Goldman Sachs economists Jessica Rindels and David Mericle examined how the current surge in artificial intelligence investment affects other economic activity. Their analysis finds that AI investment does crowd out some spending, but the macroeconomic displacement is smaller than many media and market commentaries suggest.
How large is AI investment?
The authors estimate AI investment at roughly $600 billion this year, about 2% of U.S. GDP. That amount would represent around 10% of business fixed investment and 15% of equipment investment.
Channels of crowding out
The note identifies three main channels through which AI investment displaces other activity:
-
Reallocation within tech spending: hyperscalers and companies that spend directly on AI services may cut other technology expenditures. For example, corporate IT budgets facing new costs for AI tokens or model usage might trim other software and tech spending. Such reallocation mainly shifts spending rather than reducing overall GDP substantially.
-
Construction crowding out: the data center buildout draws construction labor and equipment away from other projects. Rindels and Mericle point out that gross margins on data center construction are more than twice those on non-tech projects, which has pulled resources toward data centers and away from other builds.
-
Financing effects: the hyperscalers' very large demand for capital has driven a wave of AI-related debt issuance. That push can raise borrowing costs for the rest of the corporate sector, potentially constraining their investment. Goldman Sachs estimates this effect so far has been limited — corporate borrowing costs have risen by about 0.05 percentage point, and non-AI investment may have fallen by roughly $10 billion.
Bottom line
While the AI investment wave is materially large in dollar terms and does redirect resources — particularly within tech spending, construction, and capital markets — the Goldman Sachs analysis suggests the overall crowding-out effect is modest. The authors argue that both the claim that AI is contributing very strongly to U.S. GDP growth and the claim that it is displacing a great deal of other activity tend to be exaggerated.
