded positioning is the core reason for the outsized volatility. Nomura’s trading desk data shows 124 US equity long-short pairs hit stop-losses in the worst session, with 47 swinging beyond ±3 standard deviations, 26 beyond ±4, and six beyond ±5.
McElligott traces the root to 2026’s highly concentrated “high Sharpe” trades, where carry and trend-following strategies resonated for so long that they created a rare broad-spectrum winner effect. Now, at year-end, fund managers facing drawdowns prefer capital preservation over chasing rebounds—”de-risk, close positions” dominates.
The AI narrative reset is also accelerating systematic trend-strategy adjustments, particularly in crowded Treasury and short-rate shorts. Over the past two weeks, CTA trend strategies have covered roughly $68 billion in Treasury futures shorts and $38 billion in short-rate futures shorts, yet signals remain net short—meaning potential forced covering remains substantial if the bond rally extends.
McElligott notes that with the short-term window largely covered, the bond short squeeze is pausing. The 2-week and 1-month models have flipped to “long,” but the 3-month model needs a much larger bond rally to turn.
Macro Resonance: Stock-Bond Correlation Regime Shift
The AI narrative reset has triggered a structural shift in stock-bond correlations. The negative stock-bond correlation—the classic hedge—has re-emerged as “growth scare” credibility rises. Meanwhile, in equities, the relationship has flipped: Nasdaq and semiconductors had shown the unusual pattern of higher volatility on up days; now negative spot-vol correlation is back, meaning bigger swings on down days. This is most pronounced in Nasdaq and SMH (semiconductor ETF), directly reflecting repricing of AI capex sustainability.
Notably, index-level VIX has been relatively tame, which McElligott attributes to a “reverse dispersion” effect—tech longs getting hammered while rate-sensitive sectors and other shorts rebound sharply, offsetting each other.
Large Options Signals: Contrarians Still Betting on AI Rebound
Despite the gloom, options markets show notable contrarian positioning in core AI names:
- CoreWeave: ~$69 million premium paid, ~$170 million notional exposure
- Nebius Group: ~$25.7 million premium, ~$80 million notional
- Micron: ~$20 million premium, ~$85 million notional
These are out-of-the-money calls expiring March-April 2025, signaling some funds maintain directional bets on a medium-term AI rebound.
For stabilization, McElligott points to volatility risk premium (VRP) sellers—particularly 0DTE gamma sellers—as the key stabilizer. Whether they resume their historical “sell into vol spikes” behavior will determine if volatility can converge again.
Key Takeaways for Investors
- The AI trade is no longer a one-way bet. Both faster and slower capex are now bearish catalysts, removing the “heads I win, tails you lose” dynamic that fueled the rally.
- Positioning, not fundamentals, drove the selloff. Crowded longs and stop-loss triggers amplified moves; expect more volatility around hyperscaler earnings (Google, Amazon, Microsoft) and the Fed meeting next week.
- Watch the bond-stock correlation. The return of negative correlation means Treasuries are hedging again—a regime shift with portfolio construction implications.
- Contrarian options flows suggest some see a rebound, but the path depends on volatility sellers stepping back in.
