JPMorgan's Jamie Dimon on AI productivity and risk

Speaking at the Economic Club of Chicago, Dimon called AI “more like a tsunami” and said it was producing capabilities “literally five weeks ago” no one would have imagined. He compared the technology’s likely spread to the Industrial Revolution, saying its global adoption would be difficult to control because open-source models and systems are being developed in China and elsewhere.

That view is increasingly consistent with the scale of spending now under way. Google, Amazon, Microsoft and Meta are expected to spend more than $700 billion in capital expenditure this year, much of it tied to AI infrastructure, data centres and model development. The investment boom has supported the broader market’s AI leaders, even as investors debate how quickly that outlay will translate into profits and whether returns will justify the build-out.

Dimon’s comments also speak to a key macro question: whether AI ultimately eases the long-run productivity problem. Businesses from Australia to global development agencies are betting that AI can lift output per worker, and Dimon said it could improve healthcare, transport and workplace efficiency. If that thesis proves right, companies could preserve or even expand profit margins while wages and living standards rise, a rare combination that would reshape both labour markets and valuations.

The bullish case is that AI becomes a general-purpose technology, allowing companies to do more with fewer hours, fewer errors and lower unit costs. The bearish case is that the gains arrive unevenly, while the transition brings higher job churn, compliance burdens and security expenses that offset some of the efficiency benefits. That tension is showing up in banks’ own risk disclosures, with JPMorgan and peers warning that demand for AI products is hard to forecast and that customer behaviour may change in ways that make heavy technology investment less predictable.

Dimon was especially blunt on cybersecurity. He warned that autonomous AI agents could launch attacks around the clock or create risks if granted access to bank accounts. Bank of America chief executive Brian Moynihan has voiced similar concerns, and the issue is already prompting firms to hire outside specialists to hunt for vulnerabilities exposed by advanced models. For financial institutions, the implication is clear: the winners from AI will not just be those that deploy it fastest, but those that can secure it best.

That makes the next phase of the AI cycle less about hype and more about execution. Investors will be watching whether AI spending starts to convert into measurable productivity gains, how much of that value is captured by hyperscalers and chipmakers, and whether rising security costs or regulation slow adoption. Dimon’s message was that the technology’s economic upside could be enormous — but so is the operational risk.