Is the US economy as narrow as its stock market?

This article is an on-site version of our Unhedged newsletter. Premium subscribers can sign up to get the newsletter delivered every weekday. Standard subscribers can upgrade to Premium here, or explore all FT newsletters

Good morning. Pimco’s CIO, Dan Ivascyn, told the FT last week that the 10-year Treasury yield could hit 6 per cent. He’s right, if the US economy keeps humming and more hedge funds run for it. Six is a high but not crazy number in the current context. But remember how far we have come. The yield has already moved from 1.5 to 5.2 per cent over the past five years. We are probably much closer to the top than the bottom. Send us your thoughts: unhedged@ft.com.

Is the economy narrow too?

US stocks are hitting new highs but the AI trade is doing all the work while the average stock gets pummelled. Team Unhedged, along with every other market writer and their AI assistant, has written so much about the narrow market that you’re probably sick of hearing about it.

There is, however, a related point that gets less critical attention: the idea that the US economy is also totally dependent on AI. This remains a commonplace. On Saturday, for example, a blog post from the Center for Economic and Policy Research claimed that “The economy has become almost completely dependent on the AI bubble as most other forms of investment are being crowded out.” Is this true? Is the economy as narrow as the stock market?

No*. “No” because a survey of the data shows that the direct impact of the AI investment boom on economic growth is meaningful but not very large. The “no” gets an asterisk because the indirect wealth effect of AI euphoria in the stock market is probably an important contributor to consumption growth.

Starting with direct impacts, Stephen Brown of Capital Economics sent me the chart below, which draws on data from the Bureau of Economic Analysis. It shows the percentage of economic sectors where real value added is growing:

The share of sectors that are expanding, and the share expanding faster than GDP, is right in line with the 20-year average. Inspired by Brown, I did the same with employment growth in the Bureau of Labor Statistics’ 13 broad economic sectors. Here again, the number of sectors where employment is expanding is bang in line with the long-term average. What is more, we are clearly in the midst of a broadening cycle that started last year:

In general, an unemployment rate of a bit over 4 per cent just does not seem consistent with an economy that is totally dependent on a single sector, however large and important.

The great motor of the US economy is still the consumer, not AI. Personal consumption expenditures contributed 2.5 percentage points to real GDP growth on an annualised basis in the second quarter, more than 100 per cent of total growth. That is almost three times the contribution of all private investment, to say nothing of AI investment. Real retail sales grew at 2.6 per cent in August and have been on an accelerating trend all year. Timely data from Bank of America’s debit and credit card businesses confirms this picture. Nearly 86 per cent of spending categories are expanding, according to BofA, and spending is growing nicely in all income cohorts, undercutting the view that only the rich, with their glittering mountains of financial assets, are the only class driving aggregate consumption higher.

As I have noted recently, there have been a significant number of negative signals about the consumer in the earnings reports of retailers and staple goods producers. How much this is down to matter of industry struggles, a K-shaped economy, or the shortcomings of government economic reports is not easy to tell. But the headline numbers clearly say that in aggregate the consumer remains the US’s key economic motor.

The (admittedly incomplete) direct data we have on AI spending confirms that it is not a big driver of GDP. Last week, for example, the Census Bureau released its construction spending report for September. AI data centre construction spending is now running at an annualised rate of $85bn a year, $36bn higher than at the same time last year. That’s fast growth, but it is still the equivalent of about a quarter of a per cent of GDP and less than 2 per cent of GDP growth.

Data centre construction is not the whole of the AI investment boom — there is also software spending, power and infrastructure upgrades and so on. One Fed estimate puts the contribution of the boom to GDP growth at roughly 0.6 percentage points a year, on average, in the year that ended in the first quarter. That’s meaningful — perhaps a quarter of total growth — but not by any means the difference between the current economy and recession.

All that said, AI could grow more important from here. If recent estimates from Stijn Van Nieuwerburgh of Columbia Business School are right, current plans suggest that the AI build-out could become larger from here — to a total of $10.3tn dollars between 2025 to 2032, or the equivalent of about 3.6 per cent of GDP a year, bigger than the railroad boom (remember, though, that because the investment is equivalent to 3.6 per cent of GDP by size, this does not mean it contributes 3.6 per cent of GDP; much of the investment goes to imported chips and equipment, the cost of which is netted out of the GDP equation).

The crucial indirect effect of the AI boom on GDP growth is the wealth effect. Since the launch of ChatGPT in late 2022, US households’ shareholdings and defined contribution pension assets have grown by $34tn, according to the Fed’s distributional accounts. The Fed now puts equity ownership at 33 per cent of total household assets, far and away the all-time high (it was 27 per cent before the dotcom bubble burst).

This is undoubtedly supporting, to a greater or larger degree, the strong consumer spending detailed above. One model, from Sixtine Bigot and Aurélien Espic of the Bank of France, estimates that the wealth effect explains about 40 per cent of the growth in US consumption in 2025. If that is right, the absence of this wealth effect would not be enough to drive the US into recession. But if the effect reversed — if the AI bubble burst and created a “loss of wealth effect” that dragged on consumption — it could. In that limited sense, the US economy is dependent on AI, but on AI euphoria, rather than AI investment.

One good read

An editor.

FT Unhedged podcast

Can’t get enough of Unhedged? Keep following Rob Armstrong and Katie Martin’s incisive and unfiltered analysis on the Unhedged podcast, a twice-weekly, 15-minute lively dive into the latest markets news and financial headlines. New episodes every Tuesday and Thursday.

Recommended newsletters for you

Due Diligence — Top stories from the world of corporate finance. Sign up

The AI Shift — John Burn-Murdoch and Sarah O’Connor dive into how AI is transforming the world of work. Sign up