American inequality and the future of AI
Michael S Barr is a Member of the Board of Governors of the Federal Reserve System
In a speech earlier this year, I explored scenarios in which artificial intelligence could affect economic inequality, for better or for worse1. My focus in those remarks was on the AI side of this relationship, and today I would like to return to this topic by first providing some important context about inequality in the United States.
Inequality is an aspect of every economy, and understanding the extent and evolution of inequality in the United States can help in forming a clear view of the economic outlook and how the economy is likely to react to new developments. Widening inequality may also, depending on many factors, reflect a decline in economic opportunity for many people, which can limit the economy’s productive capacity.
I will begin by summarizing the current state of economic inequality in the United States, assisted by new results from a long-running survey of households conducted by the Federal Reserve, and then show where this fits into the history of inequality over the past century or so. I will explore some of the forces that have shaped inequality, and what these forces may indicate about how AI is likely to affect the way wealth and income are distributed.
Wealth and income inequality today
Up-to-date and detailed information about inequality in the United States comes from the latest edition of the Federal Reserve’s triennial Survey of Consumer Finances, conducted in 2025 and published just last month2. The survey confirms that both wealth and income are highly concentrated and that inequality is greater for wealth. The top 10 percent by income of households receive almost one-half of all income in the United States and the top 1 percent receive about one-fifth of income3.
As for wealth, the wealthiest 10 percent of households hold nearly three-fourths of all wealth, and the top 1 percent hold about one-third. By contrast, the entire lower half of households, as measured by wealth holdings, have only 2 percent of wealth. Low wealth means it’s also harder to get by every day. Another Fed survey of household finances conducted each year, the Survey of Household Economics and Decisionmaking, finds that 31 percent of respondents in 2025 didn’t have the cash to cover an emergency expense of $500 or more4.
Wealth is even more highly concentrated among the very wealthiest. One-third of wealth is held by the upper 1 percent—that’s nearly $60 trillion. There is even more concentration at the very top. The households in the upper 0.1 percent, which is only 135,000 individuals and families, own one-sixth of our nation’s wealth.
A household at the lower edge of the top 1 percent has 75 times as much wealth as the median household in the middle of the distribution—$16.3 million compared with $215,000 for a household in the middle. A household at the lower edge of the 0.1 percent of the wealthiest has 300 times as much wealth as the median household.
Wealth concentration has increased at the very top. Using Forbes magazine’s tally of the 400 wealthiest individuals, together with data from the Survey of Consumer Finances, a rough estimate is that their share of total wealth rose from 3 percent in 2022 to 4 percent in the most recent survey. Forbes counts nearly 1,000 US billionaires5.
Digging into the sources of widening inequality, the wealth of those at the top is primarily held in the form of business assets. Research using data from the Survey of Consumer Finances back to the 1940s shows that returns on business equity have been a major source of widening wealth inequality over the decades6.
Most of the income for these wealthy households is in capital income—interest, dividends, capital gains, and business profits—rather than wages and salaries. Unlike these higher-income or wealthier households, most income for middle-class households comes from their paychecks. And it’s harder to build wealth that way.
Although the largest asset class for wealthy households is typically stocks and other financial assets, a substantial amount of their capital income also comes from ownership of private businesses. These private businesses come in two broad types: an actively managed business—such as a plumbing business, retail store, or manufacturing firm—and a passive investment in a business, commonly known as private equity.
Owning and running a business has long been a way that households become wealthy. But private equity investments are passively run and are typically only open to the wealthiest individuals and families. These private investments have delivered returns to their owners higher than the typical mix of assets held by middle-income households. Private equity investments are 12 percent of the assets of the wealthiest 0.1 percent of households in the most recent Fed survey.
For households above the median but below the 90th percentile, the most prevalent assets are housing and stocks, but the value of housing equity has grown more slowly than that of business equity. About three-fourths of households in this group have stock holdings, mostly through retirement accounts.
Although ownership of publicly traded companies has become more widespread across the income spectrum as 401(k) plans have become common over the past few decades, stock ownership by value remains highly concentrated among the wealthiest households. The group of households in the 50th through the 90th percentiles of wealth hold only 17 percent of all public stocks by value, underscoring how much wealth is concentrated7.
