Goldman Trading Desk: AI Data Center Long Crowding Rises to Six-Month High
N.R. Finch
Goldman's trading desk data shows hedge funds have pushed long crowding in AI data centers and metals to a six-month peak, while natural-gas longs collapsed and Mag7 flipped to short-crowded for the first time — a pre-summer positioning split that is redrawing the risk map for the second half.
What does "crowding" actually measure?
Goldman uses a crowding score to gauge how concentrated hedge-fund positioning is in a given sector — the higher the score, the more funds are piled on the same side.
This means → when a crowd of funds is on one side, a market reversal can trigger stampede-style unwinds.
In plain terms = crowding doesn't tell you direction; it tells you "if everyone is wrong, how narrow is the exit."
AI data centers and metals — are longs still piling in?
AI data-center long crowding rose from 0.75 to 0.82, near a six-month high; short crowding fell to −0.36, the range low.
This means → longs are adding, shorts are retreating — both sides moving in the same direction signals consensus-long conviction is strengthening.
Metals saw an even sharper move: long crowding jumped from 0.55 to 0.73, short crowding plunged to −0.77. Goldman flagged metals as one of the most net-long-crowded sectors right now.
Why did the natural-gas long suddenly collapse?
Natural-gas long crowding dropped from 0.81 to 0.32 within a single month, hitting the six-month range low — consensus longs are unwinding fast.
Goldman points to the current geopolitical landscape as a background driver.
This reflects a directional reversal in hedge-fund sentiment on natural gas — from one of the most crowded longs to near-neutral in just weeks.
Mag7 flipped to short-crowded — what happened to Big Tech?
Mag7's crowding score moved from −0.23 to 0.13, the 93rd percentile — flipping from "uncrowded" to short-crowded.
In plain terms = hedge funds are collectively rebuilding short positions in Big Tech — not a forecast that prices must fall, but a signal that the number of shorts has risen sharply.
Media & entertainment short crowding also surged from −1.56 to −0.71 (84th percentile), with a clear acceleration in short-building.
Tech hardware — which way is money rotating?
Tech hardware & equipment long crowding rose from 0.09 to 0.30 (93rd percentile); short crowding fell from 0.67 to 0.32 (six-month low).
This means → capital is rotating clearly from "Big Tech platforms" toward "tech hardware & equipment" — long hardware, short platforms.
This reflects hedge funds' judgment: the AI beneficiary is migrating from the software and platform layer to the infrastructure and hardware layer.
What does this positioning map mean for the second half?
The current hedge-fund picture in one line: AI infrastructure + metals concentrated long, Big Tech + media concentrated short.
In plain terms = fund managers are betting "the shovel-sellers are more certain than the shovel-users" — but when everyone is on the same side, any surprise shift in macro variables could trigger a violent position rebalance.
Goldman flags that whether this structure can hold as macro conditions evolve is a key variable to watch in the second half.
Content is for reference only, not financial advice.