AI capex will destroy capital, could trigger major credit event, says Chris Wood

The latest market report highlights that The AI investment boom that has propelled semiconductor stocks and powered US earnings expansion will ultimately destroy a large amount of capital and could trigger a major credit event when the cycle turns, according to Jefferies’ Chris Wood.

Wood, one of the better-known global equity strategists, stated the first three years of the AI Capex boom were largely funded by the enormous free cash flows of hyperscalers. That is changing, with firms increasingly turning to debt to fund the next leg of spending.

“Something will give eventually, without a doubt,” Wood stated in an exclusive interview with Moneycontrol.

Hyperscaler AI Capex is anticipated to approach $1 trillion next year, up from around $700 billion this year, according to Wood. The scale of that spending has translated almost dollar for dollar into earnings for semiconductor firms, creating what he calls the biggest semiconductor cycle ever.

But Wood believes the economics of the spending will eventually come under scrutiny.

The key question for markets, he stated, will be whether hyperscalers can generate an adequate return on the enormous sums being invested in AI infrastructure. His base case is that they cannot. “My base case is that they will end up wasting a lot of money on this AI Capex, a lot of capital will be destroyed,” he stated.

Debt-funded AI boom raises credit risk

The shift from cash-funded investment to borrowing is central to Wood’s concern.

For the first three years of the AI boom, hyperscalers were spending their substantial free cash flows. Increasingly, that stated, they are borrowing to finance the next wave of investment and issuing large amounts of investment-grade corporate bonds.

That creates another source of competitive pressure for funding alongside the US Treasury.

“The more they're borrowing, the bigger the risk that when the downturn comes, it will be a major credit event,” Wood stated.

The financing burden is additionally not fully visible in headline corporate accounts, he stated, pointing to data-centre leases being kept off balance sheet. “Then you have on top of that the circular financing arrangements between NVIDIA, Microsoft and these firms, OpenAI and Anthropic,” he stated.

For Wood, the concern is as a result not simply whether AI valuations are too high. It is whether the enormous capital being committed to the sector will ultimately earn an adequate return. LLMs will become commodities

Wood's longer-term view is even more stark: the technology itself will eventually become commoditised. “My long-term outlook is that these large language models will become commodities,” he stated.

That does not mean he believes the entire AI investment story is fictitious. “What’s 100% real is the semiconductor stock firms are making real earnings,” Wood stated.

Semiconductor firms are at present capturing the economics of the spending upfront, while hyperscalers are carrying the investment and will have to generate returns from the infrastructure over time.

“The tech stocks who are selling the semiconductor, the chips, put their earnings up front,” Wood stated. “But the people who are paying for the Capex, like the hyperscalers, they are in no hurry to fully pay for the Capex via depreciation.”

The result is an AI Capex cycle that is “front-end loaded” and supportive of current earnings.

Easy money in semiconductors may be over

For market participants, that creates an awkward distinction between the reality of current semiconductor earnings and the sustainability of the broader AI investment cycle.

Wood believes the easy upside in the semiconductor trade have already been made.

“The easy money has been made in the AI Capex trade,” he stated.

For market participants who benefited from the sharp climb in semiconductor stocks, he stated, “it would make a lot of sense to take some earnings.”

The risk is that markets eventually stop accepting the premise that ever-rising AI investment will generate sufficient returns.

“As long as it continues, this can continue,” Wood stated of the AI Capex cycle. But if the market begins to question whether hyperscalers will earn a proper return on their investments, the cycle could change rapidly. Why this matters for India

The AI boom has additionally changed the relative attractiveness of India for global market participants.

Most foreign equity market participants in India are global emerging-market market participants, Wood stated. For much of the past two decades, India was a major overweight for such market participants.

But the enormous earnings being generated by semiconductor firms have redirected capital towards Taiwan and, even more dramatically, Korea.

Before the AI story emerged, India was “the best structural expansion story in global equities”, Wood stated.

The problem is not that India's expansion story has disappeared. Rather, it has become less compelling relative to the scale of the AI opportunity.

“If you wanted foreign money to come pouring back into the Indian market, the best thing that could happen would be for the whole AI story to implode,” Wood stated.