Goldman Sachs: Positioning Correlation Drops to Historic Lows as Market Enters an Era of Single-Stock Divergence
N.R. Finch
Goldman's Prime Book data shows AI-linked positioning has undergone a material reset, with the rolling correlation between the S&P equal-weight and cap-weight indices falling to a record low of 79% — U.S. equities no longer move as one, and individual stock selection matters more than it has in decades.
How severe is this sell-off?
Global equities saw their largest net selling in three months over the past week. U.S. stocks were net-sold for a second straight week at the fastest pace since "Liberation Day."
U.S. information-technology net selling hit a ten-year-plus high. Semiconductors were the epicenter — chips and chip-equipment were net-sold for eight consecutive trading days, accounting for more than half of total U.S. tech outflows.
This means → capital is systematically exiting the most crowded positions along the AI hardware chain — not trimming at the margins, but resetting exposure wholesale.
Has this much selling killed the AI thesis?
Goldman strategist Lee Coppersmith frames this as a "position reset," not a fundamental rejection of the AI theme.
Global TMT gross and net exposure still sit at 31.6% and 40.0% of the global Prime Book — the 96th and 94th percentiles over one year, and the 99th percentile over five years on both measures.
In plain terms = institutions sold a lot, but their overall AI bet is still near historic highs. They shaved the most crowded layer, not the core position.
What is happening to the Magnificent Seven?
The Mag 7 have been net-sold for five consecutive weeks. Both gross and net exposure are near three-year lows, and large-cap tech trading has visibly contracted.
This reflects a breakdown of the market's "old leadership" — the handful of stocks that drove index gains for three years are no longer the default destination for capital.
The Russell 2000 / Nasdaq ratio is testing a breakout of its three-year downtrend. If confirmed, it would reinforce a shift from mega-cap AI-beta leadership to broader, stock-level leadership.
What does record-low correlation mean?
The one-year rolling correlation between the S&P 500 equal-weight and cap-weight indices has fallen to 79% — the lowest on record, against a historical average of 96%.
In plain terms = U.S. stocks used to rise and fall largely together, like passengers on one ship. Now each stock is its own boat, drifting on its own current.
This means → stock selection matters far more than it used to. The coming earnings season will bring higher single-stock volatility and wider performance dispersion than markets have seen in years.
What do earnings expectations and the hyperscaler checklist look like?
The S&P 500's trailing-twelve-month gain has been driven almost entirely by earnings growth, not multiple expansion. The Q2 consensus calls for 22% year-on-year EPS growth — the highest going into an earnings season since 2021.
AI infrastructure stocks are expected to contribute nearly 60% of S&P 500 EPS growth this quarter. Micron (MU) and Nvidia (NVDA) alone account for over 40%.
Goldman lists three conditions for hyperscaler leadership — the largest cloud providers such as AWS, Azure, and GCP — to reassert itself: capex growth decelerating, fuller AI-revenue monetization evidence, and a macro shift from acceleration to deceleration. The first condition is not yet met, but valuations have pulled back sharply, with forward P/E ratios near a ten-year low.
What does the leveraged-ETF boom mean for volatility?
U.S.-listed leveraged and inverse ETFs — funds that amplify daily index moves by a set multiple — have swelled to nearly $200 billion in total. SOXL, a 3× long semiconductor ETF, alone accounts for roughly $30 billion.
This means → these products must rebalance daily — buying into strength, selling into weakness — mechanically amplifying intraday swings in semis and AI names, even when long-term fundamentals are unchanged.
Goldman has launched three long-volatility strategies (semiconductors, AI, emerging markets) built around monetizing realized volatility, not betting on direction. Put simply = the trade is not that stocks go up or down, but that the swings themselves will be large.
Content is for reference only, not financial advice.