Finally, on net, households in the bottom half of the wealth distribution hold just over 2 percent of all wealth, an average of $52,000. About 40 percent of them have equity in homes, and nearly 40 percent own business equity, either through a retirement account or as their own business. But the values are low compared with the upper half of households.
The lesson from history is that the level of inequality in American society is not immutable—it changes over time. These historical patterns suggest that policy choices play a crucial role in shaping inequality levels
The evolution of inequality
By many estimates, today’s inequality levels are among the highest in the past 100 years8. But wealth and income inequality in America have changed significantly over our history, widening and, at times, also narrowing, driven by a variety of factors, including booms and busts in financial assets, structural changes in the economy, and changes in tax and regulatory policy.
In the Gilded Age, roughly 1870 to 1910, incomes rose strongly for most people, but it was also the era when large fortunes were accumulated by industrialists and their families. With no income tax or estate tax for the wealthiest and limited public education for everyone, the gap between those at the top and most others widened9.
But over the subsequent decades, especially during the New Deal, legal and institutional change strengthened worker power, and the introduction and broadening of progressive income and estate taxes helped limit the concentration of income and wealth.
Inequality narrowed in the decades after World War II, an era of fast economic expansion that benefited many households. There were improvements in public education, including the GI Bill, income tax changes that made rates more progressive, and growing unionization that raised wages for workers10.
Beginning in the late 1970s, however, wealth and income inequality began to increase again. This was a period of slower economic growth and sometimes high inflation. Top income tax rates declined, and fewer households were required to pay the estate tax. As in the 1910s and 1920s, business equity grew as a share of total assets, helping the wealthiest households become wealthier11.
The wealth generated from the typical middle-class asset—housing—increased but at a slower pace than business assets, exacerbated by the decline in housing wealth during the Global Financial Crisis. Over the same period, real wages grew slowly. Moreover, minimum wage laws failed to keep up with inflation and the growth of the economy12.
The lesson from this history is that the level of inequality in American society is not immutable—it changes over time. These historical patterns suggest that policy choices play a crucial role in shaping inequality levels. Economic growth, technological change, education, and globalization also play key roles13.
The evolution of the labour and capital shares of national income
Some of these trends are most clearly seen in the decline in the share of national income going to labour over the past 25 years. Though individuals and families across the spectrum hold business equity from which they receive income, the main source of income for most households is wages and salaries—labour earnings.
Since lower- and middle-income households depend heavily on wages, another useful way of exploring income and wealth inequality is to consider how the share of national income earned from labour has changed over time, compared with income earned from capital—dividends, interest, and private business earnings.
After World War II, the labour share of national income was fairly steady between 60 percent and 65 percent, but it began falling in fits and starts beginning in 2001. The decline accelerated after the pandemic and has been especially steep in the past year, reaching a record low of 52.8 percent in the second quarter of 202614.
Some of this reflects measurement issues associated with how income has been allocated to labour and capital over time15. But, even after adjusting for those factors, there is a clear downward trend.
Economic research suggests three causes of the decades-long trend in labour versus capital income, each highly relevant to the potential effects of AI on inequality. The first is technological innovations that have reduced demand for labour and decreased wages; the second is the concentration of market power in many industries with a small number of firms; and the third related cause is the reduction in the bargaining power of workers.
Technology replacing labour has been going on for centuries, but new technology can also augment the productivity of existing jobs and create new jobs, including those that are hard to envision when the technology is initially adopted, a point I have noted in a number of speeches16.
The internet and the businesses it created reduced employment in some occupations—such as travel agents—and eliminated others: think video store workers. But the development of new technologies has expanded jobs in other areas. Consider that there are millions of full-time social media creators earning a living today in the United States, something unimaginable a decade ago17.
What matters is how returns to labour compare with those to capital from the adoption of new technology. The available evidence from across the globe suggests that advances in digital technology have lowered the price of investment in technology relative to labour18. If capital is cheaper relative to labour, firms use relatively more of it. Capital can substitute for labour, too—for example, through the adoption of automation and robots19.
Some evidence shows that firms that adopt robots may also become more productive and expand their labour force. But in this expansion, labour is outpaced by the expansion in capital, seen in rising profits and equity appreciation, which widens inequality20. It will be important to watch these developments closely given the immense capital investment going into the AI sector.
