Goldman tells clients the profit boom is a cyclical peak, not a bubble

Goldman Sachs chief US equity strategist Ben Snider pushed back on the idea that US corporate profits are in a bubble¹. “Our base case is for S&P 500 earnings growth to decelerate, not collapse, in coming years,” he wrote in a client note³.

The starting point is unusually high. Snider noted that S&P 500 earnings per share grew 51 per cent in the second quarter and 26 per cent over the past four quarters, lifting profits well above trend². That pace suggests companies are over-earning as AI investment surges¹. His team attributes the above-normal earnings to three temporary factors, AI-related effects among them⁷, which one summary described as hyperscaler capex, high memory margins and non-operating income⁹.

Why it matters to investors

The bank has put a number on the deceleration. Snider’s team forecasts S&P 500 earnings per share growth of 11 per cent in 2027 and 2028⁴, a sharp step down from the current run rate². That is the bull case stated plainly: a slowdown to a normal growth rate rather than a reversal. It also concedes the bears’ premise, that the present level of profitability is not sustainable, while disputing only the shape of the descent^1,3.

The tailwind has a visible end date. Goldman warns that AI spending will not sustain stock-market momentum into 2027, with S&P 500 earnings slowing as capex wanes⁵. Business Insider framed the same research as the uncomfortable truth the capex boom has been masking⁶. Snider’s own caveat is that while there are indeed factors contributing to over-earning today, the base case is for deceleration⁸.

What to watch

Not everyone on the street is as calm. The veteran market strategist Wood, speaking to CNBC-TV18, expressed apprehension about the enormous amounts being invested in AI by American hyperscalers and warned of massive capital destruction in the United States¹⁰. That is the tail the Goldman base case excludes rather than refutes^3,10. The two views are not reconcilable by argument, only by the capex schedule itself^5,10.

Three things test the thesis. Whether hyperscaler capex plateaus on the schedule Goldman assumes^5,9; whether memory margins and non-operating income normalise together or separately^7,9; and what replaces AI capex as an earnings driver once the temporary factors roll off^4,7.