Record Sales but Still Heavy on Tech Stocks: Hedge Funds Compress Positions into "Low-Leverage Certainty"

According to Trading Intelligence APP, data from Goldman Sachs' Prime Brokerage Department show that hedge funds are exiting technology stocks at the fastest pace in at least a decade. In the five trading days ending October 15, hedge funds cut tech stock holdings at the largest rate since Goldman began tracking the data in 2016. Meanwhile, overall trading activity dropped at the fastest rate in nearly three months.

This wave of reduction comes as global tech stocks face significant sell-offs. With AI trades unwinding globally, the Nasdaq 100 index dropped 5.5% last week, marking its worst performance since April 2025. During the five-day period tracked by Goldman, the tech sector was also the worst-performing in the MSCI World Index.

Goldman Sachs data reveal that global equities as a whole saw net selling, reaching a scale equal to two standard deviations of capital flows over the past year. Information technology led the decline, with the largest reduction in holdings in two months; the semiconductor sector suffered the greatest loss in dollar terms.

Nevertheless, hedge funds still hold significant technology stock positions. According to a Goldman report, the tech sector's net allocation accounts for 23.5% of Goldman’s global prime brokerage book, remaining at the 90th percentile over the past five years. However, funds have become more cautious, with their net leverage ratio—a measure of the extent long positions exceed short positions—dropping to 48.6%, the lowest in a year.

The True Trigger: Credit fractures in the AI financing chain, not valuation itself

AI trades have been unwound worldwide, with a deeper catalyst being that the financing structure of AI infrastructure is exposing credit risks. Take CoreWeave as an example: Its 2032 bonds saw spreads break above 1000 basis points, its five-year CDS reached a record high of 1117 basis points, implying about a 63% chance of default within five years; the stock price plunged 15% in a single day, and bond prices fell below 80 cents on the dollar. Nscale’s IPO also faced discounts, with some investors suggesting a 37%-47% cut from the $30 billion valuation target.

This points to an issue more thorny than "overvaluation": there is a structural mismatch between the expenditure pace and revenue pace in AI infrastructure—costly outlays for GPUs, data centers, and power occur up front, while income must wait for operations, acceptance, and actual customer activation. The higher the financing cost, the weaker a project’s ability to endure delays and underutilization. When market pricing focus shifts from "how much is the compute demand" to "who pays for the high capital costs," the first to be sold off will inevitably be those with the most fragile balance sheets and the most sensitivity to financing conditions—semiconductors, new cloud service providers, and highly leveraged AI stocks—rather than cash-rich tech giants.

The Essence of Disagreement: How far has deleveraging gone?

The current core disagreement on Wall Street is not whether the long-term demand for AI exists, but whether most of the deleveraging has already taken place.

The cautious side (some teams at Goldman Sachs): From a one-year perspective, AI positions seem to have been mostly cleared; but switching to a three-year view, global AI stocks' net allocation is still at the 83rd percentile, with an AI long-short ratio of 3.27 at the 64th percentile. JPMorgan’s TPM position indicator is at around the 46th percentile—neutral, but well above the 19th percentile low at the end of July—suggesting room for further deleveraging remains. Feedback from Goldman’s Asia trading team shows that the “capitulation” in the Taiwan market during this sell-off was not as severe as in July, hinting at possible further selling pressure.

The positive side (Morgan Stanley, Bank of America): Morgan Stanley believes that the months-long valuation correction for most stocks is nearing completion. After a significant correction, valuations for tech giants are near historical cycle lows, while fundamentals such as core AI infrastructure chip prices, orders, and profit trends have not shown a similar deterioration, creating a buy-the-dip opportunity. Bank of America’s bullish logic is similar—the market has already discounted future financing pressures, yet spot prices are resilient and inference demand can support profit expansion for some time.

The key difference boils down to a judgment: whether credit tightening on the financing side is a substantive blow to the long-term AI narrative, or merely a forced cleansing of excessive leverage, while demand itself remains intact. Goldman’s own balanced view is reflected in analyst Mark Wilson’s admission that the "second derivative" of capital expenditure and profit growth will slow from current levels; however, he emphasizes that prices have largely priced in this expectation, and cites OpenAI’s CFO, who says there’s no sign of a slowdown on the compute demand side.

In short, the current exit is not "smart money" systematically abandoning tech stocks, but a forced compression of positions from "high-leverage AI exposure" to "low-leverage certainty holdings" under tighter financing conditions. The intensity of selling reflects the fragility of leverage, rather than a final verdict on the AI narrative.

Thus, a rebound in AI stocks should track bottoming signals for junk-grade AI assets and whether net leverage stabilizes and recovers. If the yearly low of 48.6% continues to fall, deleveraging is ongoing; if subsequent data show persistent stabilization, this points to diminishing selling pressure at the margin.

Moreover, whether the divergence between the bond market and the stock market narrows is also a key signal. During this correction, the 10-year US Treasury yield, despite higher-than-expected inflation, actually fell to its lowest in several weeks, with bonds tightening for the Fed even as they price in growth risks. If the spread on junk-grade AI bonds stops widening and yields peak and reverse, this may be a more reliable turning point signal than a rebound in stock prices.