Goldman Sachs forecast a 7 percentage point drop in average return on equity for the largest U.S. tech companies, according to a report published June 12.1 The bank identifies AI capital expenditure as the primary driver, with hardware and infrastructure investment cycles typically lagging revenue generation by 18–24 months.1
The S&P 500 has returned 9% year-to-date on earnings growth.1 Current valuations appear to price in continued profitability — a assumption the Goldman forecast directly challenges.
Five companies sit at the center of the analysis: Microsoft, Alphabet, Meta, Amazon, and Apple.1 Each has committed to multi-year AI infrastructure buildouts. Revenue from those investments has not yet scaled to match the pace of spending.
The ROE compression mechanism is straightforward. Capital flows out now into data centers, chips, and compute capacity. Revenue from AI products and services follows later. The gap between those two timelines is where margin gets squeezed.
For equity investors, a 7pp ROE decline across Big Tech is not a rounding error. ROE is a core input in valuation models. Lower ROE reduces the theoretical justification for premium price-to-book multiples these stocks currently carry.
The 18–24 month lag Goldman cites has precedent in prior infrastructure cycles. Cloud buildouts in the early 2010s also depressed returns before hyperscaler revenues caught up. The question this time is whether AI monetization follows a similar curve — or takes longer.
Traders watching this thesis should track quarterly ROE for each of the five companies over the next four quarters.1 A confirmed average decline in the 5–9 percentage point range would validate the Goldman view and likely weigh on sector multiples.
The near-term risk is asymmetric. If AI revenue ramps faster than expected, ROE holds and the forecast proves too bearish. If capex continues to accelerate while monetization lags, the compression could exceed 7pp — and re-rating risk rises accordingly.
