AI investment collapse could make US economy vulnerable: Jefferies

New Delhi [India], January 9 : A sudden collapse in artificial intelligence (AI) investment could make the United States economy vulnerable, as AI-related capital expenditure has emerged as a key driver of economic growth in recent years, highlighted a report by Jefferies.
The report said the macroeconomic risks from an unsustainable AI investment cycle are becoming increasingly clear, noting that AI-related capital expenditure was one of the main contributors to US economic growth last year.
It stated "the macro vulnerability to a sudden collapse in AI capex is clear since that was the main driver of US economic growth last year after personal consumption, based on national accounts data".
According to Jefferies, US real GDP increased by USD 438 billion, or an annualised 2.5 per cent, in the first three quarters of 2025.
Over the same period, real personal consumption rose by USD 268 billion, while real AI-related fixed investment, including private investment in information processing equipment, software and data centre construction, increased by USD 193 billion.
This translated into annualised growth rates of 2.2 per cent and 19.8 per cent, respectively.
As a result, real AI-related investment accounted for 44 per cent of the increase in real GDP during the first three quarters of 2025, compared with a 61 per cent contribution from real personal consumption, the report said.
Jefferies noted that a key issue for the US stock market at the start of 2026 is whether hyperscalers will be able to monetise their AI capital expenditure. AI-related spending by the four major US hyperscalers is forecast to reach USD 480 billion this year.
The report warned that investors are likely to start questioning the returns, or lack thereof, from the AI capital expenditure arms race by the middle of this year.
It added that signs of this process were already visible by the end of last year, with increasing focus on companies that have been borrowing to fund AI capex rather than using cash reserves.
As concerns around AI returns grow, Jefferies said this could also lead to worries about potential excess capacity in data centres. Such a scenario may raise further concerns about those financing data centre construction and related infrastructure.
The report added that similar risks continue to apply to US energy stocks, which rallied last year on expectations of increased electricity demand to address America's power bottlenecks driven by AI and data centre expansion.
Jefferies said that while AI agents are expected to arrive in the future and may have significant commercial implications, large investments have already been made.
It noted that if an over-investment bust occurs, costs related to AI "inference" could fall sharply, potentially boosting demand, similar to the dotcom bust, when excess fibre optic capacity led to a collapse in broadband costs and a surge in e-commerce demand.
The report also cautioned that AI chips have a much shorter shelf life of around 3-4 years, compared with roughly 25 years for fibre optic cables, making the current investment cycle more vulnerable.