AI Boom Widens Divide as Banks Expand Leverage

JPMorgan Asset Management says the K-shaped U.S. economy could persist: lower-earning households are struggling while higher-earning Americans do well, and capital gains are expected to outpace labor income in coming years. The bank warns that this divide could become a source of economic and market volatility. People who own assets are collecting the boom; those without stocks have been left out.

Who Owns the Upside

The S&P 500 has risen 70% in five years. In the second quarter, households in the top income quintile held $56.3 trillion in stock and mutual-fund wealth, or 87% of all household stock and mutual-fund wealth, according to the latest Federal Reserve data. JPMorgan expects U.S. large-cap stocks to grow about 7.5% a year.

Artificial intelligence could lift U.S. total factor productivity growth to about 1.2% annually in JPMorgan’s base case. But the bank says it’s unclear how those productivity gains will reach lower- and middle-income Americans. Its report calls the K-shaped economy “a metastable state — surprisingly durable, but sensitive to stress or shock,” and describes a miscalculation in the balance between its upper and lower parts as “the greatest risk” to its outlook.

Workers could demand more resources than economists expect, the report warns, destabilizing the broader economic and financial system. Efforts to redistribute AI-related capital returns, it says, could discourage investment and trigger knock-on effects. If the investment boom fails to significantly raise U.S. productivity, “widespread capital misallocation” could slow economic growth. Even higher-earning Americans are pulling back on spending amid hotter inflation and high gas prices.

The Banks’ Leverage Machine

The five biggest investment banks generated about $67 billion in equities revenue through Sept. 30, $7 billion more than they earned during all of 2025, which was itself a record. Goldman Sachs, JPMorgan Chase and Morgan Stanley brought in $52 billion in stock-trading revenue so far this year, nearly 80% of the total.

Rising stock-market values and increased institutional and retail trading in stocks, ETFs and derivatives have accompanied the banks’ trading gains. Their prime brokerages have loaned money to hedge funds so they can increase their bets. Bank executives said demand for leverage to invest in AI hardware, hyperscalers and chip suppliers reached new highs. Goldman’s equities financing revenue rose 96% in the third quarter to a record $3.37 billion.

JPMorgan CFO Jeremy Barnum called demand “insatiable,” saying, “In this moment where AI is such an overwhelming theme ... there is capital being tapped through all kinds of different mechanisms.” Wall Street executives say the activity has turned the Big Three’s equities desks into more durable, high-margin lending businesses. Wells Fargo banking analyst Mike Mayo said the Big Three “consider each other their main competitors” in the AI trade and have been best positioned to benefit.

Regulators have also made more room for that business. Trump administration regulators eased some leverage rules for the largest U.S. banks; they finalized the revised rules in 2025, and banks could adopt them as early as January. The changes eased balance-sheet caps that had treated Treasurys and collateralized prime-brokerage loans like riskier assets. Stanford University finance professor Darrell Duffie said, “Capital did go down, so you have to say the system is somewhat more levered because of the relaxation in regulations.”

A Boom With Known Unknowns

The International Monetary Fund warned this month that rising leverage, crowded positions and correlated selling by hedge funds could threaten stability. It estimated global hedge-fund assets at $13 trillion. In August, JPMorgan CEO Jamie Dimon said industrywide leverage was “the highest it’s ever been.” After some investors had to unwind positions during the summer, JPMorgan global head of equities Rachid Alaoui said, “They deleveraged, risk came down, and the market moved on.”

The AI buildout is also lifting prices for computing capacity and chips. Nebius said prices for AI computing were rising more than expected. Its chief revenue officer, Marc Boroditsky, said Nebius would sell out its 2027 capacity if it brought it online, and that 2028 could show “the same type of dynamic.” Chase Lochmiller, CEO of Nvidia-backed AI company Crusoe, said prices for older Nvidia chips were higher than when they were first deployed.

Markets have shown how quickly the boom can wobble. The S&P 500 fell 1.1% last Friday during a final wave of AI selling, while the Cboe Volatility Index rose from below 16 to above 22 in two hours. The State Street SPDR S&P 500 ETF Trust had its lowest full-day trading volume since 2003 on Oct. 12. JPMorgan CFO Barnum put the remaining uncertainty plainly: “Hedge fund leverage as a contributor of risk to the system is very much in the category of known unknowns. It doesn’t mean it’s riskless.”