Another factor driving the decline in the labour share of national income is market concentration, which researchers estimate has increased significantly in three-quarters of US industries since 2000. Since that time, wages have grown at a slower rate than labour productivity21. Since larger firms are often more productive, these new technologies, in turn, can contribute to market consolidation, with so-called ‘superstar’ firms as the result22.
Superstar firms also have lower labour shares of income because they can establish higher product markups—the difference between their costs and what they can sell their product for—because there may be less price competition and fewer alternatives for customers23. At the same time, even with higher markups, their ability to deploy new technologies means these firms may be able to charge lower prices than before, which could serve to further increase their size and may result in lower costs for consumers.
These forces—technological change, market concentration, and declining bargaining power for workers—may have contributed to the declining share of labour income over time24. All these changes have tended to favour those with the highest incomes and wealth, driving greater wealth inequality.
How will AI affect the labour share of income?
The discussion of superstar firms probably brought to mind the handful of companies that are dominating AI: the hyperscalers and frontier AI labs. We have seen tremendous investment in the AI sector, with announced plans suggesting that capital deepening in this sector will escalate further. AI could be what some researchers call capital-biased innovation, in which the returns go primarily to capital rather than to labor25.
In previous remarks, I have explored scenarios for how AI could evolve as a technology and how it could affect labour markets. Key questions include the extent to which AI will substitute for human workers—or largely complement them—and the degree to which AI will create new jobs. Market competition and the bargaining power of workers will also affect the labour share of income.
The nature of AI, particularly its ability to learn how to complete complex, nonroutine tasks that require extensive knowledge difficult for humans to achieve, raises the question of whether AI’s effects on the labour market will be stronger than other technologies26. In previous speeches I have explored whether labour displacement could be larger than in the past, and today I consider the risk that the benefits of this technological revolution could be more narrowly shared than those that came before27.
There are questions also about the concentration of market power. Competition and distributed innovation could make AI a cheap and ubiquitous commodity, with gains widely shared. In this case, most businesses would have access to AI resources sufficient to support improved productivity and further innovation. AI could democratize access to skills and information needed for small businesses to thrive.
But some researchers believe that AI has characteristics that have, in past examples of high-tech advances, reinforced concentration of market power28. AI depends on access to data, model improvements, and computing power, so it benefits from economies of scale and scope. AI itself is a powerful tool to train and accelerate the development of new AI models, including through self-reinforcement learning.
The extent of the advantage possessed by AI leaders may prove to be unprecedented. If a small number of AI firms dominate the market and receive the lion’s share of investment returns, then the wealth of the managers and owners of these firms will surge. Large businesses could afford AI investments and continue to grow, while smaller firms would see declining market share.
In this scenario, the vast majority of businesses and their employees would have limited access to ever-improving AI resources, which means they would see smaller returns. Overall, the labour share of income would decline, market concentration would increase, and inequality would go up.
So far, there has been relatively little displacement of workers by AI and some significant evidence that the transition to AI is enhancing the productivity of some workers, increasing demand in some sectors, and even creating new occupations. Vigorous competition among the AI firms so far has meant that access to AI remains fairly broad.
Some believe that many businesses will find that most of their AI needs can be fulfilled by smaller-scale language models that will not need to be on the cutting edge of technology. And in that case, benefits would be more broadly shared.
As fast as AI has grown, its implementation is still limited so far, and no one knows the degree to which its benefits will be shared. We are already seeing signs that AI may be a labour substitute in some sectors, especially for younger, less-experienced workers.
Some tasks that are easily automated with clear guardrails and predictable outcomes might see rapid labour substitution, while other tasks that require human judgment, management, coordination and relationships, creativity, or outputs that are hard to measure might see more labour augmentation29.
If labour market changes happen quickly, it will be hard for workers to adjust, and dislocations might be large, whereas a more gradual adoption might permit more orderly adjustments.
That said, we should be prepared for the possibility of serious short-term disruptions in the labour market that need to be managed effectively to ensure the benefits of AI are realized over the long term. If AI proves to be capable of work that displaces humans, the extent of the disruption to the labour market will depend in part on whether society undertakes the investments needed in new job creation and worker training, connecting workers to new jobs, and other efforts to mitigate adverse effects for the long term.
