AI Investment Collapse Could Expose US Economy to Shock: Jefferies

America’s economic boom is riding on artificial intelligence. That is also what makes it fragile. A sudden collapse in AI investment could leave the US economy exposed, according to a report by Jefferies, which warns that AI-related capital expenditure has become one of the single biggest drivers of US growth, and therefore a key point of vulnerability.

The numbers underline the risk. During the first three quarters of 2025, US real GDP rose by $438 billion, or an annualised 2.5%. Of that, $193 billion came from AI-related fixed investment—spending on data centres, information-processing equipment and software—accounting for 44% of the total GDP increase. Only personal consumption contributed more, at $268 billion, or 61%.

AI investment, in other words, has moved from being a growth tailwind to a macro pillar.

Jefferies cautions that this concentration creates a clear risk: if the AI capex cycle proves unsustainable, the shock would ripple far beyond the tech sector. The concern is no longer whether companies will keep spending, but whether they will be able to monetise that spending.

The question is already looming over markets.

AI-related capital expenditure by the four largest US hyperscalers is forecast to reach $480 billion in 2026. By mid-year, Jefferies expects investors to start scrutinising returns more closely, especially at firms borrowing heavily to fund AI build-outs rather than using internal cash flows.

As doubts over returns grow, fears of excess data centre capacity could follow—raising risks for lenders, infrastructure investors and utilities that have bet on a sustained surge in AI-driven electricity demand. US energy stocks, which rallied last year on expectations of AI-fuelled power shortages, remain exposed to this reversal.

The report draws a cautionary parallel with the dotcom era.

If an AI investment bust does occur, Jefferies says costs related to AI “inference” could fall sharply, potentially unlocking a new wave of demand, much as collapsing fibre-optic costs after the dotcom crash accelerated broadband adoption and e-commerce growth. But the adjustment could be painful.

Unlike fibre cables, which last decades, AI chips have a shelf life of just three to four years, making today’s investment cycle inherently more vulnerable to miscalculation and overcapacity.

The bottom line: AI may still reshape the economy but when so much growth depends on one investment engine, the risks of a sudden slowdown become systemic.

(ANI and yMedia are the content partners for this story)