U.S. corporate profits reached a record seasonally adjusted annual rate of $4.71 trillion in the second quarter of 2026, according to the Bureau of Economic Analysis. The result supports the case that artificial-intelligence investment has lifted earnings, but it also raises a harder question: how long can capital spending grow faster than the productivity it is supposed to create?
A powerful earnings cycle
BEA data show profits from current production increased by $400.9 billion in the second quarter after a $74.4 billion rise in the first. Goldman Sachs Research estimated that AI investment accounted for almost half of S&P 500 earnings growth in 2026. The source also cited 51% year-over-year S&P 500 earnings growth in the second quarter and a 29% estimate for the third.
Those S&P growth rates are far above the roughly 7% historical average cited by Goldman. They need careful interpretation because index earnings can be concentrated in a small group of large companies and affected by comparisons with a weak prior period. National corporate-profit data and S&P earnings are also different measures, so their growth rates should not be treated as interchangeable.
Spending is the bridge to productivity
Goldman estimated that the largest hyperscalers were on pace to invest $800 billion in 2026, 94% more than in 2025, with capital spending projected at $1.2 trillion in 2027 and $1.4 trillion in 2028. Those forecasts show how much of the earnings outlook depends on continued infrastructure demand.
The spending has two effects. It generates revenue for chip, networking, power and construction suppliers today. It also raises depreciation and creates pressure on cloud and software companies to earn adequate returns from the new capacity. If AI products improve worker output or create durable new revenue, the investment can support profits beyond the buildout. If utilization or pricing disappoints, the same spending can reduce future margins.
Valuation is not a full defense
Goldman argued that equity prices had not kept pace with earnings and therefore near-term valuations did not show an earnings bubble. That claim is narrower than saying the market is cheap. Earnings can decelerate while valuation multiples remain elevated, especially when profits are concentrated in companies exposed to the same investment cycle.
The next evidence should come from cloud revenue, AI-service pricing, data-center utilization and depreciation growth. Corporate profit growth can slow without collapsing, as Goldman expects, but the quality of earnings will matter more as the comparison base rises. Investors should look for productivity and cash returns from AI spending rather than assume that another year of larger capital budgets is sufficient.