AI Drove the Majority of US Business Investment in 2025

Last Friday’s US GDP reading for the fourth quarter was overshadowed by the Supreme Court’s decision to strike down the emergency tariffs imposed by Donald Trump. Yet the data—published nearly a month late because of the government shutdown—shed some light on the strength of the US economy at the end of an especially turbulent year. At first glance, the biggest surprise was softer-than-anticipated real GDP growth External link.. Goods consumption, net exports and federal spending were disappointingly weak. In contrast, business investment was more robust than expected.

Despite numerous headwinds—including heightened uncertainty, higher input costs due to tariffs and persistent supply chain issues—investment held up relatively well. Non-residential business investment increased by 4.2% for 2025, more than the 2.9% recorded in 2024 (though still less than the 7.3% invested in 2023). Can this improvement be credited to Trump administration policies? Attributing this performance to the regulatory, tax and protectionist measures implemented last year would be hard to justify, if not premature External link..

And business investment was far from broad-based in 2025. We also didn’t see much growth in manufacturing, mining and fossil fuels, even though they were heavily supported by President Trump. These industries lost workers in 2025, while their output barely increased or even fell (+0.9% for manufacturing, +0.6% for mining and -6.8% for oil and gas extraction). This weakness was also reflected by the decline in real construction in these sectors. In fact, real construction investment rose in only two categories: data centres and infrastructure for renewable energy—wind, solar, dry waste and geothermal (graph 1). Even construction in the manufacturing sector, which had trended upward sharply from 2022 to 2024, contracted in 2025.

Data centre construction is an obvious consequence of the rise of artificial intelligence (AI). But building these centres is just the first step. They’re useless if they remain empty. Consequently, in 2025, most of the uptick in real equipment investment came from companies purchasing computer hardware (graph 2). The contrast was even more pronounced in the fourth quarter of 2025. Without the surge in real investment in IT equipment, which came to an annualized 80.1%, non-residential investment in equipment would have fallen at an annualized rate of 10.3%.

Finally, computer hardware needs software. The National Economic Accounts categorize software investment as part of the “intellectual property products” component. It’s the same story we’ve already seen in construction and equipment. If it wasn’t for the 11.0% jump in software purchases, private sector investment in intellectual property products would have stalled in 2025 (graph 3). And that doesn’t even take research and development spending into account. Although detailed data for 2025 are not yet available, a large share of this investment is related to IT, which accounted for 27% of this category in 2024.

All of this leads to an obvious conclusion: Much of the strength in business investment growth in the United States in 2025 came from the AI sector. Now, that’s not a bad thing in and of itself. If the momentum lasts, the broader economy will continue to reap the benefits. But the implications of the AI boom don’t end there. There’s a lot that could be said about its impact on the job market, productivity, corporate profits, wealth distribution, energy consumption, natural resource use, and consumer and investment decisions.

We could also ask whether AI mania is crowding out other types of investment. Is AI’s rapid expansion limiting the potential growth of other industries, either by stretching resources (including financial ones) or driving up input costs? Could that be the only reason for the relative weakness in other industries or has non-IT investment dropped off for reasons specific to those sectors?

This may be where factors frequently mentioned over the past year—starting with the uncertainty caused by the Trump administration’s trade policy—come into play. The US government’s about-face on climate policy and the end of support programs introduced under the Biden administration may have also played a role. Manufacturing investment in the automotive sector was weak (-7.8% in 2025), as motor vehicle and parts manufacturers abandoned several projects related to electric transportation.

So What About That $18 Trillion?

Upon closer inspection, private investment seems lacklustre once we take AI out of the equation. That raises questions, especially since President Trump often claims that new investment commitments, particularly from abroad, have never been higher. In this week’s annual State of the Union address to Congress, he declared: “In four long years, the last administration got less than $1 trillion in new investment in the United States. And when I say less, substantially less. In 12 months, I secured commitments for more than $18 trillion pouring in from all over the globe. Think of it—much less than $1 trillion for four years, versus much more than $18 trillion for one year. What a difference a president makes. A short time ago, we were a dead country.”

Unfortunately for Americans, this astronomical figure (equivalent to 58% of the country’s annual nominal GDP and four times its annual nominal non-residential investment) isn’t guaranteed to materialize, nor is it even realistic. The official White House web page External link. tracking these investments lists commitments totalling US$9.7 trillion, which is still an enormous sum. And circling back to our earlier point, roughly 40% of those investments involve AI, data centres or semiconductor manufacturing. It’s hard to predict how and when all that money will be deployed across the US economy. Some of these commitments also appear to be tied to the partial trade deals that the Trump administration negotiated with various partners. Given the wild shifts in US trade policy, it’s far from certain they’ll be honoured. The European Union has already admitted it can’t guarantee that US$600 billion would be invested in the United States by 2028, as stipulated in last August’s agreement. A Peterson Institute for International Economics analysis External link. of foreign pledges to invest in the United States has also raised doubts, especially regarding these countries’ ability to pay, the wide range of time horizons involved, the definition of what counts as an investment and the general lack of clarity in the process. All of this undermines confidence in the likelihood these investments will actually come through. Based on the data available so far, which run to the third quarter of 2025, there’s been no observable acceleration of foreign direct investment in the United States. But it’s still far too early for the country to see any of the benefits of President Trump’s aggressive investment policy, assuming they ever materialize.

In the meantime, the US economy will have to rely on artificial intelligence to keep driving strong gains in private investment. Our forecast scenario calls for continued growth in this area, though at a less feverish pace than in 2025. The outlook could also improve with greater clarity around trade policy and the lasting impact of the tax cuts passed last year. If these factors fail to materialize, business investment in the United States may prove far less robust.