Will Artificial Intelligence Do More Harm Than Good for U.S. Growth?

Rebecca Patterson is a senior fellow at the Council on Foreign Relations, a globally recognized investor, and macroeconomic researcher. Ishaan Thakker is a research associate for geoeconomics at the Council.

A discussion of the broad adoption of artificial intelligence (AI) often prompts forecasts of longer-term U.S. Goldilocks-type growth. The idea of artificial general intelligence (AGI)—the theory that AI will eventually understand, reason, and problem solve across different domains—only extends these conversations further. Typically, this dialogue surrounds assessments and scenarios where workers become significantly more productive, boosting gross domestic product (GDP) without lifting inflation.

For now, there is little visibility around when and to what degree such AI-driven benefits would emerge. Indeed, research published in August [PDF] by the Massachusetts Institute of Technology found that despite massive investments, 95 percent of U.S. businesses that had launched generative AI pilot programs had not yet seen tangible business benefits.

What is already clear, however, is that AI is contributing to U.S. growth in at least two major ways: massive capital expenditures and wealth creation. At the same time, it is creating what could become a meaningful drag—job cuts.

How these offsetting factors evolve in the year ahead will heavily influence U.S. consumption and the broader economy, directly and indirectly, through monetary policy. Barring a rapid, unexpected AI breakthrough—or significant, growth-supportive policy out of Washington—risks appear skewed so that Bloomberg consensus estimates of 1.7 percent U.S. GDP growth for 2026 will be disappointed.

More AI-Driven Job Losses Will Mean a More Cautious Consumer

The biggest challenge to U.S. economic growth in the year ahead is the labor market. Incomes provide consumers with the means to spend, and consumption accounts for more than two-thirds of overall GDP. With fewer jobs, a subsequent reduction in spending slows overall growth.

While recent months have seen signs that the labor market is cooling, there is little evidence so far that AI is to blame. The July monthly report on U.S. employment cuts from job placement firm Challenger, Gray & Christmas showed that AI was directly related to 10,375 layoff announcements so far this year. That’s only 1.3 percent of the total job cuts announced through July 2025. This is notable given that 2025 has seen the most overall layoffs announced since 2020 (there were a total of 806,383 cuts from January to July).

Even when the definition of AI layoffs is broadened to the entire technology sector, the number appears relatively modest. While tech is the leading private-sector industry for layoffs—nearly ninety thousand through July, up 36 percent from the same period a year earlier—the pace of layoffs remains well below what occurred in 2023. It’s also noteworthy that the tech sector layoffs pale in comparison to the federal government’s so far this year.

There is reason to believe that AI will, directly and indirectly, start displacing a more significant number of jobs in the months ahead. Consider the position of C-suite executives today. They face pressure both internally and from stakeholders to find efficiencies through AI adoption, which means new costs for tech integration and staff training. At the same time, a highly uncertain, moderating economy requires tight budget controls. One way to thread this corporate-planning needle is by slowing hiring or incrementally reducing headcount—both where AI can be a cheaper substitute and a way to manage overall costs.

This scenario was captured by an August quarterly survey of CEOs by Conference Board, a global nonprofit business membership and research organization. More than one-third of respondents said they expect their workforce to shrink over the next twelve months, which is more than the share expecting their workforce to increase for the first time since 2020. CEOs also leaned more toward reducing than expanding investment over the coming year.

The survey also showed that CEOs see AI both as a tool to manage costs and as a business risk, like cybersecurity and tariffs. Indeed, 93 percent of respondents said they are relying on technology to increase productivity.

Greater AI-driven job losses are also being predicted by a number of tech company executives. Anthropic CEO Dario Amodei said in a May Axios interview that AI could eliminate half of all entry-level, white-collar jobs within the next five years and cause the unemployment rate, now just over 4 percent, to soar to somewhere between 10 and 20 percent.

An increasingly cautious corporate mindset could easily set a negative feedback loop in motion in the quarters ahead that leads to a consumer pullback and slower GDP growth. The question now seems less about whether AI will negatively affect the labor market and more about by how much.

AI’s Upside: Wealth Creation and Significant “Private-Sector Stimulus”

Fortunately for today’s AI-focused economy, jobs are only one factor affecting the growth outlook. There are two major economic supports coming from AI: extraordinary capital expenditures as companies invest in future innovation, and wealth creation from soaring company share prices that then facilitates consumption.

Especially since ChatGPT’s launch in November 2022, optimism that AI will become a transformative technology like the internet has driven incredibly strong demand for related companies’ equities—even as valuations have grown increasingly rich relative to both historical norms and the rest of the equity market. In the first quarter of 2023 alone, AI hopes fueled by ChatGPT’s rapid adoption helped lift Nvidia shares by 90 percent, Facebook parent company Meta by 76 percent, and Advanced Micro Devices by nearly 51 percent.

