Written by: Rita

US AI investment is expected to reach $600 billion by 2026, equivalent to nearly 2% of GDP, accounting for more than 10% of corporate fixed investment. Goldman Sachs raised an important question in its US economic analysis report published on August 10, which is rarely asked by the market: What is AI investment crowding out? The answer is that the crowding-out effect is limited. Three main channels—substituting other tech investments, squeezing other construction, and pushing up interest rates—together crowd out about $50 billion. The direct contribution of AI to GDP growth is only 0.1 percentage points, far lower than the intuitive impression given by the 2% share of AI investment in GDP. Goldman Sachs believes that the market's narrative about AI driving growth is somewhat exaggerated, and concerns about AI crowding out other activities are also somewhat overblown.

Three main crowding-out channels total about $50 billion, actual impact is mild

Goldman Sachs dissected the three channels through which AI investment crowds out.

The first is substituting other tech investments. Over the past few years, large-scale firms have financed AI investments by reducing share buybacks, and now that capex has exceeded cash flow, they have begun to incur debt, with high interest rates not deterring them. Corporate IT spending surveys show that AI costs still only account for 1% to 5% of IT budgets for most companies, with one-third coming from incremental spending and two-thirds from cuts in other expenditures. Goldman Sachs estimates that this portion crowds out about $30 billion.

The second is squeezing other construction. Data center construction has risen to 9% of private non-residential building spending, but this is offset by a decline in manufacturing facilities construction supported by the Inflation Reduction Act and the CHIPS Act. Goldman Sachs estimates that data center construction has crowded out about $10 billion of other construction activities. The crowding-out effect is significant in certain regions, with some states seeing data center construction account for more than half of non-residential building spending, but national data does not show a significant crowding-out effect. In states with more data center construction, the growth rate of construction wages has not significantly exceeded the national average.

The third is pushing up interest rates. AI-related bond issuance has accounted for nearly a quarter of the total issuance of investment-grade bonds. Goldman Sachs estimates that AI issuance has raised borrowing costs for non-AI companies by about 5 basis points, corresponding to a reduction in investment of about $10 billion.

AI directly drives 0.1 of GDP, real drive is 0.3

The 2% share of AI investment in GDP is a concept of total expenditure, not GDP contribution. A significant portion of the $600 billion consists of imported equipment; some AI investment is categorized as intermediate inputs in the national accounts, not as final expenditures. Goldman Sachs' calculations show that the officially reported direct contribution of AI to GDP is about 0.5% of GDP, far lower than the total expenditure scale of 2%. In terms of annual growth rates, the direct drive of AI investment on GDP growth in 2026 is about 0.1 percentage points.

If we adjust the statistical criteria to reclassify semiconductor investments used for training AI models as final expenditures and include exports of chip design services in the statistics, the real GDP drive is about 0.3 percentage points. After considering the wealth effect (the boost to consumption from rising stock prices of AI companies), the drag on real income from rising electricity prices, and the crowding-out effect of about $50 billion, the net impact is about 0.2 percentage points.

Market narrative is overestimated: AI is neither a panacea nor a vampire

The core contribution of Goldman Sachs' report is to provide a calibration framework. There are two voices in the market: one believes that AI is driving growth on a large scale, while the other believes that AI is significantly crowding out other activities. Goldman Sachs' data shows that both situations have been exaggerated.

The $600 billion is a huge number, but most of it is spent on imported equipment. Data center construction has indeed created a tightness in construction resources in certain areas, but it is offset at the national level by a decline in manufacturing facilities. AI bond issuance has indeed pushed up interest rates, but the impact is only 5 basis points. AI investment is neither a panacea nor a vampire. Its actual impact on the economy is smaller than it sounds, and the crowding-out effects are less than feared. For investors, this means that trading in AI themes may need to shift from a "macro narrative" to "micro verification," focusing on the capital returns of specific companies rather than looking broadly at "AI is changing everything."

Disclaimer

This article is a compilation and interpretation of a third-party brokerage research report (Goldman Sachs, August 10, 2026) by Trend Research, along with the整理 of publicly available market information. The ratings, target prices, earnings forecasts, and related judgments cited in the article represent the opinions of the analysts from the brokerage and do not necessarily reflect the views of Trend Research, nor do they constitute any investment advice.

The market carries risks, and decisions should be made independently. This article should not be used as a basis for buying or selling any securities.