Goldman Sachs research published a new report.

• “Almost half of S&P 500 growth in EPS in 2026 comes from AI investment. The largest US hyperscaler companies are on track to spend $800 billion on capital expenditures this year, an increase of 94% over 2025.”

• “AI investment is expected to transition from an earnings tailwind this year to a marginal drag in 2028.”

i.e., the AI capex boom itself may stop boosting earnings by 2028. From there, the real test becomes whether AI can actually deliver enough productivity gains to replace it.

• AI-era memory economics are running far above normal: roughly 80% gross margins, more than 2× the historical level, with Goldman expecting that margin boost to fade.

• $1.2 trillion of total hyperscaler capex in 2027 and $1.4 trillion in 2028. The catch is depreciation: the equipment keeps generating charges even as capex growth slows, so the earnings lift from the buildout starts getting squeezed.