The AI trade's real risk showed up in July, and it was not the drawdown

Goldman Sachs told clients that July produced one of the sharpest hedge fund de-grossing episodes of the past decade, as the AI rally lost momentum and managers unwound heavy positions [s1c1]. The figure that carries more information than the losses is this one: the bank's Hedge Fund VIP list of the most popular long positions suffered its worst one-month underperformance versus the S&P 500 in more than 20 years of history, according to Goldman [s1c2].

That is a statement about positioning, not about skill. A basket of consensus longs underperforming the index by a record margin means the crowd was in the same seats, and the seats emptied at the same time. Goldman says performance, leverage and key long positions all shifted considerably as the AI trade changed course [s1c3], and that hedge funds have been pulling back quickly from AI names [s1c4].

The crowding was built in plain sight. Goldman's data show tech stocks took 14 of the 20 spots among the fastest-growing hedge fund favourites [s1c5], which is 70 percent of the list [s1c6]. HFR reported that strong performance pushed total industry assets up by $409bn in the second quarter, to $5.6 trillion [s1c7], implying roughly a 7.9 percent rise from a base near $5.19 trillion [s1c8]. Shenan Dhanani, co-chief executive at Trium Capital, called it a possible "golden era" for the funds [s1c9]. In late July, Vincent Lin, co-head of Prime Insights and Analytics in Global Banking & Markets, described the wave of tech selling as a healthy correction amid high volatility rather than a loss of conviction in AI [s1c10].

The mechanics of the reversal are the point. Semiconductor, cloud and AI infrastructure names had absorbed heavy institutional demand during the rally, leaving many managers exposed to the same trades [s1c11]. When momentum weakened, the simultaneous rush to cut exposure amplified the losses [s1c12]. JPMorgan, in early August, said tech sell-offs wiped out 3 percent of hedge fund gains in July, and that managers were trapped in overcrowded tech positions that created a bottleneck preventing exits before profits vanished [s1c13][s1c14].

Two caveats, both worth taking seriously. JPMorgan notes that since 2018 hedge funds have tended to dump unprofitable positions in July and buy back in later, and suggested traders could return to tech by September [s1c15][s1c16]. And the P&L is not broken: Goldman says US equity long/short funds have returned 10 percent through mid-August [s1c17]. The read that follows from the source material is that July was a warning about valuations and positioning being stretched rather than a rejection of AI as a theme [s1c18].

For anyone running money or raising it, the practical conclusion is that concentration risk here is not idiosyncratic. It is shared, and shared risk clears through the same door.

What to watch: whether September re-grossing arrives on the seasonal schedule JPMorgan describes [s1c16], whether the VIP basket's underperformance versus the index persists past one month [s1c2], and whether allocator preference keeps rotating toward macro strategies, which HFR data show were the most sought-after style this year [s1c19].