One of the great mysteries of recent history is why tightening monetary policy has failed to significantly reduce economic activity, particularly capital investment. For example, as the Fed funds rate increases, non-residential fixed investment increases in 2022-23.
Figure 1: GDP growth, q/q AR (black line, left scale), contribution from non-residential investment (blue bar), imports of computers, computer accessories, semiconductors (green bar), rest-of-GDP (red bar), Fed funds rate (green line, right scale). 2026Q2 growth rates from Atlanta Fed GDPNow of 7/10; computer purchases in 2026Q2 estimated in 2 months of trade data. Source: BEA, Federal Reserve, Atlanta Fed, and author’s calculations.
This result – driven by AI-related capex growth – has fueled the usual description of the US economy as remarkably resilient. The question is whether this capex boom will continue, and by extension, the growth of the US economy.
In recent quarters, the contribution to growth from non-residential investment has been offset by computer and semiconductor imports – in the sense of counting. In 2026Q2, it looks like the bulk of GDP growth is accounted for by fixed non-residential investment, which is offset by imports, although this is dependent on 2 months of import data that accurately represents overall Q2 imports. That means continued US growth is highly dependent on what happens with AI-related capex.
What has the structure of non-residential fixed investment in AI looked like over the past few years. We do not have a specific allocation of investment in AI, but we can look at the change in investment from 2022Q3 (we take the November 2022 release of ChatGPT as a kind of marker for AI capex).
Figure 2: Change in investment in software (green bar), information assets (red bar), and energy and communication infrastructure (blue bar) from 2022Q3, by bn.Ch.2017$ SAAR. Source: BEA and author’s calculations.
BIS estimates 2026 investment by US hyperscalers and other AI companies at $800 bn, up from around $750 bn in 2025.
Is there any reason to believe that this level of investment will not be achieved? One leading indicator of non-residential investment in the dot-com peak of 2000 was imports of telecom equipment and computer equipment (the latter peak was one quarter ahead of the previous peak). We only have data on information technology investments in 2026Q1, while we have monthly data on imports of computers, computer accessories and semiconductors for May. Connecting the series together, one gets the next picture.
Figure 3: Change in investment in information goods (green line), imports of computer equipment, components and semiconductors (red line) from 2022Q3, in bn.Ch.2017$ SAAR. The 2026Q2 outlook is based on April and May import data. Source: BEA and author’s calculations.
The 2026Q2 outlook is just a projection, based on two months of data, which will be updated over time. It may be the case that imports continue to rise in Q2. However, there is reason to believe that there may be a surprise downside to AI-related capex.
First, Magnificent 7 share prices have fallen, and are below the peak. As the stock market cools, there are (at least) two consequences. First, the cost of capital will increase for AI-related firms (and others). Second, the wealth effects that support consumption will disappear. For AI capex should be slower compared to what would otherwise be the case.
Source: Bloomberg.
Second, investment now exceeds cash flow, so firms must now rely on external financing, that is, through bond markets.
Source: Economist.
Therefore, hyperscalers scale up the financial system. Internal funds (cash flow) were cheap, and not influenced by the market rate. Note that as the corporate bond rate increases, investment will now decrease (as shown in Figure 1 below).
Source: Fazzari et al. (1988) as modified by Chinn.
Access to bond markets means that investment will now be subject to a higher hurdle rate, which is very close to bond yields – and at the same time some firms will have to pay a virtual risk premium. (Below, Microsoft’s bond yield serves as the risk-free rate).
Source: Economist.
So, while AI-related capex seemed immune to high interest rates in the past, this may not be so true going forward. And indeed uncertainty about expected future cash flows may weigh more heavily on investments than ever before. The biggest reason is to reduce policy uncertainty, and to reduce upward pressures on interest rates from suppressive inflation.
