Jamie Dimon believes artificial intelligence will ultimately justify the excitement surrounding it, but the JPMorgan chief has identified a harder question for investors pouring trillions into the technology.
Speaking to The Economic Times, Dimon said AI would eventually pay off much as the internet did, while questioning whether $3 trillion of investment would generate enough revenue to cover the bill.
His warning cuts to the central tension facing Wall Street’s hottest trade: AI does not need to fail for investors to lose money.
If spending races ahead of monetisation, some of today’s biggest beneficiaries could still struggle to earn adequate returns.
AI can change everything and still destroy investor capital
Dimon is not dismissing AI as speculation. JPMorgan has been investing in the technology for 14 years and already uses it across risk, fraud, marketing, document reading and customer generation.
But he told The Economic Times that only part of the infrastructure being financed today has actually been deployed into companies.
“The question is whether, after investing $3 trillion, there will be enough revenue to pay for it,” Dimon said.
His comparison with the internet is important. The technology ultimately transformed the economy, but many companies and investors that financed the late-1990s boom failed to capture the value created.
Jefferies strategist Chris Wood is more bearish. He told Moneycontrol that his base case is that hyperscalers “will end up wasting a lot of money on this AI Capex”.
Wood expects hyperscaler AI spending to approach $1 trillion next year, from roughly $700 billion this year, and argues growing reliance on borrowing could eventually create credit risk.
Also read: S&P 500 record masks $1.5 trillion leverage trap beneath Wall Street’s AI boom
Wall Street can still make the $3 trillion maths work
Morgan Stanley offers the strongest counterargument to the capital-destruction thesis.
The bank estimates major cloud providers could spend more than $1.4 trillion on the AI build-out next year alone, while compute capacity could roughly quadruple between 2025 and 2028.
Yet US internet analyst Brian Nowak’s bottom-up modelling remains constructive. “We see paths to roughly 25 to 50 percent return on invested capital,” he said in a recent Morgan Stanley podcast.
A large next-generation data centre could generate about 30% ROIC by renting AI compute, according to the bank.
For model developers that own their infrastructure, Morgan Stanley’s base case points to returns above 40%, although economics depend heavily on token pricing and improvements in computing efficiency.
Investors may face trouble before AI itself does
The bigger market risk is that AI succeeds while stock returns become less extraordinary.
Goldman Sachs chief US equity strategist Ben Snider estimates the AI investment boom has accounted for nearly half of S&P 500 earnings growth this year. But he expects that contribution to fade in 2027 even as capital expenditure continues rising.
“This tailwind should begin to fade next year,” Snider wrote, according to Business Insider.
That matters particularly for semiconductor companies. Surging AI spending has translated into unusually strong chip demand and margins, but Goldman warns greater supply, lower chip prices or slower margin expansion could weaken S&P 500 earnings growth without any collapse in AI adoption.
The same tension extends to Microsoft, Meta, Alphabet, Amazon, Oracle and the data-centre ecosystem.
Companies can continue investing heavily while shareholders receive weaker incremental returns.
Dimon’s $3 trillion question is not whether AI will succeed, but whether today’s valuations and spending plans assume too much of that future value will arrive quickly and accrue to the companies funding the boom.
The post Jamie Dimon says AI will win but investors may still lose on the $3 trillion bet appeared first on Invezz