Does the market need to be concerned about AI adoption?

Good morning. President Donald Trump lashed out at Fed chair Jay Powell yesterday after Powell emphasised the risks posed by tariffs in a speech on Wednesday. Before Trump, it would have been unusual, and even alarming, for a president to openly rail against the Fed chair. But the market seems to be well prepared for such skirmishes: the S&P 500 was flat yesterday, and yields on the 10-year Treasury only ticked up 5 basis points.

Unhedged will be off for Easter Monday, but back in your inboxes on Tuesday. Email me: aiden.reiter@ft.com.

AI adoption

Big Tech has had a rough year: the Magnificent 7 stocks are down 22 per cent, and semiconductor stocks have taken a beating. With the exception of the January wobble caused by Chinese AI upstart DeepSeek, this seems to be more about market volatility than dents in the AI narrative.

Still, the threat of slowing adoption is worth watching. Sam Tombs at Pantheon Macroeconomics points out that according to the recent regional surveys from the Federal Reserve, services businesses expect to dial back on IT and capital expenditure, having already cut spending in prior months (chart from Tombs; the capex intentions trend is expressed as an average of standard deviations relative to its 2015-2024 mean):

The most plausible justification for this is fear of slower economic growth. Most companies have not found a use case for AI yet, and the best models (ChatGPT, Gemini) have free versions. If you are the IT manager at a medium-sized company, with a potential recession looming, do you really want to approve a big AI line item?

Another explanation is that this could just be the new technology lifecycle in action. Other historical tech adoptions have also had moments of market underperformance and lower uptake, says Joseph Davis, chief economist at Vanguard and author of an upcoming book about tech cycles:

It’s not always a straight line — there are hiccups along the way . . . In every cycle, the tech sector underperforms for a significant period. [The] market underestimates new entrants, while [companies] ask: why are we putting money on this tech stack when we can go on cheaper tech stack in the future? We saw this with electricity [and other technologies].

Then there’s DeepSeek. The Chinese company’s low-cost models demonstrated that end users do not necessarily need the best in class, and cheaper offerings may come from smaller players in the not-distant future. That could justify going more slowly on capex and adoption spending now.

Even with falling expectations, many analysts see this as just momentary turbulence, and expect high adoption going forward. Here’s Joseph Briggs at Goldman Sachs:

The real need of AI-related capex from an end user perspective is still seven years off. Just 7 per cent of companies currently report that they are using AI for the regular production of goods and services. While we have seen a pullback in capex expectations more broadly, I would think about this as being separate from the A theme, but rather related to a near-term headwind to investment related to trade policy uncertainty.

Goldman still forecasts $300bn in AI-related investments by the end of 2025. But, as Briggs told me, that number is based on AI-exposed companies’ revenue forecast revisions. As the outlook worsens, AI spending will probably drop, too.

The AI narrative is not dead. The market and business pullbacks look like an example of the familiar non-linearity of growth in new technologies. But if the US enters a multi-quarter recession — and AI customers really start to cool their jets — that could change.

Friday interview: Brent Neiman

Brent Neiman is a professor at the University of Chicago Booth School of Business and recently served as the assistant secretary for international finance in the US Treasury department. Earlier this month, he made headlines when the White House’s ‘reciprocal tariff calculations misleadingly cited research done by Neiman and his colleagues. Unhedged spoke with him about that calculation, the price effects of tariffs and the future of the dollar.

Unhedged: Could you walk us through the research cited by the White House?

Neiman: The paper was written to measure the pass-through of the first Trump administration’s 2018 tariffs into prices. At the time, there was a lot of discussion of how much foreign countries would pay for the tariffs, rather than the US. Theoretically, there’s nothing incoherent about that — it was possible that foreign exporters would reduce their prices to offset any imposed tariff. But it was also possible that there’d be very little change in pricing, forcing US importers or consumers to cover the tariff.

We decided to do an empirical analysis on this question, using data meant to represent the full basket of US imports. We found that US importers paid around 95 per cent of the 2018-19 tariff. For example, if there were a 20 per cent tariff, there would be a one percentage point reduction in the price charged by foreign exporters, and a 19 per cent increase in the prices faced by US importers.

We also looked at the price effects of Chinese retaliatory tariffs against the US. Interestingly, there was not the same effect. We found that US exporters dropped prices by more in response to China’s tariffs than Chinese exporters did in response to US tariffs. So in some sense, US exporters paid a greater share of Chinese tariffs than the Chinese exporters paid of the US tariff.

Finally, we traced it through as best we could to retail prices, using information from two large US retailers. Our research showed that pass-through was actually much lower for the retailers.