Investment in AI strengthens US economy while investors remain concerned
- On March 10, 2000, the NASDAQ index of US technology stocks peaked at a value of 5,047. By May 26, 2000, it had fallen to 3,205—a decline of 36% in two and a half months. It was the start of the collapse of the dot-com bubble. And yet, in the second quarter of 2000, US real GDP grew at a stunning annualized rate of 7.5%. It would be the last quarter of strong growth for a while. The collapse of the tech-equity boom ultimately led to a modest recession.
Why am I discussing ancient history? The reason is that the current economic and financial situation in the United States bears some resemblance to the situation a quarter of a century ago. Last week, shares of one of the biggest hyperscalers fell sharply, wiping out half a trillion dollars in valuation. The technology-heavy NASDAQ fell 1.4% in response.
Yields on tech-company bonds have risen and the cost of insuring those bonds against default has soared. And yet, the US government, last week, released data showing that real GDP grew at an annualized rate of 3.4% in the third quarter. The point is that strong GDP growth, fueled by strong investment about a new technology, can take place at the same time as a major market correction that could ultimately contribute to an economic slowdown. The US economy is thriving in large part because of AI investments. Yet, many investors are increasingly concerned about the sustainability of this level of investment. Let’s look at the details:
In the third quarter of 2026, US real GDP grew at an annualized rate of 3.4%—the best performance since the third quarter of 2025. Growth was fueled by a combination of massive investment in technology and strong growth in consumer spending. Real (inflation-adjusted) consumer spending grew at a rate of 3.8%, including a 4.1% increase in spending on services. Within the latter, the biggest contributor to the increase was spending on healthcare. Of the 3.4 percentage points of GDP growth, 2.5 percentage points were attributable to increased consumer spending. The strength of the consumer came despite a decline in real wages as income gains failed to keep pace with higher inflation.
Meanwhile, the star performer was nonresidential fixed investment, which grew at a rate of 12.7% in the third quarter. This included a 20.9% increase in spending on business equipment. In fact, spending on equipment accounted for 1.16 percentage points of GDP growth. Within that category, spending on information technology equipment, which includes computers and related equipment, accounted for 0.86 percentage points of GDP growth. In addition, spending on software accounted for 0.23 percentage points of GDP growth. Thus, combined spending on information technology equipment and software accounted for one-third of the increase in real GDP.
On the other hand, much of that equipment was imported. In fact, real imports of goods grew at an annual rate of 26.4%, subtracting 2.6 percentage points from GDP growth. This demonstrates how the AI buildout is having a strong, positive impact on many US trading partners. Separately, we have seen strong export growth in such countries as Japan, China, Taiwan, and South Korea, partly reflecting global or US demand for AI-related products.
Also, not all categories of GDP were strong in the third quarter. Residential investment fell at a rate of 3.3%, spending by state and local governments fell 0.1%, while exports of goods fell at a rate of 7.4%. And investment in transportation equipment fell as well. Interestingly, real nondefense federal government spending grew at an astonishing rate of 22.2%. Yet, this was mostly due to a winding down of government sales of crude oil from the Strategic Petroleum Reserve.
Finally, there was a sizable increase in private inventories, accounting for 1.6 percentage points of GDP growth. Thus, absent inventory accumulation, real GDP grew at a rate of 1.8%, a majority of which was due to investment in AI.
The US government also released data on consumer income and spending in September, including data on the Federal Reserve’s favorite measure of inflation. The data indicates that, in September, inflation-adjusted disposable personal income (after taxes) fell 0.3% from the previous month. Meanwhile, real consumer spending was unchanged from the previous month. This was a big change from August when spending increased sharply while income was flat. Thus, it appears that consumer spending was starting to slow down by the end of the third quarter.
In addition, the government reported that, in September, the personal consumption expenditure deflator (PCE-deflator), which is the Fed’s preferred measure of inflation, was up 3.6% from a year earlier, the highest since May. When volatile food and energy prices are excluded, the core PCE-deflator was up 3.1% from a year earlier—not very different from core inflation in recent months. As such, underlying inflation remains elevated but is not accelerating. Energy prices were up 21.3% from a year earlier, contributing to a weakening of consumer purchasing power.
Getting back to the financial markets: the sharp decline in equity prices came as questions were being raised about the ability of AI companies to generate sufficient revenue to service their debt obligations. The Bank for International Settlements (BIS), which serves the world’s central banks, wrote that there is a “risk of firms over-committing resources to investment projects with still uncertain returns, leaving all firms vulnerable to disappointments in AI payoffs.” It added that “disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust.”
Finally, Stijn Van Nieuwerburgh, a professor at Columbia Business School, has estimated that average annual capital expenditures on AI in the period from 2025 to 2032 will be 3.6% of GDP, based on current plans. This far exceeds investment booms in past episodes in which new technologies were rolled out. He also estimated that, during the dot-com boom of 1996 to 2003, capital expenditures on telecom and fiber averaged 1.1% of annual GDP. Going back further, the railway boom of 1870 to 1890 involved annual capital expenditures equivalent to 2.24% of GDP. Thus, the AI boom is historically large. Moreover, it is worth noting that the dot-com and railway boom were both interrupted by the collapse of a financial bubble due to excess investment, the failure of revenue to keep pace with capital expenditures, and a sudden reversal of investor sentiment. The temporary collapse of the railway boom in 1873 led to a very deep recession.
