Wall Street's Biggest Banks Are Funding the AI Boom and Warning About It at the Same Time

The banks that lend to and raise money for the AI build-out spent Tuesday's earnings calls sounding a common warning. Goldman Sachs (GS), JPMorgan Chase (JPM) and Wells Fargo (WFC) all said demand for AI computing remains enormous. All three also said the boom is increasingly being paid for with borrowed money, and that rising financing costs or overbuilding could eventually force a painful correction. None of them is pulling back yet, but the tone has shifted from cheerleading to caution.

Three CEOs, one message

Goldman: a reset is coming, the question is when

Goldman Sachs CEO David Solomon warned that rising financing costs and possible overbuilding could trigger a painful reset in AI infrastructure investment. His concern is simple math: as capital requirements grow, borrowing rates could rise, cutting into project returns and making some data center developments financially unviable.

For now, Solomon said demand for computing power far exceeds supply, which supports continued construction and financing. Goldman sees big opportunities in AI infrastructure financing but is being selective about how much capital it commits and how much risk it takes on. He expects an eventual recalibration, but said its timing and severity remain uncertain as the technology and demand keep evolving.

JPMorgan: few would escape a real downturn

JPMorgan's message was about the system as a whole. CFO Jeremy Barnum said the bank's direct exposure to AI is relatively modest, but it is monitoring risks across data center lending, corporate credit lines and capital markets. He warned that few institutions would escape a major AI downturn. JPMorgan is now evaluating whether the financial system could withstand a sharp deterioration in AI investment sentiment.

The scale explains the concern. Goldman Sachs estimates global AI capital spending will total $7.6 trillion over the next five years, raising questions about whether there is enough financing capacity to go around and how fiercely projects will compete for capital.

CEO Jamie Dimon added two more points. On rates, he said a rise in borrowing costs from 5% to 6% would be more painful than the move from 4% to 5%, especially as existing debt is refinanced, though he stressed he was not forecasting such an increase. On AI security, he said JPMorgan will block outside autonomous AI agents from operating inside the bank, citing customer security and the risk of unauthorized activity.

Wells Fargo: strong demand, familiar risks

Wells Fargo CEO Charlie Scharf described demand that is still surging across the whole AI supply chain: semiconductors, networking equipment, power generation, data centers, cooling systems, storage, cybersecurity and advanced manufacturing equipment. But he compared today's spending to previous infrastructure booms and warned that heavy borrowing and rising equipment costs could threaten the sustainability of some investments.

Inside the bank, Wells Fargo is expanding its own use of AI in coding, marketing, risk management and customer service, while weighing the security risks of third-party AI agents. The efficiency push shows up in staffing: head count fell 2% from the prior quarter, 8% from a year ago and 30% from its 2020 peak, and it has declined every quarter for six straight years.

Why it matters

The AI build-out is moving from a phase funded mostly by the cash of giant tech companies to one that leans more on debt: data center loans, private credit and bond deals. That puts the banks in the middle. They earn fees and interest from the boom, which is why none of them is walking away. But they are also the ones who would absorb losses if projects stop paying off, and their comments suggest they are building in more caution now, while demand is still strong.

The warnings land on a tense day for AI stocks. Earlier on Tuesday, Broadcom CEO Hock Tan compared AI investing to a casino, Microsoft unveiled a chip that could reduce its reliance on Nvidia, and a selloff in Asian chip stocks wiped out about $800 million in leveraged crypto bets.

What to watch

• Credit spreads on AI debt. Rising borrowing costs for data center developers would be the first sign that Solomon's financing squeeze is arriving.
• Hyperscaler capex guidance. Microsoft, Alphabet, Amazon and Meta report later this month, and their spending plans set the pace for the whole build-out.
• Long-term interest rates. Dimon's 5%-to-6% warning makes yields a direct risk for AI projects that need to refinance.
• Other banks' commentary. Whether more lenders echo the caution as earnings season continues.