An ’AI winter’ is likely within one to three years, BCA warns

Investing.com -- An “AI winter” is becoming an increasingly likely outcome for the industry over the next few years, according to BCA research.

Stay ahead of shifting AI trends with premium market research and exclusive stock insights by upgrading to InvestingPro - get 55% off today

The rapid boom in AI-related spending is entering a vulnerable phase as expectations embedded in tech valuations far exceed what is likely to materialize, BCA chief strategist Jonathan LaBerge said in a Wednesday report.

“Over a three-to-five-year time horizon, the balance of probabilities points to the emergence of an ‘AI Winter,’ likely beginning over the next one-to-three years,” LaBerge wrote.

“That would involve a slowing in the pace of AI-related CAPEX/data center construction, as well as a meaningful correction in tech/growth stock prices,” he added.

BCA outlines three possible long-term paths for AI. it assigns only a 5% chance to a true artificial general intelligence scenario, where AI breakthroughs validate the sector’s most optimistic assumptions.

Instead, the research house sees an 80% likelihood that AI delivers “moderate macro-level productivity gains,” boosting output by roughly 0.4–0.5% annually—helpful for the economy but insufficient to meet the aggressive profit and cash-flow expectations currently priced into leading tech names.

A more pessimistic scenario, involving a major data center capital misallocation and a productivity bust, carries a 15% probability.

LaBerge believes that U.S. equity markets have already priced in a much more powerful AI tailwind. He estimates that $9–$12 trillion of the market’s gains since late 2022 cannot be explained by earnings or interest rates.

That gap reflects investor assumptions of a powerful and long-lasting AI-driven productivity boom, far stronger than what BCA expects.

The strategist cautions that “it is not enough for artificial intelligence to be productivity enhancing: It needs to boost productivity growth / corporate profits very significantly for current pricing of the overall equity market to be justified.”

AI adoption trends are also showing signs of strain. BCA highlights that business uptake of AI tools has recently slowed, even as capital spending continues to accelerate.

At the same time, the industry is becoming increasingly dependent on debt to finance data-center expansion, with hyperscalers and infrastructure providers issuing large amounts of bonds.

This pattern resembles earlier periods of overbuilding in sectors such as telecoms and U.S. shale, where CAPEX eventually overshot demand, LaBerge said.

Another warning sign is the capital-heavy nature of today’s AI boom. Unlike the asset-light internet era, AI requires continuous investment in chips, power, networking, and cooling.

LaBerge argues that this structure raises the risk of a future compute glut and puts pressure on returns on invested capital, particularly for enablers whose valuations already embed “extremely optimistic assumptions.”

For investors, BCA advises positioning for a long-term environment where expectations gradually reset. The firm recommends underweighting tech-related companies relative to the broader market until valuation excesses have been worked off.

It also sees adopters as better positioned than enablers and monetizers during a cooling phase, noting that industries such as financials, pharmaceuticals, biotechnology, life sciences and defense are likely to be among the biggest beneficiaries of practical AI deployment.