News & analysis
AI Stocks Shed Billions as Traders Cut Crowded Bets Before Midterms
Goldman Sachs says the AI complex failed to hold an early rebound, with high-momentum names down about 7% as money rotated into homebuilders and value.
Key points
- Goldman Sachs described a continuing de-grossing cycle, with crowded long positions being cut while shorts squeezed higher.
- Homebuilders, alternatives and value stocks outperformed as energy prices collapsed and flows moved toward discretionary and retail.
- Apple beat on the quarter but its December-quarter gross margin guidance of 45-46% reflected higher memory costs, and services growth guidance of 8-9% missed expectations.
The AI trade failed to hold an early rebound, and the selling spread across the complex as investors cut crowded positions heading into the U.S. midterm elections, Goldman Sachs said. Gen AI stocks fell about 5%, AI healthcare names lost 4%, optical stocks dropped 7.5% and neoclouds — smaller cloud providers built around AI computing capacity — slid 6%. High-momentum stocks, the shares that had run hardest, dropped roughly 7%.
The pattern Goldman Sachs described is a rotation rather than a uniform selloff. Homebuilders, alternatives and value stocks outperformed while the AI complex fell, which the firm tied to two forces: fading support for AI shares and a sharp collapse in energy prices that redirected money toward consumer discretionary and retail names. In a de-grossing cycle, investors reduce total borrowed exposure — cutting both winning and losing positions — rather than simply turning bearish. Goldman said crowded longs were being cut while shorts continued to squeeze higher, meaning some of the worst-performing bets were actually rising as they were closed out.
Apple added to the cautious tone. The company beat on its quarter but disappointed on the outlook, and the shares fell about 3% after hours. Its December-quarter gross margin guidance of 45-46% reflected higher memory costs, while services growth guidance of 8-9% missed expectations. The detail matters beyond Apple: AI demand is supporting memory pricing, which in turn tightens component supply and pressures margins at hardware companies that must buy that memory. That is a cost channel running through the same AI buildout that has driven the sector's gains.
The timing sharpens the stakes. Investors are reducing risk into the midterms, when control of Congress and the policy agenda are decided, and the AI complex is where positioning had been most concentrated. Goldman's account suggests the pressure is coming from portfolio mechanics — leverage, crowding and momentum — as much as from any single piece of bad news. The Apple guidance is the clearest fundamental input in the packet, and it cuts in an unusual direction: stronger AI demand is feeding higher memory costs that hurt margins elsewhere in technology.
The inference that follows is conditional. If de-grossing is the dominant force, the selling should ease once leverage is worked down and the crowded longs are cleared, because the underlying businesses have not all deteriorated. If instead the memory-cost squeeze proves persistent, margin pressure at hardware makers becomes a genuine earnings problem rather than a positioning one, and the rotation out of AI would have more room to run. The distinction between those two mechanisms is what separates a temporary unwind from a change in the trade's fundamentals.
