Axios reported on September 30 on a Stanford economists' analysis finding a nearly $1 trillion gap between hyperscaler spending and the revenue they have taken in from AI since 2024. One summary of the report names the economists as Jared Bernstein and Ryan Cummings. They argue that a large share of the spending goes to chips that lose value after about five years.
By their math, hyperscalers would need to triple or quadruple AI revenue next year and sustain that kind of growth every year for the following decade for the investment to pay off. A Goldman Sachs analysis cited alongside it found hyperscaler AI revenue remains below the level needed just to break even on capital expenditure. The bulls dispute the assumptions, so readers should treat this as a scenario, not a settled fact.
Source: Axios