Companies driving AI’s fortunes have seen their shares continue to gain in value, with investors reluctant to reduce exposures by much in advance of another possible “ChatGPT-like” moment, perhaps when AGI emerges. Chipmaker Nvidia and the top tech hyperscalers—large service providers offering scalable AI infrastructure like Amazon, Alphabet (parent of Google), Meta, Microsoft—have collectively gained over $10.3 trillion in market value since the fourth quarter of 2022. Just looking at 2025 through July 30, they created nearly $2.25 trillion in wealth and accounted for more than 80 percent of the S&P 500 equity index’s gains.

This market rally, and expectations that it will continue, have provided a measure of support to household confidence. The Conference Board’s July survey showed that 48 percent of respondents expect U.S. stock prices to rise over the next twelve months, up from 38 percent in April. Increased confidence often translates to more consumer spending. Using conservative wealth effect estimates from the National Bureau of Economic Research and the Federal Reserve [PDF], the $2.25 trillion in gained wealth implies additional consumer spending of $63 billion to $112 billion over the next one to two years.

At the same time, in what feels like a race among leading hyperscalers to reach AGI first, much of these companies’ free cash flow is being reinvested into their businesses.

The top four tech hyperscalers—Amazon, Google, Microsoft, and Meta—had record capital expenditures of $212 billion in 2024 and are on track to spend around $361 billion in fiscal year (FY) 2025. Google and Microsoft have indicated in recent earnings calls that this spending trend, which is focused on building AI infrastructure, should continue into 2026. Meanwhile, Amazon and Meta have hinted at annualized spending of nearly $100 billion.

The degree of this “private-sector stimulus” is easier to appreciate when viewed against other public-sector stimulus efforts and historical analogies. Consider the American Recovery and Reinvestment Act of 2009, which passed in the wake of the 2008–09 global financial crisis. It directed about $420 billion to be spent in FY 2010. That’s only slightly above the annual pace for AI capital expenditures today. The top four hyperscalers alone are expected to total more than $1 trillion from 2020 to 2025.

Major past innovation also helps put the scale of AI spending today into context. Estimated capital spending for FY 2025 constitutes around 1.2 percent of U.S. GDP. That is comparable to the 2.5 percent of GDP that railroad investment [PDF] contributed to the U.S. economy from 1828 to 1860, or the 1.2 percent GDP contribution from telecommunications infrastructure investment [PDF] in 2000.

While the benefits of this AI spending do not filter across the entire economy, they are still large enough to meaningfully boost growth. The contribution to GDP between the fourth quarter of 2024 and the end of June 2025 from data center-related investments—key infrastructure needed to support AI—slightly exceeded the contribution from consumer spending. Put another way, the U.S. economy would be growing at a much slower rate today without AI’s influence.

This growth flows back to equities as well. More bullish expectations for the U.S. economy, driven in large part by AI investments, generally help make U.S. equities more attractive to investors.

A New Challenge: The Sustainability of AI’s Equity Boom

The outsized market gains and future expectations for these already mammoth companies, as well as the speed and scale of capital deployment, are prompting debates about how much further the AI equity rally and spending boom can continue. There are at least two clear risks: Macro-economic catalysts that weigh on U.S. equities broadly and physical constraints on future AI capital expenditures.

Risks to U.S. equities are plentiful. There is, for example, the potential for the Federal Reserve to undershoot market rate-cut expectations or for disappointment about U.S. AI technology advances to spike at home—perhaps driven by a comparison to China’s success. While risks like these are always a possibility, they become more relevant when U.S. equity ownership is elevated and valuations are rich relative to history.

The challenge for the economy is what happens when a catalyst causes equities to fall. With more than 60 percent of Americans owning U.S. equities, a meaningful, sustained market pullback can quickly weigh on confidence and spending.

This pattern was recently illustrated between 2022 and 2023. As the COVID-19 pandemic and Russia’s invasion of Ukraine contributed to sharply rising inflation, the Fed reacted with an historically aggressive monetary tightening cycle, hiking policy rates by 5 percentage points between March 2022 and July 2023. Before ChatGPT’s November 2022 launch provided critical market support near the end of the tightening cycle, U.S. equities fell by 9 percent. Notably, these losses were led by tech stocks. Nvidia, for example, dropped by 36 percent over the same period.

Not surprisingly, the combination of higher interest rates and lower share prices left consumers—and businesses including AI-focused firms—feeling more cautious. The pace of U.S. retail sales slowed markedly and investments by hyperscalers declined modestly. Both factors served to weigh on growth, all else equal.

Fast forward to today, the Fed faces a different dilemma: Maintain restrictive policy rates to return inflation to its 2 percent target, or  ease policy to limit the labor market slowdown—at the risk of sustained, higher inflation. A more hawkish-than-expected Fed would create a potential catalyst that could pull equities, including those focused on AI, lower. So too could White House decisions that undermine the Fed’s independence, potentially pushing long-term U.S. bond yields higher as investors demand more “risk premium” to lend the government money.

