UBS Group's trading desk believes that last week, U.S. stocks were essentially undergoing "deleveraging rather than de-risking"—AI and momentum sectors faced targeted selling, with memory chips plunging 16% in a single week, as capital shifted toward defensive and value-oriented sectors; however, overall risk exposure has not been systematically unwound. Hedge funds' aggregate exposure remains at a high 90th percentile, leaving further deleveraging risks intact. At the time, the "AI stock guru's margin call" was also driven by forced liquidation of crowded positions, rather than a fundamental‑driven, systemic sell-off.
Last week's U.S. stock market did not experience a broad-based sell-off, but rather a targeted position reset. In his latest weekly report, UBS Group trader Torsten Sippel noted that last week's market dynamics closely resembled the deleveraging episode triggered several months ago by "Situational Awareness," a well-known AI-driven quantitative fund—AI and momentum stocks remained under pressure, with capital rotating into defensive and less crowded sectors, yet overall risk exposure was not systematically abandoned.
AI and the momentum sector have become the epicenter of this round of market correction. Semiconductors, memory chips, and AI infrastructure have faced sustained selling pressure, with virtually no fresh capital stepping in to buy the dip. The memory sector posted a weekly decline of 16%, speculative growth stocks fell 12%, and the momentum long-short strategy dropped 8%. Meanwhile, consumer staples advanced 5%, and the value long-short strategy gained 6%, signaling a clear shift in market style—from growth and crowded momentum toward quality and value.
Hedge fund net exposure has fallen to the 30th percentile of its five-year historical range, yet total exposure remains above the 90th percentile. This structure suggests that, should profit-and-loss pressures persist, the market still faces the risk of further deleveraging. In his report, Sippel explicitly stated that the current market environment is "fragile rather than outright bearish," with risks being reallocated rather than abandoned.
Deleveraging, Not Risk Reduction: The Essence of This Round of Adjustment
Sippel characterized last week's market behavior as "de‑grossing rather than de‑risking."
From a capital‑flow perspective, high‑touch trading volume in the U.S. market has risen markedly, with the long‑only fund net‑long ratio at 36/64 and the hedge‑fund long‑short ratio at 42/50/8. Long positions continue to be reduced, yet new single‑stock short positions remain relatively limited—features that are inconsistent with a full‑scale risk‑clearing dynamic.
This pattern closely mirrors the deleveraging episode earlier this year, when the hedge fund Situational Awareness—founded by Leopold Aschenbrenner, the so‑called "AI stock guru" and former OpenAI researcher—experienced a similar dynamic: once again, it was crowded positions being forced to liquidate, rather than a fundamental‑driven, systemic sell‑off. In the current market phase, software, defensive sectors, and stocks with low positioning have absorbed some of the rotational capital midweek, while consumer discretionary, housing, healthcare, and select industrial sectors have also seen inflows. Meanwhile, small‑cap stocks and equal‑weight indices have outperformed the broader market.
AI financing logic under scrutiny: CoreWeave has become a key indicator.
CoreWeave has become an important benchmark for observing AI transactions. Sippel noted that, as the company's credit concerns continue to grow, market discussions around AI have shifted from the demand side to more fundamental questions: Can the economics of AI infrastructure development and its financing structure sustain the next phase of capital expenditure expansion?
Notably, both ASML and Taiwan Semiconductor reported strong underlying fundamentals, yet this has failed to allay the aforementioned concerns. This divergence is critical—it suggests that even as hardware demand remains robust, the market has developed an independent pricing rationale for the sustainability of AI capital expenditures.
Against this backdrop, the upcoming earnings reports of hyperscale cloud providers will face heightened scrutiny: investors will need to see that cloud revenue, order backlogs, and cash‑generation capabilities can reliably support the scale of AI‑related capital expenditures and financing needs.
Macroeconomic factors amplify, rather than trigger, deleveraging.
Sippel emphasized that, in this round of market adjustments, macroeconomic factors have acted as amplifiers rather than triggers.
Inflation and interest-rate shocks have accelerated systemic sell-offs, while the escalating situation in the Middle East is increasingly spilling over into the volatility market. On Friday, the VIX surged, and the 1‑month/3‑month VIX term structure inverted for the first time in six months, signaling that hedging demand has broadened from AI‑specific risks to broader market‑wide concerns.
The market's weakness on Friday afternoon is particularly noteworthy—the sell-off in the technology sector has exhibited a distinctly indiscriminate character. Sippel argues that this pattern aligns more closely with ETF redemptions, options expirations, and hedging roll‑overs than with incremental selling driven by individual stock fundamentals.
Earnings season has not yet become the dominant factor.
So far, earnings season has been broadly solid, though its impact on market trends has been limited. To date, 10.4% of the S&P 500's market capitalization has reported results, with third-quarter revenue expected to grow 12.1% and earnings per share up 26.5%. Actual earnings exceeded expectations by 4.1%, and 75% of companies beat estimates.
However, neither the banking sector's stellar earnings nor the positive AI‑related data have been enough to absorb the prevailing selling pressure; financial reports remain subordinate to positioning dynamics.
Over the next five trading days, 80 companies will report earnings, collectively accounting for 13.8% of the S&P 500's market capitalization, including Netflix, General Electric, Meta Platforms, Tesla, Procter & Gamble, and Western Digital. Sippel specifically noted that Meta released its earnings report a week earlier than usual, and hinted at this with a rhetorical question:
"Are they moving ahead to beat the deadline before the additional‑share issuance window closes?"
Overall, Sippel's assessment of the near-term market is that it is "fragile rather than outright bearish."
Risk is being reallocated rather than abandoned, but with the AI sector still lacking genuine bargain‑hunting capital, overall exposure remaining elevated, and the volatility curve recently under pressure, there is still room for further deleveraging. Moreover, the recent rise in volatility has sparked a fresh wave of skepticism about the sustainability of the capital markets' issuance window.
Editor/lambor