Get ready for lower returns on capital from the largest, most important companies on the planet.

In discussing the outlook for US equities in 2026, Goldman’s Ben Snider — who took the reins this year from the retiring David Kostin — noted that over the past 10 or so years, the companies we now know as the “hyper-scalers” generated profits which amounted to between two and three times trailing capex (the ratio dipped in 2022 which, you’ll recall, was a rough year for the Mag7).

Capex of course exploded higher since 2023 as the arms race was joined, with YoY growth rates for outlays peaking near 80% midway through last year. As the by-now-familiar refrain goes, these traditionally capital-light businesses are now capital-intensive, or well on their way to becoming capital-intensive.

The figure on the left, below, is an updated version of a chart we’ve all seen before: It shows the capex breakdown by company and underscores just how quickly these outlays have ballooned.

The figure on the right shows you that on the current trajectory, which is to say assuming consensus is borne out, hyper-scaler capex as a share of operating cash flow will roughly match the tech peak from the dot-com bubble.

Is that a problem? Well, not if the AI dream pans out. If it doesn’t, then yeah, this is an issue. You can’t spend all your free cash on capex (there are buybacks to fund, after all), and as soon as you start resorting to debt, the clock starts ticking in terms of proving these outlays are worth leveraging the company for.

“Given consensus estimates for an annual average of $500 billion in capex from 2025-2027, maintaining the returns on capital to which their investors have become accustomed would require these companies to realize an annual profit run-rate of over $1 trillion,” Snider went on to write, before quickly noting that’s “more than double the 2026 consensus estimate of $450 billion in income.”

To reiterate: Get ready for lower returns on capital from the largest, most important companies on the planet. The figure on the left, below, shows you a decade of history for the profits/TTM capex metric.

The chart on the right (the red annotation’s mine) shows you what happens to the FCF multiple when companies embark on unprecedented capex sprees in hot pursuit of a fever dream.

If you’re concerned about this, the good news is that even if hyper-scaler capex continues to swell, the growth rate will invariably slow which, assuming these companies continue to mint money (grow earnings), will help “boost the diminishing free cash flows of some of the largest AI spenders,” Snider wrote.

But… well, suffice to say there are far more questions than answers currently, and that’s likely to be the case for the foreseeable future.

As Snider put it, “The rise of a new technology inherently introduces uncertainty regarding the size and shape of future profit pools, and these uncertainties are unlikely to be resolved in the next 12 months.”