Another risk to watch, for equity sentiment as well as the AI firms themselves, is the ability to build sufficient infrastructure quickly enough to meet investor expectations.   

Even with ample cash, hyperscalers face supply constraints—including energy, materials like transmission cables and critical minerals, and data centers. Data centers are especially important to the AI buildout because they provide computing power and storage capacity to train and operate advanced AI models. However, they are becoming more expensive to build, as the supply of real estate, power access, and equipment such as transformers struggles to keep up with demand. Existing and upcoming tariffs on aluminum, copper, semiconductors, and steel are imposing costs on AI infrastructure providers and raw material suppliers.

Today’s push to build data centers creates a related risk to watch in the year ahead: electricity price inflation. The U.S. Energy Information Administration estimates that average household electricity prices rose 6.5 percent between May 2024 and May 2025, with greater electricity inflation in parts of the country where there are more data centers, such as Virginia and Ohio. Meanwhile, a report by Monitoring Analytics, the independent market watchdog for the mid-Atlantic grid, attributed nearly 70 percent [PDF] of increased energy costs from the last year, or $9.33 billion, to data center demand. According to research and consulting firm Wood Mackenzie, proposed data centers have increased their electricity usage requirements from 50 to 134 gigawatts since last year. (One gigawatt can power 1.1 million homes for an hour.) This far outstrips the sixty-four gigawatts that U.S. utilities have publicly committed to supplying. 

Given households’ greater sensitivity to inflation since the pandemic, a continuation of electricity price increases could result in a push for hyperscalers to bear more of these costs. Greater costs and physical constraints could combine to slow AI capital expenditures and, in turn, progress. Rising electricity prices could also feed into broader household inflation expectations, something the Federal Reserve is carefully watching.

Putting AI Influences in Context

Even if AI capital spending remains robust in 2026, the potential for more job losses suggests risks for the U.S. consumer. By itself, that doesn’t guarantee a slower economy. Other influences, such as spillovers from tariffs, deregulation, and fiscal and monetary policy, will matter as well.

What could we expect from those variables? Economists at the nonpartisan Yale Budget Lab see tariffs offsetting near-term growth benefits from this year’s fiscal package, except for the top decile of consumers. Meanwhile, deregulation should be positive for growth by reducing frictional business costs—though there isn’t yet clarity around when or to what degree such benefits may accrue.

What appears most notable in terms of economic risk heading into next year is monetary policy—both the potential for expected easing to disappoint and the possibility of a politicized Fed resulting in a greater risk premium in government bonds.

After Fed Chair Jerome Powell suggested at the August central banking conference in Jackson Hole, Wyoming, that a slowing labor market could require easier policy, financial markets discounted that the bank would cut policy rates by more than 125 basis points by the end of 2026.

Such aggressive easing will likely require more evidence of a labor slowdown, which seems likely. But it would also require further disinflation, and that remains questionable. The Fed’s preferred inflation measure, core personal consumption expenditures (PCE), remains closer to 3 percent rather than the bank’s 2 percent target. This is at a time when a growing number of U.S. firms suggest they will be raising goods prices in the coming months in response to tariffs and consumers’ elevated long-term inflation expectations. If inflation remains stubbornly high or rises further, policy rate cuts may disappoint expectations. This in turn would weigh on equity sentiment and potentially consumption.

The other monetary policy risk comes from easing that is seen as politically rather than economically driven. This has become a greater risk as the Donald Trump administration pushes to replace more Federal Reserve policymakers. In such a scenario, easing could push up long-term U.S. borrowing costs as investors position for higher levels of sustained inflation.

Despite AI’s Potential, Year-Ahead Seems Biased to Downside Growth Risks

AI is contributing to an incredibly wide range of potential longer-term economic outcomes. The proposed aftermath of broad AI adoption ranges from productivity booms that leave large swaths of the population out of work, to incremental innovation and role augmentation, to the stuff of dark sci-fi movies.

Looking ahead to 2026, there is relatively greater visibility around some immediate outcomes, though unexpected policy changes and the timing of AI breakthroughs make even short-term forecasts uncertain.

AI capital expenditures should continue to provide a notable measure of support to GDP growth as companies race to achieve AGI. However, physical supply constraints may temper the scale and speed of that spending, as could a possible reversal in equity markets. At the same time, greater job losses are likely to weigh on sentiment and consumption, while offsetting rate-cut support is not assured given potentially sticky inflation. Taken together, the risk to consensus expectations for 2026 economic growth seems biased lower, with the consumer pullback dominating.

This work represents the views and opinions solely of the authors. The Council on Foreign Relations is an independent, nonpartisan membership organization, think tank, and publisher, and takes no institutional positions on matters of policy.