Goldman Sachs says the rapid expansion of artificial intelligence investment in the United States could be starting to displace spending elsewhere in the economy, although the bank has found relatively little evidence of a widespread crowding-out effect so far. With AI-related investment approaching hundreds of billions of dollars annually, analysts are increasingly examining whether the boom is diverting capital, construction resources and financing away from other business activities.
U.S. AI investment could approach $600 billion
Goldman Sachs analyst Jessica Rindels estimates that U.S. investment in artificial intelligence will reach almost $600 billion in 2026, “equivalent to nearly 2% of US GDP.”
AI expenditure has also represented more than 10% of business fixed investment during recent quarters, highlighting the extraordinary scale of spending on data centres, computing infrastructure and related technology.
The pace of expansion raises questions about whether companies are reallocating capital from other projects to fund their AI ambitions.
That issue is particularly relevant because a significant portion of AI spending is directed towards imported technology equipment, potentially limiting the amount of investment that translates directly into domestic economic activity.
Hyperscalers find alternative ways to fund AI spending
Goldman examined whether major technology companies are sacrificing other technology investments to finance their rapidly expanding AI infrastructure programmes.
The bank found that hyperscalers have funded a substantial portion of their spending by reducing share repurchases rather than making major cuts to other investment.
These companies have also been “willing to borrow and appear undeterred by high interest rates.”
Their ability to access capital markets and generate substantial internal cash flows means the AI investment boom has not necessarily required equivalent reductions elsewhere in their businesses.
AI customers are reallocating some existing budgets
The picture is somewhat different among companies purchasing AI services rather than constructing the infrastructure itself.
Goldman’s survey suggested that AI-related costs remain relatively modest for many of these businesses.
However, approximately two-thirds of that expenditure is being financed through reductions in other areas of corporate spending.
This provides some evidence of direct crowding out, although the overall amounts remain relatively small compared with the vast infrastructure investments being made by major technology companies.
Data centres take larger share of construction spending
The rapid construction of data centres is another area where AI investment could potentially compete with other projects for workers, materials and capital.
Goldman estimates that data-centre expenditure has increased to around 9% of total private nonresidential construction spending.
However, this expansion has occurred alongside declining construction of subsidised manufacturing facilities.
That slowdown has freed resources that can instead be absorbed by data-centre development, reducing the risk that the AI construction boom creates severe shortages elsewhere.
Goldman consequently sees “only limited signs of crowd-out nationally.”
AI financing has surged in credit markets
Artificial intelligence is also becoming an increasingly important source of corporate borrowing.
Goldman estimates that AI-related financing now accounts for almost one-quarter of investment-grade bond issuance.
Such a large share could theoretically increase borrowing costs for other companies if AI projects absorb a disproportionate amount of available capital.
So far, however, the evidence of wider disruption remains limited.
Goldman said financing spillovers “look limited so far,” noting that credit spreads for companies outside the AI sector remain close to historical lows.
Economic impact of AI may be less dramatic than assumed
Goldman’s analysis suggests that both the positive and negative effects of the AI investment boom may be smaller than commonly perceived.
Massive spending on computing equipment and data centres is undoubtedly contributing to investment activity, but imports reduce some of its direct contribution to U.S. GDP.
At the same time, fears that AI investment is absorbing capital and resources at the expense of the wider economy appear overstated based on current evidence.
Goldman estimates that incremental crowding out will amount to roughly $50 billion in 2026.
Overall, the bank concluded that both AI’s contribution to economic growth and its crowding-out effects “are smaller than often thought.”