The U.S. equity market is not rising solely because investors have become more optimistic about earnings, AI and economic growth. A growing part of that optimism is being expressed through margin borrowing, leveraged exchange-traded products, derivatives and hedge-fund financing.
Leverage can extend a rally by allowing investors to increase their market exposure faster than their underlying capital. The danger emerges when the direction changes. At that point, the same mechanism that amplified gains can turn an otherwise manageable correction into forced deleveraging.
The rally is becoming more dependent on borrowed money
Debit balances in U.S. customer securities margin accounts reached a record $1.502 trillion in June 2026, up approximately 49% from $1.008 trillion a year earlier. The acceleration has been particularly sharp, with margin debt rising from $1.304 trillion in April to $1.416 trillion in May and then to $1.502 trillion in June.
A nominal record alone is not evidence of a bubble, especially as the size of the equity market has also increased. The speed of the rise, however, shows how aggressively investors have expanded financed exposure during the latest stage of the rally.
Margin debt, however, is only one part of the picture. Assets in U.S. domiciled leveraged exchange-traded products nearly doubled to around $200 billion during the first quarter, while primary dealers’ equity-repo exposure surpassed a record $220 billion.
Barclays estimated the equity-financing market and hedge funds’ gross equity exposure at approximately $10 trillion. Options demand has created another layer of synthetic exposure, while the Federal Reserve reported that hedge-fund gross notional leverage remained near the highest level recorded in its comprehensive dataset and was disproportionately concentrated among the largest funds.
These figures measure different things and must not be added together, but collectively they show that equity demand has become increasingly dependent on financing and derivative exposure.
The Federal Reserve continues to describe the banking system as sound and resilient, while broker-dealer leverage remains low relative to its historical range. The more immediate vulnerability is therefore a market-structure and forced-selling risk rather than clear evidence of an imminent banking crisis.
High leverage is not an immediate sell signal
Leverage is inherently procyclical. Rising equity prices increase the value of collateral, which creates additional borrowing capacity. That borrowed capital supports further purchases, pushing prices and collateral values higher again. As long as investment returns exceed financing costs and volatility remains contained, leverage can reinforce the prevailing trend rather than interrupt it.
This is why high leverage should not be treated as a mechanical market-timing indicator. Barclays noted that rising equity-financing costs often accompany periods of strong investor confidence or euphoria and are not necessarily bearish by themselves. The problem begins when prices flatten or fall while financing remains expensive. At that point, the expected return no longer compensates investors for the cost and risk of maintaining leveraged exposure, encouraging or forcing them to reduce positions.
The market can therefore continue rising despite the warning signs. Positive earnings revisions, persistent AI-related capital expenditure, resilient economic growth and further inflows could attract even more financed demand. Leverage may help produce an additional extension of the rally before becoming the mechanism that accelerates its reversal.
The BofA survey contains two different messages
Bank of America’s
July Global Fund Manager Survey adds an important positioning dimension to the leverage data. Its findings suggest that the market has become increasingly vulnerable to a positioning reset.
Average cash allocations fell from 4.1% to an “uber-low” 3.6%, triggering BofA’s contrarian sell signal. Managers raised their U.S. equity allocations to the highest overweight since December 2024. Long global semiconductors remained the market’s most crowded trade, cited by 82% of respondents, while no surveyed manager reported being short on technology.
AI bubble risk was identified as the market’s largest tail risk by 45% of respondents. When positioning becomes this one-sided, there are fewer uncommitted buyers and a substantially larger pool of potential sellers if the narrative weakens. 
The same survey also supports a constructive longer-term interpretation. A record 54% of respondents expected a “no landing” outcome for the global economy, while only 2% expected a hard landing. In addition, 61% believed hyperscalers were unlikely to reduce capital expenditure during 2026, compared with 28% who expected spending cuts. Investors therefore recognised that AI positioning was crowded without necessarily concluding that the underlying investment cycle had ended. 
These conclusions are not contradictory. The AI investment cycle can remain fundamentally intact while the securities associated with it become temporarily overowned, overvalued or overleveraged. A correction could remove excessive positioning without ending the structural trend. If earnings, capital expenditure and final demand remain credible, lower valuations after deleveraging could provide the foundation for the next stage of the rally.
The fundamental support remains substantial. Earnings for semiconductor and semiconductor-equipment companies were expected to rise 133% year over year in the second quarter. The group now represents roughly 18% of the index’s market capitalisation, yet it was expected to generate approximately 44% of the S&P 500’s year-over-year earnings growth. This does not prove that current valuations are justified, but it shows that the rally is supported by genuine profit growth rather than leverage alone. 
