Summary of Key Points
A Goldman Sachs report indicates that although the scale of AI investment in the United States is growing rapidly (reaching $600 billion by 2026, accounting for 2% of GDP), its actual impact on the U.S. economy is much lower than expected due to the heavy reliance on imported AI hardware and the classification of certain investments as “intermediate goods” that are not included in GDP calculations. Currently, there are limited displacing effects of AI investment on other sectors (such as technology, construction, and financing costs). Considering all factors, AI’s contribution to U.S. GDP growth in 2026 is estimated to be only 0.1 percentage points.
I. AI Investment: Big in Scale, but Small in Actual Impact – High Import Dependency is the Key Factor
Many people assume that a large investment in AI ($600 billion) will significantly boost the economy, but in reality, there are several mitigating factors:
- High import proportion: A significant portion of AI purchases consists of imported technology products (such as chips and servers), and the money spent on these imports does not contribute to U.S. domestic GDP.
- Statistical constraints: Some AI-related expenditures are classified as “intermediate inputs” (for example, companies buy AI software to improve efficiency, not as final products for consumers), which are also not included in GDP calculations.
As a result, the report states that the direct contribution of AI to GDP, according to official statistics, is only 0.5%, far lower than the 2% represented by total investment—meaning that out of every $10 invested in AI, only $2.5 actually contributes to U.S. economic growth.
II. Three Limited Displacing Effects
Displacing effects refer to the situation where AI investment takes money and resources that could have been allocated elsewhere. The main areas affected are:
1. Compression of Other Technology Investments
- Actions by Cloud Providers: Companies like Amazon and Microsoft have reduced stock repurchases to fund AI investments, shifting funds towards purchasing AI equipment. Although they are still borrowing to expand, they have not yet significantly cut back on other technology-related spending.
- Corporate AI Spending: 89% of companies’ AI costs account for only 1%-5% of their IT budgets, with two-thirds of the additional funding coming from reductions in other IT expenses (e.g., by canceling software subscriptions). However, this is mostly a replacement of existing services rather than new investments, so the impact on GDP is minimal.
- Estimation: This displacement is estimated to amount to approximately $30 billion in 2026.
2. Dredging Up Construction Resources
- Data Center Development: Data center construction accounts for 9% of private non-residential building activities, with higher profit margins than other types of projects. However, the overall impact is limited due to two factors: a decline in government subsidies for semiconductor manufacturing facilities and a sluggish residential housing market.
- Regional Variations: In some states, data centers account for more than half of local construction spending, but nationwide, the competition for building resources (such as materials and labor) is not intense. Nevertheless, if government subsidies for data center construction decrease in the future, the pressure on other construction projects could increase.
- Estimation: Displacement in this area is estimated to be around $10 billion in the past year, with potential increases under reduced subsidy scenarios.
3. Rising Financing Costs for Other Enterprises
- Increase in AI-related Bonds: AI-related bonds (e.g., for data center financing) make up nearly a quarter of investment-grade bonds. The increased demand has raised interest rates by about 5 basis points over the past year, resulting in higher borrowing costs for other businesses and a reduction in their investment capacity by approximately $10 billion.
III. Overall Impact of AI on GDP
The report calculates the overall impact as follows:
- Direct Contribution: Only 0.1 percentage points according to official statistics due to import and statistical constraints.
- Additional Contributions: If imported investments and AI-related exports are included, the contribution rises to 0.3 percentage points.
- Offsetting Factors: Displacing effects (about $50 billion in 2026), rising energy costs (negative impact on consumption), and increased stock market activity (positive impact on consumption) offset each other, leaving the net contribution at 0.1 percentage points.
In summary, while AI can drive economic growth, factors such as high import dependency and displacing effects significantly reduce its actual impact on GDP.
IV. Future Risks
The potential risks include increased pressure on the construction sector once government subsidies for semiconductor manufacturing facilities are phased out. If these subsidies disappear and data center construction resumes at full pace, the competition for building resources will intensify, potentially affecting other construction projects (such as factories and commercial buildings) more severely. Additionally, if the issuance of AI-related bonds continues to rise, it may lead to further increases in interest rates and increased financing costs for businesses.
Conclusion
The impact of AI investment on the U.S. economy is not as significant as initially thought. Despite the large scale of investment, import dependencies and statistical limitations reduce its actual contribution. While the displacing effects are currently limited, there are potential risks in the future. Overall, AI’s contribution to GDP growth is only 0.1 percentage points. This highlights the importance of considering where the money is being spent and whether it is competing with resources needed for other industries when assessing the economic impact of AI.