(Bloomberg) -- Wall Street staged a tentative recovery from one of the wildest weeks of the AI boom, belying a historic rotation beneath the surface that briefly exposed growing cracks in some of the grand assumptions underpinning global markets.
Most Read from Bloomberg
The week brought a rout in technology shares, a fire sale by a prominent hedge fund and a surge in long-term borrowing costs, followed by a decisive snapback. By Friday's close, investors were left with a pressing question: whether they had witnessed a technical accident in an overcrowded trade, or a glimpse of a market becoming less forgiving of both Federal Reserve policymakers and the assumptions behind the AI investment boom.
At the center of the turmoil was Situational Awareness, the once high-flying hedge fund run by Leopold Aschenbrenner. Losses on leveraged AI bets triggered margin calls, forcing the firm to dump public equities into a falling market. The spiral finally broke after Ken Griffin's Citadel bought the bulk of its remaining public-equity portfolio in one swoop.
Yet Situational Awareness was an accelerant, not the original spark. AI stocks had already begun falling as investors questioned whether the vast sums being poured into the technology would pay off. Then a supposedly uneventful Fed meeting raised fresh doubts about new Chair Kevin Warsh's commitment to curbing inflation, pushing long-term Treasury yields higher. The liquidation helps explain why the selloff became so violent; it does not fully explain the pressures that set it in motion.
The moves were particularly stark through the lens of quant factors — characteristics such as value and volatility that underpin systematic portfolios and risk models across Wall Street. While cross-asset volatility gauges remained relatively subdued, many individual factors swung by the most in at least four years. Momentum, the strategy of betting on recent winners, has gone from its biggest four-day plunge since the vaccine-fueled rotation of 2020 to its biggest one-day gain also since 2020.
Laggards including value, quality and low volatility surged, tentatively overturning the hierarchy of investment styles that had prospered during the AI-led rally.
"Recent momentum weakness looks more like a rotation than a crash or even a correction," said Wai Lee, head of systematic equity research at Allspring Global Investments. "Now the market is rewarding those stocks which are showing better return on the investments, free cash flows and the like."
South Korea's chip-heavy Kospi surged a record 18.5% at one point on Friday, while semiconductor shares snapped higher around the world. Government bond yields finished the week higher, while the S&P 500 eked out a 1% gain.
There was plenty of evidence that leverage and positioning magnified the upheaval. The three sessions through Tuesday marked the sharpest degrossing by hedge funds since November 2022, Goldman Sachs's prime brokerage wrote in a midweek report. Momentum exposure among fundamental long-short funds fell to the 89th percentile of the past five years, while among systematic funds it dropped to the 34th percentile.
The speed of the subsequent rebound strengthened that case. Once a major forced seller was removed, investors rushed back into many of the trades that had suffered most as economic data remained resilient.
"A sharp reversal after such gains is common, and often it's not about any single catalyst as it is fear-of-missing-out giving way to vertigo," said Lewis Grant, senior portfolio manager at Federated Hermes. "Given the sharp drawdown in high momentum names and the valuation of many AI-winners now less stretched, I expect the worst of the rotation is behind us."
JPMorgan strategists reckon the rotation has further to run. In a note recommending the quality factor, the quant team led by Khuram Chaudhry cited worsening sentiment and peaking money supply growth. While momentum reversals have been frequent over the past year, "this month felt different," they wrote Friday.
The next threat lies in the bond market.
Warsh's press conference after the Fed held interest rates steady revived concerns about policy credibility and sticky inflation. The Treasury yield curve steepened sharply, with the gap between five- and 30-year yields widening by the most in a week since August 2025.
While such moves are often associated with stronger economic growth, the recent gains in long-end yields reflected higher inflation expectations and term premiums, according to Yin Luo, a quant analyst at Wolfe Research. The latter is the additional compensation investors demand for holding long-dated securities rather than repeatedly investing in shorter maturities. "Higher bond yields further amplify concerns about AI investment by raising discount rates and the cost of capital," Luo wrote in an email. "This reduces the present value of long-duration growth expectations and makes capital-intensive data-center projects more expensive to finance."
That creates a particularly awkward combination for the AI trade. The investment cycle increasingly requires companies to commit enormous sums today to data centers, power and networks in pursuit of profits that may arrive years from now. Higher borrowing costs make those projects more expensive at the same time that higher discount rates reduce what investors are willing to pay today for their hoped-for returns.
If yields retreat and AI earnings keep delivering, the episode may ultimately look like a spectacular clearing of leverage from an overcrowded trade. If borrowing costs remain elevated, the lesson of the week may prove more consequential: the valuations and crowded positions built on those assumptions have left parts of the market unusually sensitive when either is challenged. "This week was a wake-up call for anyone bullish the broad AI trade or anyone passively invested with concentrated US large-cap indexes," said Paisley Nardini, head of multi-asset solutions at Simplify Asset Management Inc. "Markets showed us that in order to navigate this new reward system we can't simply buy the index. Active management should provide better upside and risk management."
--With assistance from Julien Ponthus, Isabelle Lee and Greg Ritchie.
Most Read from Bloomberg Businessweek
©2026 Bloomberg L.P.