Goldman Sachs: U.S. Momentum Stocks Plunge Over 20% in 5 Days, Steepest Since 2020
Miles Bennett
Goldman Sachs called July 7 an "ugly day" — its high-beta momentum factor fell 20%+ in five sessions, the sharpest sell-off since the 2020 Covid narrative switch, yet the trading desk says panic has not set in.
A 20% drop in five days — how rare is that?
Goldman's high-beta momentum factor (GSPRHIMO — a gauge tracking the hottest-running stocks as a group) plunged roughly 6% on the day and over 20% across five sessions.
This means → a basket that gained 57% in H1 gave back more than a third of those gains in under a week.
The last sell-off this fast was 2020, when markets violently rotated from "stay-at-home" to "reopening" plays. Goldman trader Guillaume Soria notes a key difference: that rotation had a stronger fundamental catalyst; this one does not.
In plain terms = it is falling as hard as 2020, but without a 2020-sized reason.
What lit the fuse?
The trigger came from the global AI supply chain. Samsung reported preliminary results: operating profit hit a record, but revenue slightly missed buy-side expectations. Its shares fell 9% in Seoul.
At the same time, DeepSeek announced plans to develop its own AI chips, intensifying concerns about the AI hardware pecking order and sparking a global AI-sector sell-off.
Leveraged semiconductor ETF SOXL saw volume 35% above its 30-day average and sat 50% below its all-time high set just 10 trading days earlier; DRAM-related volume ran 60% above average.
This reflects a market that is not pricing "AI is over" — it is pricing "the winner map in AI hardware may be redrawn."
Is there panic? Who is selling and who is buying?
Goldman's trading desk was explicit: no panic mode observed. De-risking remains orderly, driven by systematic and factor flows rather than panicked liquidation.
UBS's cash trading desk confirmed: selling pressure comes mainly from long-only accounts; hedge funds are operating tactically — covering shorts and adding selectively. This is consistent with "de-leveraging driven," not "bearish driven."
The standout structural signal: retail investors bought the dip. Goldman data show retail turned net buyer intraday and kept adding; by the close, net inflows hit the 90th percentile over a three-year lookback.
283 S&P 500 constituents finished higher on the day, showing the sell-off concentrated in momentum and AI names rather than a broad-market collapse.
Are positions still crowded? How much further can it fall?
Despite the 20% drawdown, the high-beta momentum factor and the broad AI index have not entered oversold territory; year-to-date gains still stand at 23%, and on a five-year basis positioning remains extremely crowded.
This means → without a fresh upside catalyst, further selling is possible — Goldman's trading desk said so in plain terms.
Historical reference: the factor's maximum historical drawdown is roughly twice the current decline; July has historically been one of the worst months for momentum, and this month is on track to be the worst July on record.
Soria's read: the correction is likely in its "late stages," but if negative news widens cracks into a chasm and market leadership rotates, the drawdown could expand significantly.
What is the key question for investors right now?
Goldman TMT specialist Peter Bartlett notes that the fundamental view on tech remains positive — a major reason the sell-off has not escalated into panic.
But buying in recent weeks has turned noticeably "selective" — a clear cool-down from the frenzied pace of May and early June.
In plain terms = the market has not rejected the AI story, but it has stopped buying every AI stock indiscriminately. The dividing line ahead: is this correction a breather after a position reset, or the beginning of a leadership rotation?
Content is for reference only, not financial advice.