The concentration of market power may help drive a widening of inequality that would strain the American Dream. In my judgment, now is the time for society to begin to consider how to address these potential challenges—while AI adoption is in its relatively early stages—so we can realize the long-term benefits for society.
Endnotes
1. See Michael S Barr (2026), “Will Artificial Intelligence Broadly Raise Living Standards or Drive Income and Wealth Inequality?” speech delivered at ‘Next-Gen Financial Inclusion’, the third annual Financial Inclusion Conference hosted by the Federal Reserve Board, Washington, July 14.
2. Nearly 4,500 single-person households and families were interviewed.
3. The Survey of Consumer Finances measures income for the calendar year before the survey—in the most recent case, for 2024. The sources of income collected in the survey include wages, self-employment and business income, taxable and tax-exempt interest, dividends, realized capital gains, unemployment insurance, food stamps and other related support programs provided by the government, pensions and withdrawals from retirement accounts, Social Security, alimony and other support payments, and miscellaneous sources. Note this survey gathers information from single-person households and families, referred to here simply as households.
4. See Board of Governors of the Federal Reserve System (2026), Economic Well-Being of U.S. Households in 2025 (PDF) (Washington: Board of Governors, May).
5. See Chase Peterson-Withorn and Matt Durot (2026), “Forbes 400: The Definitive Ranking of America’s Richest People 2026,” Forbes, September 15.
6. See Moritz Kuhn, Moritz Schularick, and Ulrike I Steins (2020), “Income and Wealth Inequality in America, 1949–2016,” Journal of Political Economy, vol. 128 (September), pp. 3469–3519. This paper discusses how differential changes in equity and house prices shaped wealth dynamics in postwar America.
7. This is 20 percent of all corporate equities held by U.S. households, consisting of both direct and indirect holdings of stock mutual funds and exchange-traded funds, including through retirement accounts.
8. See Emmanuel Saez and Gabriel Zucman (2016), “Wealth Inequality in the United States since 1913: Evidence from Capitalized Income Tax Data,” Quarterly Journal of Economics, vol. 131 (May), pp. 519–78; also, see Vincent J Geloso, Phillip Magness, John Moore, Philip Schlosser (2022) “How Pronounced Is the U-Curve? Revisiting Income Inequality in the United States, 1917–60,” Economic Journal, vol. 132 (October), pp. 2366–91.
9. See Peter H Lindert and Jeffrey G Williamson (2016), Unequal Gains: American Growth and Inequality since 1700 (Princeton: Princeton University Press).
10. See Saez and Zucman, “Wealth Inequality in the United States” (in note 8).
11. See Saez and Zucman, “Wealth Inequality in the United States” (in note 8); also, see Jesse Bricker, Joseph Briggs, Sarah Friedman, and Kamila Sommer (2025), “The Business Cycle Dynamics of the Wealth Distribution,” Journal of Political Economy Macroeconomics, vol. 3 (September), pp. 343–81.
12. See Ann Harrison (2024), “Disentangling Various Explanations for the Declining Labor Share: Evidence from Millions of Firm Records,” NBER Working Paper Series 32015 (Cambridge, Mass.: National Bureau of Economic Research, January; revised July 2025).
13. See Lindert and Williamson, Unequal Gains (in note 9).
14. Some of the steep decline in the labour share for several years after the pandemic was due to the relative rise of less labour-intensive industries; see Richard Audoly, Miles Guerin, Srinidhi Narayanan, and Rachel Schuh (2026), “The Post-COVID Decline in the Labor Share,” Federal Reserve Bank of New York, Liberty Street Economics (blog), June 24. But that decline has continued.
15. See Matthew Smith, Danny Yagan, Owen Zidar, and Eric Zwick (2022), “The Rise of Pass-Throughs and the Decline of the Labor Share,” American Economic Review: Insights, vol. 4 (September), pp. 323–40. This paper estimates that the decline in the labour share between 1986 and 2017 is overstated by approximately one-third as a result of the shift to pass-through firms.
