What happened
Goldman Sachs' prime brokerage desk told clients Wednesday that hedge funds sold AI-exposed equities at the fastest pace in months, according to a note first reported by CryptoBriefing. The selling was broad. It covered semiconductor leaders, cloud infrastructure names, and the mid-tier software layer that rode the 2025 AI capex wave.
Goldman's flow data, which tracks net and gross exposure across its hedge fund book, showed net AI exposure dropping sharply from the elevated levels that defined positioning for most of the year. The desk framed the move as a de-risking rotation rather than an outright bearish call, with funds trimming the most crowded longs first. Short interest in second-tier AI names ticked up alongside the long unwind, a combination that typically precedes further downside in the weakest cohort.
Why it matters
AI equities and crypto majors have traded as correlated risk assets for most of 2025, both benefiting from the same pool of discretionary risk capital. When hedge funds crowd into one theme, the other usually catches a bid on the margins. When they unwind, the capital doesn't vanish.
It rotates. Goldman's note matters because the prime desk sees aggregate positioning before it shows up in 13F filings or quarterly letters, and a flow signal this clean is rare. The second reason it matters: AI stocks were the single most consensus long on the Street going into Q4.
A fast unwind of a consensus trade creates dry powder. Crypto majors, sitting near local highs on improving ETF flows, are the obvious candidate for a portion of that capital. The counterpoint is simpler and uglier.