Goldman, JPMorgan and Morgan Stanley earn $52 billion in AI-fueled trading boom

Goldman Sachs, JPMorgan Chase and Morgan Stanley earned a combined $52 billion in stock-trading revenue through September 2026, nearly 80% of the top five investment banks’ total. CNBC’s October 21 report ties their lead to surging demand for loans backing AI-related trades—a business that rewards large balance sheets while raising concerns about leverage.

The five largest investment banks generated about $67 billion in equities revenue through September 30, according to company filings cited by CNBC. That was already $7 billion above their full-year 2025 total, itself a record. These are revenues across their stock-trading businesses, not a tally of income earned solely from AI investments.

More room to lend

US regulators finalized looser leverage rules for the largest banks in 2025, with adoption allowed from January 2026. The changes gave banks more room to use their balance sheets for businesses such as lending to hedge funds.

Those loans sit inside prime brokerage: services that include lending against investors’ holdings, lending securities and helping hedge funds trade. Borrowing lets funds make larger bets. Bank executives said last week that demand had reached new heights for trades involving AI hardware, large cloud operators and chip suppliers.

JPMorgan chief financial officer Jeremy Barnum agreed with an analyst’s description of “insatiable demand for capital” on the bank’s third-quarter earnings call. He said AI had become an overwhelming theme, with capital being tapped through many different mechanisms.

$3.37 billionGoldman Sachs equities financing

Third-quarter revenue reached a record, up 96% from a year earlier.

The advantage of financing a giant trade

JPMorgan and Morgan Stanley do not disclose the same equities-financing figure as Goldman, but both described similar demand. Wall Street executives say this borrowing has helped turn the biggest banks’ equity desks from businesses dependent on individual transactions into more durable, high-margin lenders.

Size is a practical advantage. Financing multibillion-dollar bets on volatile AI stocks or sector funds can require an enormous balance sheet. Trading executives told CNBC the largest banks can absorb a potential $100 million loss without threatening their quarterly results. Smaller rivals may have to decline that exposure or surrender the business.

The fastest-growing regional opportunity was Asia, trading executives said. JPMorgan’s global equities head, Rachid Alaoui, described institutional investors targeting memory-chip makers and semiconductor manufacturers in South Korea and Taiwan. The three banks’ regional networks positioned them to serve those trades.

The revenue comparisons have limits. Transaction taxes and stamp duties can inflate reported Asian prime-brokerage revenue. Banks also count different activities within equities revenue; some include clearing services that others leave out. The totals therefore are not perfectly comparable measures of each bank’s business.

When borrowers have to retreat

The International Monetary Fund warned this month that rising hedge-fund borrowing, crowded bets and investors selling together could threaten financial stability. Analysts also warned that the leading banks could be hit harder than rivals if AI proves less transformative than investors expect.

Banks counter that collateral and frequent stress tests help contain the danger. Alaoui pointed to forced hedge-fund unwinds this summer: investors reduced borrowing, risk fell and markets moved on. Prime brokers have learned to demand additional collateral and cut exposure quickly when clients’ positions move against them, he said.

Stanford finance professor Darrell Duffie offered a more guarded assessment. Relaxing capital requirements had made the system somewhat more leveraged, he said. He called that unwise, while adding that conditions had not become dramatically out of balance. Barnum likewise acknowledged hedge-fund leverage was not riskless, describing it as a known uncertainty rather than an unfamiliar threat.