16. See, for example, Michael S Barr (2026), “What Will Artificial Intelligence Mean for the Labor Market and the Economy?” speech delivered at the New York Association for Business Economics, New York, February 17.
17. See Mark Schaefer (2023), “How Big Is the Creator Economy? Three Times Larger Than We Thought!” LinkedIn, November 14.
18. See Loukas Karabarbounis and Brent Neiman (2014), “The Global Decline of the Labor Share,” Quarterly Journal of Economics, vol. 129 (February), pp. 61–103.
19. See Daron Acemoglu and Pascual Restrepo (2020), “Robots and Jobs: Evidence from U.S. Labor Markets,” Journal of Political Economy, vol. 128 (June), pp. 2188–2244.
20. See Daron Acemoglu, Claire Lelarge, and Pascual Restrepo (2020), “Competing with Robots: Firm-Level Evidence from France,” AEA Papers and Proceedings, vol. 110 (May), pp. 383–88.
21. See Roc Armenter (2015), “A Bit of a Miracle No More: The Decline of the Labor Share,” Federal Reserve Bank of Philadelphia, Business Review,; also, see Gustavo Grullon, Yelena Larkin, Roni Michaely (2019), “Are U.S. Industries Becoming More Concentrated?” Review of Finance, vol. 23 (July), pp. 697–743.
22. See Emin Dinlersoz and Zoltan Wolf (2024), “Automation, Labor Share, and Productivity: Plant-Level Evidence from U.S. Manufacturing,” Economics of Innovation and New Technology, vol. 33 (May), pp 604–26.
23. See David Autor, David Dorn, Lawrence F Katz, Christina Patterson, John Van Reenen (2020), “The Fall of the Labor Share and the Rise of Superstar Firms,” Quarterly Journal of Economics, vol. 135 (May), pp. 645–709.
24. See Gene M Grossman and Ezra Oberfield (2022), “The Elusive Explanation for the Declining Labor Share,” Annual Review of Economics, vol. 14 (August), pp. 93–124; also, see Stansbury and Summers, “The Declining Worker Power Hypothesis” (in note 12).
25. See Karabarbounis and Neiman, “The Global Decline” (in note 19); also, see Grossman and Oberfield, “The Elusive Explanation” (in note 24).
26. See David H Autor (2025), “Polanyi’s Paradox and the Shape of Employment Growth,” in Re-Evaluating Labor Market Dynamics: A Symposium Sponsored by the Federal Reserve Bank of Kansas City (Kansas City: Federal Reserve Bank of Kansas City, pp. 129–77).
27. See, for example, Barr, “What Will Artificial Intelligence Mean,” (in note 16).
28. See Anton Korinek and Jai Vipra (2025), “Concentrating Intelligence: Scaling and Market Structure in Artificial Intelligence,” Economic Policy, vol. 40 (January), pp. 225–56.
29. See Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen (2026), “No Widespread Displacement, but the AI Employment Gap for Young Workers Has Widened to 19%,” Stanford Digital Economy Lab, August 12; David Autor, Caroline Chin, Anna M Salomons, and Bryan Seegmiller (2026), “What Makes New Work Different from More Work?” NBER Working Paper Series 34986 (Cambridge, Mass.: National Bureau of Economic Research, March); Isabella Loaiza and Roberto Rigobon (2024), “The EPOCH of AI: Human-Machine Complementarities at Work,” MIT Sloan Research Paper No. 7236-24 (Cambridge, Mass.: MIT, November; revised October 2025); Wilbur Xinyuan Chen, Suraj Srinivasan, and Saleh Zakerinia (2024), “Displacement or Complementarity? The Labor Market Impact of Generative AI,” Harvard Business School Working Paper 25-039 (Cambridge, Mass.: Harvard, December); and Kathryn Bonney, Cory L Breaux, Emin Dinlersoz, Lucia S Foster, John C Haltiwanger, and Aditya A Pande (2026), “The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks,” NBER Working Paper Series 35141 (Cambridge, Mass.: National Bureau of Economic Research, April).
The views expressed here are my own and are not necessarily those of my colleagues on the Federal Reserve Board or the Federal Open Market Committee. This article is based on a speech delivered at Yale University, New Haven, Connecticut, 12 November 2026.