Leverage still needs a catalyst
A heavily leveraged market does not collapse merely because leverage is elevated. It requires a catalyst that raises financing costs, reduces the value of collateral or undermines the fundamental narrative supporting the most concentrated positions.
The Federal Reserve is one possible trigger. The next FOMC meeting is scheduled for July 28–29, 2026. The July BofA survey showed that 83% of respondents did not expect the Fed to raise interest rates before the November midterm elections. A materially more hawkish message would therefore challenge both consensus expectations and the economics of leveraged exposure. 
Oil represents a second potential catalyst. moved above $100 per barrel on July 23 as renewed conflict in the Middle East raised concerns about supply disruption and inflation. A temporary price spike would not necessarily alter monetary policy. A sustained move above $100, however, could revive inflation concerns, push inflation expectations higher and keep policy rates elevated for longer.
The transmission mechanism would be particularly damaging for leveraged growth exposure:
Higher would attack the rally through two channels simultaneously. They would increase the direct cost of financing leveraged positions while raising the discount rate applied to the distant cash flows of high-growth technology companies. The stocks with the greatest sensitivity to changes in rates are also among those where positioning and leverage appear most concentrated.
A stronger US dollar would add another layer of pressure by tightening global financial conditions, increasing the effective burden of dollar-denominated financing and reducing the translated value of foreign earnings generated by U.S. multinational companies. The dollar was already receiving support from renewed U.S.–Iran tensions and the inflationary implications of higher oil prices.
These macro catalysts matter because they can undermine several assumptions supporting current valuations, including cheap financing, stable inflation and lower future rates. A leveraged market is most vulnerable when the same shock weakens expected returns while simultaneously increasing the discount rate applied to them.
However, macro risks are not the only potential trigger. A separate fundamental tail risk comes from the rapid improvement of Chinese open-weight AI models. Moonshot AI’s Kimi K3, a 2.8-trillion-parameter system, demonstrated that Chinese developers are narrowing the capability gap with leading U.S. models while offering systems that can be deployed at materially lower cost. This is unlikely to eliminate demand for computing infrastructure. Cheaper models may instead accelerate adoption and inference demand.
The bigger risk is lower returns on the capital invested in AI infrastructure. Cheaper and increasingly interchangeable Chinese models could weaken proprietary pricing power and make it harder for hyperscalers to justify their enormous spending. That would challenge the earnings assumptions behind the market’s most crowded semiconductor positions.
What a leveraged unwind would look like
A disorderly adjustment would probably begin inside the most crowded part of the market rather than across the entire index. An initial decline would raise volatility and weaken collateral values, potentially triggering tighter financing conditions, margin calls, hedge-fund deleveraging and leveraged-product rebalancing. These flows could then turn a concentrated correction into broader selling.
July’s semiconductor pullback offered a limited preview. The (green line) fell just over 20% from its late-June high. Yet the broader S&P 500 (red line) remained comparatively resilient, suggesting that the episode was still a concentrated positioning unwind rather than a market-wide rejection of equity risk.
The constructive scenario therefore remains credible. If earnings expectations remain resilient, hyperscaler spending continues and financing conditions stabilise, the correction could remove excess positioning without breaking the underlying AI cycle. An improvement in market breadth would be particularly important, as it would reduce the rally’s dependence on leveraged semiconductor and momentum exposure.
What ultimately matters is whether this remains a positioning correction or develops into a deterioration in fundamentals. The first could create a healthier foundation for the next advance. The second would challenge the assumptions supporting the entire AI trade.
Conclusion
The scale of leverage in the U.S. equity market is a legitimate warning signal, but it is not a countdown clock. Margin debt, leveraged products, hedge-fund exposure and equity financing can continue to expand while the market advances. The danger appears when an external catalyst changes the economics of maintaining those positions. A hawkish Federal Reserve, sustained oil above $100, higher Treasury yields, a stronger dollar or a fundamental challenge from low-cost Chinese open-weight AI could provide that catalyst.
Even then, a correction would not necessarily invalidate the longer-term bull case. If economic growth, AI adoption and corporate earnings remain intact, deleveraging could remove speculative excess and create the conditions for the rally to resume.
The path would be unlikely to remain orderly. The point is not that leverage tells us when the market will turn. It tells us how violent the move could become once the turn begins.
