Short interest reaches record high in U.S. equities amid AI-driven bull market

• S3 data shows that the short interest in U.S. equities has risen to record levels, with increased unwinding of positions in technology stocks.
• This is more like a high-level hedge and does not mean the bull market has reversed; the degree of earnings realization will determine the direction of volatility.
• Underlying assets: SPX, NDX, XLK, SMH, NVDA, Mag 7.

While the S&P 500 remains at elevated levels, short positions in U.S. equities and hedge fund reductions in technology stocks are both increasing, prompting the market to reassess the safety margin for AI-related trades.

This set of signals has unsettled investors because it contradicts the dominant trend of the past two years: while AI narratives have driven index gains, institutions are buying more insurance on this trade. The question isn’t just whether U.S. stocks have peaked, but whether bad news will be amplified after prices have priced in substantial optimism.

Let’s clarify the concept first. The short interest ratio is the proportion of shares sold short relative to the total tradable shares. It may reflect a direct bet on price declines, or it may simply serve as a hedge. Funds may still hold long positions but reduce downside risk through short selling, options, or reducing exposure.

Therefore, a record long position does not automatically mean institutions are fully bearish. A more accurate statement is that U.S. stocks are still trading on the long-term potential of AI, but institutions are beginning to reprice short-term volatility.

Rising prices and increasing short positions may seem contradictory, but they often occur simultaneously—especially when valuations are high, positions are crowded, and earnings season is approaching. Funds often maintain their long positions while simultaneously increasing protective hedges.

According to media reports citing data from S3 Partners, the short interest in S&P 500 components accounts for approximately 3.79% of freely tradable shares, the highest level since S3 began tracking in 2010. For Russell 3000 components, this ratio is around 6.3%, also at a record high.

These figures cannot be simply added together with data from other sources. Different institutions use varying scopes of measurement, and exchanges typically disclose short interest numbers rather than a consistent set of proportional metrics. The methodology disclosed by the New York Stock Exchange in July shows that, as of June 30, 2026, the total short interest for the NYSE Group increased compared to the prior period, indicating only that the size of short positions has grown.

Clients of Goldman Sachs Prime Brokerage also reported a similar trend. According to Reuters on July 6, U.S. hedge funds posted their fourth consecutive week of net selling in technology hardware and semiconductors, with the information technology sector also marking its fourth straight week as the most net-sold U.S. sector.

The key point here is not a specific percentage, but that large holders are reducing their net tech exposure. The market has not seen a collective retreat—overall prices are still supported by retail buying, corporate buybacks, and trend-driven capital—but the cushion during rallies has grown thicker.

Institutions are not concerned that AI lacks value, but rather that the current price has already priced in much of the future returns.

Over the past few years, the primary explanation for the rise in U.S. stocks has been AI. Cloud providers and tech giants increased capital expenditures, benefiting NVIDIA and the semiconductor supply chain through spillover orders, as the market believed these investments would ultimately translate into higher revenue, profit margins, and productivity gains.

For investors, capital expenditure is not the story itself, but an investment that must generate returns. If AI infrastructure spending continues to rise without a corresponding acceleration in end-user revenue, enterprise payments, and profit contribution, valuations will come under pressure first.

Semiconductors are most prone to becoming amplifiers of volatility. Positioned at the front end of the AI supply chain, they react fastest to orders and expectations, and their valuations are most sensitive. Whenever earnings reports show slowing order growth, margin pressure, or concerns over customer concentration, the market typically first reprices semiconductors before transmitting the impact to the Nasdaq and S&P 500.

Geopolitical risks serve as external catalysts. Conflicts in the Middle East, energy prices, and supply chain uncertainties may not alter the long-term demand for AI, but they can affect how much multiple investors are willing to pay for high-valuation assets. The biggest concern in an overvalued market is not a single negative event, but rather the overcrowding of positions when such events occur.

This explains why the increase in short positions resembles a rise in insurance premiums: institutions may not believe the AI bubble is bursting, but they prefer not to be overly exposed to earnings reports and external risks.

Morgan Stanley's recent strategic framework perfectly illustrates the core of this divergence: while the index may still have upside potential in the medium term, technology and semiconductor sectors need to digest recent gains in the short term.

On July 14, Morgan Stanley strategist Mike Wilson mentioned on the official podcast that semiconductors may pull back, with continued volatility and corrections expected before the next bull market rally. On July 15, a Morgan Stanley Wealth Management article noted that its Global Investment Committee anticipates the S&P 500 could rise to between 8,000 and 8,300 over the next year, while recommending taking some profits in the semiconductor sector.

This is not simply a matter of being bullish or bearish, but rather a dual perspective common in elevated markets. In the long term, if earnings continue to be revised upward and AI investments generate real revenue, the index could continue to rise. In the short term, if valuation expansion outpaces earnings growth, a pullback could occur.

For investors, do not interpret position signals as one-way predictions. An increase in short positions could lead to a short squeeze after positive earnings reports, forcing short and hedging positions to cover, thereby pushing prices higher. Alternatively, it could amplify declines when negative news emerges.

What determines the direction is not how many bears there are, but whether, when the catalyst materializes, the market finds that its previous valuation assumptions were too conservative or too optimistic.

The upcoming earnings reports from tech giants and semiconductor companies will serve as a stress test for AI trading. The market won’t be satisfied with just “strong demand”—it will look for alignment between cloud revenue, AI orders, gross margins, and return on capital expenditures.

If the earnings report shows cloud revenue continuing to accelerate, AI orders remaining strong, and gross margins holding steady, short positions could become fuel for a rally. Those who are short or hedging may need to cover their positions, and momentum buyers will reaffirm AI as the key theme.

If financial reports only demonstrate continued expansion of capital expenditures without proving corresponding returns, the market will reassess paying high valuations for future growth. At that point, short positions won’t be the cause of the decline, but they will act as amplifiers of volatility.

The more reasonable assessment is not that the bull market has ended, nor that short sellers are certainly being squeezed. The U.S. stock market is entering a more discerning phase. The AI narrative remains valid, but valuations require continued earnings performance to justify them. Semiconductors remain the core theme and the first to face risk reevaluation.