As artificial-intelligence capital expenditures rise, so do the risks for AI stocks, Goldman Sachs tells investors

By Jules Rimmer

Forecasts of $920 billion in AI-related capex in 2027 are likely too conservative

The market's expectations for capital spending among so-called hyperscalers in 2027 are way too conservative. Combined with valuations of AI-related stocks expanding to post-ChatGPT-launch highs, this is a recipe for volatility in the sector, and investors must brace themselves for it.

A note from Goldman Sachs's portfolio strategy research team warned investors on Wednesday about the mounting risks associated with hyperscaler capex estimates that seem to rise inexorably. At present, that analyst consensus for 2027 is $920 billion, which, though higher than in 2026, still shows a deceleration in the rate of growth from 84% this year to 22% next.

Taking historical precedents for the build-out of new technologies into account, like the development of railroads or the automobile industry, the Goldman team thinks a more realistic figure for next year is $1.1 trillion. It points out, though, that a more extreme upside scenario could suggest a figure of $1.4 trillion. That would represent a growth rate similar to this year's.

Hyperscaler 2027 capital-expenditure scenarios.

The Goldman report, authored by Ryan Hammond and five colleagues, highlights the risk of this unprecedented capital investment, however. The price-earnings multiple of the median AI infrastructure stock is now 26, the highest multiple since ChatGPT was launched in the fall of 2022.

That median multiple has increased most sharply among semiconductor and power-generation stocks while the median multiple among hyperscalers and memory-chip makers has not, Hammond points out. The challenge facing investors then is to pick the winners between the AI-enabled and the AI-disrupted because the research suggests return dispersion among these stocks will be meaningful.

For example, the Goldman note highlighted that 54% of companies discussed AI productivity on their first-quarter earnings calls, but only 11% quantified the benefits, while even fewer, just 2%, quantified the impact on earnings. The reaction of the market to these announcements, however, showed little differentiation.

Looking forward, Goldman anticipates lower-cost competition will start to eat into the growth and margins of the incumbents, and this largely explains why software valuations peaked at 39 times earnings last year and now stand at 25 times, having touched a low of 21 in March.

The upside potential of AI capex should drive further upside to earnings and share prices of AI infrastructure companies in the near term, but Goldman does emphasize that "rapid price appreciation, elevated valuations and crowded positioning increase the risk of volatility."

-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

06-11-26 0756ET