Midterm elections provided the hyperscalers with the excuse they needed to cut back on spending, says Jefferies strategist

By Jules Rimmer

Reining in capex may prove to be positive for AI plays, if it prevents 'malinvestment'

Slowing down the proliferation of data centers may turn out to be a positive for the AI sector over time, argues Chris Wood.

Chris Wood, the veteran Jefferies equity strategist, thinks the AI capital-expenditure cycle has peaked. He doesn't necessarily think that's a bad thing.

When it comes to investing capital, Wood favors "disciplined deployment" over "indiscriminate spending," and so the constraints that may be imposed by a Democratic majority in both houses of the U.S. Congress could be to the advantage of the hyperscalers.

In the latest edition of his long-running weekly publication "Greed and Fear," dispatched to clients on Thursday, Wood relates that his conversations with investors confirm his theory on AI capex. This year it's forecast to reach a number just shy of $1 trillion, and the consensus for 2027 hovers around the $1.6 trillion level.

However, Wood and his quant colleague Desh Peramunetilleke observe a "dissonance" between the accelerating returns and the declining investment. In fact, even as AI capex consensus rises, the predicted return on invested capital is falling. In 2024 the median ROIC was 27.2%, but it's expected to drop as low as 15.7% by 2028.

Capital intensity among hyperscalers Amazon, Meta, Alphabet and Microsoft vs. the ROIC trend.

The hyperscalers (AMZN) (META) (GOOG) (MSFT), Wood notes, must now consider their spending in light of the so-called blue tsunami of the midterm elections last week. The expansion of data centers will probably slow and the tax breaks provided by the Republicans' One Big Beautiful Bill Act to incentivize capex are threatened also. This explains why the forward price-earnings multiple of the four largest hyperscalers has declined from around 30 times in the summer of 2023 to around 21 times now.

For Wood, though, the slowdown is no bad thing. It may actually deter what he calls "malinvestment." Putting it in an historical context, outsized capital-spending booms like Japan's in the mid '90s or the wider Asia's in the late '90s "rarely generate proportionate investment returns." Wood quotes a Brookings Institution paper, published recently, that predicted AI capex would average 3.63% of GDP between 2025 and 2032. It's a lot higher than the figure of 2.24% of GDP for the late 19th century's railroad boom.

A history of U.S. capex booms depicts average annual spending as a percentage of GDP.

For a turnaround in the AI sector's fortunes, Wood asserts that fresh evidence that the large language models have pricing power would restore confidence. That would prove they have recognizable moats.

Since the AI thematic has peaked, Wood considers the rotation that's taking place to be a consequence. Pharmaceutical stocks are an obvious beneficiary. Wood points out the pharmaceuticals index XX:SP600.35202010 has outperformed AI and "Big Data" stocks by 31% since early October. The Democrats' victory might mean more healthcare spending.

However, the real beneficiaries may be in emerging markets, where Wood identifies India is profiting from a rotation in asset allocation, having been out of favor for a couple of years. The market there is now much more reasonably valued, with its Nifty index IN:NIFTY50 trading at 18 times forward earnings, domestic inflows from mutual funds resilient and private capex surging.

One more sector Wood highlights amid diversification away from AI is energy XLE. He's worried that as a lame duck, Trump may, like others before him, turn to foreign policy, where he has fewer constraints, for distraction. This geopolitical risk to investors, Wood argues, can only be hedged via energy stocks. He recommends Petrobras (PBR), for example, whose U.S.-listed depositary receipts have plunged by a third since Oct. 23.

Lastly, one other contrarian and non-AI trade Wood suggests is the Chinese stock market CN:SHCOMP.

Bond yields continue to fall, last printing a 1.6% BX:AMBMKRM-10Y yield to maturity for 10-year notes while the equity market trades on just a 9 times forward PE multiple.

-Jules Rimmer

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

11-13-26 0843ET