Goldman Sachs: AI Investment Boom to Last at Least Through 2028
nashnova research
Goldman Sachs chief economist Jan Hatzius sees almost no chance of an outright decline in AI spending before 2028 — this means → the AI tailwind for corporate earnings has at least two more years to run.
Why is S&P 500 earnings growth hitting 36% this year?
Goldman forecasts S&P 500 earnings per share (EPS) will grow 36% in 2026.
Two drivers: a one-off boost from private-equity investment gains, and windfalls from AI spending.
Semiconductor and memory-chip companies see the sharpest margin expansion — this means → the engine behind this earnings surge is AI hardware, not software or applications.
How much does AI infrastructure actually add each year?
Goldman estimates AI infrastructure contributes roughly 15 percentage points to S&P 500 EPS growth in 2026, 11 pp in 2027, and 8 pp in 2028 — fading year by year, yet still a double-digit push each time.
Hyperscalers — Amazon AWS, Microsoft Azure, Google Cloud — add another 4, 3, and 3 pp respectively.
In plain terms = the "dosage" of AI spending on earnings is shrinking, but withdrawal is nowhere close — 8 pp of support remains even in 2028.
Why does growth slow in 2027?
Goldman expects S&P 500 EPS growth to ease to 11% in 2027, mainly as private-equity gains fade and AI spending nears its peak.
Hatzius notes that an outright decline in AI investment would steepen the slowdown — but Goldman sees this scenario as unlikely before 2028.
This means → the 2027 "deceleration" looks more like shifting from the highway to a fast road, not slamming the brakes.
Does the "earnings bubble" argument hold up?
Some market voices warn current earnings levels are a "bubble." Goldman calls this overblown.
The bank acknowledges earnings are elevated relative to a normal cycle but attributes the gap to a structural AI-driven lift, not a bubble.
This reflects Goldman's core thesis: the moment AI investment truly peaks will be the defining pricing variable for the next two years — in other words, debating "bubble or not" matters less than tracking when AI capex rolls over.
市场有风险，内容仅供研究参考，不构成投资建议。