What gets squeezed as America's AI spending boom accelerates?

If companies shifted capital expenditure budgets toward imported goods and thereby displaced domestic output, "the net effect on US GDP could be negative." The firm, however, said it sees "only limited crowding-out" from technology investment, construction, and borrowing costs, according to a note dated August 10.

Goldman estimated that AI spending crowded out roughly $50 billion of other spending over the last year. It estimated about $30 billion came through displaced technology investment, just over $10 billion through construction, and about $10 billion through higher borrowing costs.

Hyperscalers financed much of their AI investment by reducing buybacks in recent years. Goldman said hyperscalers are now willing to borrow as capital expenditure is set to exceed cash flow, limiting pressure to cut other spending.

According to the brokerage, data-center spending has risen to 9% of private nonresidential construction. It said this coincided with a larger decline in spending on manufacturing facilities supported by the Inflation Reduction Act and CHIPS Act.

Goldman estimated AI investment will add 0.1 percentage point to measured GDP growth in 2026, concluding that claims about AI's GDP contribution and crowding-out effects are "exaggerated."

U.S. artificial-intelligence investment will reach almost $600 billion in 2026, or nearly 2% of U.S. gross domestic product, according to a report from Goldman Sachs economists Jessica Rindels and David Mericle. They wrote that the rapid buildout is causing limited crowding out of other economic activity.