The artificial-intelligence boom is wobbling even if the bosses of big US banks don’t see it. Chipmakers’ stock prices have dropped in recent days, and there were signs of a hiccup in the third-quarter earnings of Goldman Sachs Group Inc., JPMorgan Chase & Co. and others on Tuesday.

Top executives acknowledged a slowing in trading activity and investment banking fees after an extremely active second quarter but said demand for capital to fund AI infrastructure was insatiable. However, pressures on the economy and the plans of large technology companies are building. The rest of this year looks potentially much trickier for markets and banking.

For the third quarter, profits and returns remained robust for the banks that reported Tuesday, which included Citigroup Inc. and Wells Fargo & Co. along with JPMorgan and Goldman. But beneath the headlines were several missed forecasts and spots of concern.

The sharp rise in Treasury yields and some indigestion from AI-related bond issuance knocked trading in fixed income and slowed fees from new debt and mortgage deals, too. All except Citigroup undershot analyst revenue forecasts for their bond and currency trading desks.

For JPMorgan and Wells Fargo, the selloff in government debt also meant losses on their own bond portfolios. Investors knew that markdowns were coming, but there was still sticker shock at the size of the hits, according to Brian Foran, an analyst at Truist Securities.

Fees for underwriting equity sales and initial public offerings also slowed significantly from the second quarter, which included the listing of SpaceX and Alphabet Inc.’s massive secondary share sale, both of which set records for fundraising. Dealmaking fees held up well, but the pipeline of coming activity in mergers and acquisitions, as well as new listings, has shrunk.

The uncertainty and volatility that has hit markets lately has made investors cautious and put some companies off planned deals. A string of pending IPOs has been pulled in recent weeks, including one for fitness tracker Oura Inc., while OpenAI decided last month not to pursue one this year. Anthropic PBC still plans to go ahead with its own multibillion-dollar IPO between the midterms and the Thanksgiving holiday, according to Bloomberg News, but that’s a tight window that could be difficult to hit if the elections produce unsettling results.

Bank executives said activity had been delayed but that many deals and capital raisings were still waiting to be done. Goldman Chief Executive Officer David Solomon said that while May or June this year had been a top-decile environment for bankers, today was still a top-quartile environment. “That’s still really, really good when you think about long periods of time,” he said.

Banks normally expect a seasonal slowdown in the final few months of the year because of holidays and efforts by many investors to protect the returns they’ve made through the year. But JPMorgan CEO Jamie Dimon, who is normally the first to highlight storms on the horizon, said this year might be different. “I think the fourth quarter will have not the same seasonal decline,” he said. “Too much going on in the world.”The trouble, though, is that the AI infrastructure trade is everywhere.

Jeremy Barnum, JPMorgan’s chief finance officer, said it wasn’t just in construction lending; it was also in working capital revolvers, stock markets and almost anywhere else you could look.

The bank’s total lending exposure to the AI theme was modest in terms of JPMorgan’s overall corporate and investment bank loan books, Barnum said, but that doesn’t mean the bank wouldn’t be hurt by a significant change in sentiment in the AI space. “I think I would be surprised if anyone, frankly, was immune to that at this point,” he added.

And that’s the key point. Citigroup CEO Jane Fraser tried to highlight other factors helping bank earnings. “While much has been made of the AI boom, the consumer was still driving much of the growth,” she said. But AI spending is flooding through the whole economy, supporting employment, wages and growth beyond the immediate tech sector.

Executives talked repeatedly on Tuesday about how they were closely monitoring the risks being taken in financing companies involved in the AI boom, private credit firms helping to fund it, and the hedge funds betting on the stock markets being inflated by it.

But credit and equity markets are still priced for something close to perfection from the current explosion of capital expenditure. And yet, plenty of people, including bank executives, admit they have no idea how much of this infrastructure is truly needed.

“No one knows how the demands for capital and compute will evolve as the broader AI ecosystem matures,” Solomon said.

The pressures are building. Energy costs aren’t going to fall while Eastern Europe and the Arab Gulf are wracked by war; the AI boom itself is helping to keep inflation higher for longer; and the elevated levels of interest rates and Treasury yields are going to crimp borrowing and spending by companies, consumers and the government sooner or later.

The fear of missing out on AI supremacy is a powerful force, but it will hit a wall. And in the US, the political atmosphere around rising living costs and the giant footprints of tech companies across the countryside might just be the barrier that this boom hits.