As concerns about an AI bubble grow, one economist warns that investors may overlook other vulnerabilities in the boom.
“The trigger does not have to be endogenous to the boom. It could be external,” Dario Perkins, managing director of global macro at Global Data.TS Lombard, wrote in a note Monday.
Perkins says much of the industry’s revenue growth still comes from spending within the AI ecosystem rather than from businesses and consumers paying for AI services.
“The question is whether this revenue comes from companies and end users outside the ecosystem – presumably to pay for productivity-enhancing tools – or from simple recycling of capital expenditures,” he writes.
The warning follows a recent selloff in chip stocks that has intensified concerns about whether massive spending on AI infrastructure will pay off.
Last week, the Philadelphia Semiconductor index fell more than 11% before recovering 2% on Monday.
The weak links
One potential fault line is debt outside of big tech.
While hyperscalers such as Microsoft, Amazon and Google have strong balance sheets, smaller cloud providers, model developers and infrastructure providers have gained increasing influence.
Perkins estimates that the debt for parts of the AI ecosystem could reach $2 trillion by the end of 2027.
He doesn’t view AI debt as a systemic financial risk, but warns that disappointing results and order cancellations could ripple across the entire ecosystem. Even cash-rich hyperscalers could decide not to renew data center leases, exposing weaker companies.
Another risk is an unexpected glut of chips, as companies would order too many semiconductors to guard against shortages, potentially creating an oversupply if demand slows.
“Let’s not forget that we are talking about technology hardware and chips, which are notoriously cyclical and prone to bullwhip effects,” Perkins wrote.
Higher interest rates could further harm the profitability of AI investments, particularly for companies that rely on borrowing to finance expansion.
But Perkins sees another potential trigger that has little to do with corporate profits: political backlash.
“AI boosters on social media live in a bubble,” Perkins wrote.
“They don’t realize that ordinary people don’t like AI. They don’t find it useful, they REALLY don’t want to live near data centers and they worry about their jobs. And they vote,” he added.
Perkins suggested that a single high-profile incident could quickly turn this unease into broader opposition.
“High-profile accident makes AI toxic,” he wrote.
Gn bussni
