Inequality is such a fact of American life that it’s easy to shrug off. But we are in uncharted terrain. The amassed wealth of today’s tech titans makes the Rockefellers and the Vanderbilts look quaint. Over the past two years, 19 households have added $1.8 trillion to their coffers, the economist Gabriel Zucman told me — roughly the size of the economy of Australia.
Into this fragile state enters artificial intelligence. It threatens to make a bad situation much worse.
Left on its current course, A.I. could deliver a bleak picture: lower- and middle-income jobs automated away, with top earners remaining unscathed. Income shifting from middle-wage workers doing the bulk of the labor toward those wealthy enough to bankroll the technology. Growth headwinds. Worsening affordability. So, too, a federal government less able to respond, thanks to a shrinking tax base.
For any society in which this much wealth gets concentrated in so few hands, and is then so easily parlayed into political clout, the question becomes one not just of economics but of basic civic standing. At some point soon, we are no longer sharing in self-government.
Start with A.I.’s impact on jobs. Technologists are convinced that a labor apocalypse is nigh. In this story, A.I. is sometimes posited as a great equalizer, gutting white-collar jobs and salaries, giving more clout to trades like plumbing and dimming the luster of that Ivy League degree. The theory has gotten the nod from academics, industry associations and institutions such as the O.E.C.D.
In truth, whether A.I. will lead to widespread job loss remains guesswork. But the notion that it will narrow inequality by pushing downward on top earners seems far-fetched. What’s already clear: As A.I. transforms anything touching a keyboard, it will land first and hardest on the income ladder’s middle and lower rungs. The jobs most at risk, say government forecasters and economists, are administrative and office support staff, sales and lower-level computer programmers — all roles with salaries of $40,000 to $100,000.
Those losses on the lower half of the scale are underway. One-quarter of computer programming jobs disappeared in 2023 and 2024. IBM’s chief executive said in 2023 he could “easily see” 30 percent of the company’s back office roles getting replaced by A.I. in the next five years. With at least three major rounds of job cuts in 2024 and 2025, IBM appears to be following through on that idea (though it has also signaled plans to grow entry-level hiring). A Stanford study found that early-career employees in A.I.-exposed fields like customer service have seen a 13 percent drop in employment since 2022 — unlike more experienced workers and those in other sectors.
At the same time, premiums for elite graduates with hefty Rolodexes full of powerful people, and tacit knowledge (like how to generate a laugh at a cocktail party on Park Avenue), aren’t going anywhere. Chatbots are no substitute for people who can call the right people when high-stakes deals go awry.
“Goldman Sachs’s Manhattan offices aren’t likely scathed,” Justin Searls, a writer and a founder of the software company Test Double who tracks A.I. advancement, told me in an email. “But its Jersey City operations? Well, that’s likely another story.” In a recent survey, 750 chief financial officers were twice as likely to say that A.I. could lead to job cuts in low-skill office work as they were to say it would enhance this work. And a majority thought A.I. would augment, rather than replace, higher-skill roles — especially those requiring high levels of education.
When more people are out of work, those who still have middle-class jobs — the marketers, the air traffic controllers and so on — will have less bargaining power, reducing their wages. A.I. can reinforce this downward spiral by lowering the expertise required for these remaining middle-class jobs. With A.I., we can all be coders now, right? As the pool of capable workers widens, wages fall once again. Younger workers who are un- or underemployed may need to work for decades to recover the lost opportunities and wages.
And that’s just the job market. Now let’s turn to how the A.I. investment boom seems poised to further grow the top end of the income scale. Thanks to years of handsome returns from the stock market, the top 1 percent holds more wealth than the bottom 90 percent combined.
Most new technologies tilt the scales toward investors and away from workers, but A.I.’s thumb is particularly heavy. Look at all the companies blaming A.I. for job cuts. While some argue that this explanation may just be a convenient cover for cuts that were going to happen anyway, investors are buying the story, rewarding these firms and their investors handsomely with surging share prices. Workers are losing, and stock prices are soaring, with the gains most benefiting wealthy investors. Inequality widens at both ends.
What’s worse, much of the trillion-plus-dollar investment in the A.I. boom isn’t happening in the stock market at all — it’s happening in private funds out of reach to all but the wealthiest, most connected among us. In earlier technology-fueled booms, companies like Amazon sold their shares in the public markets. As the value of its shares soared, they enriched Amazon’s early investors, yes, but thousands of employees also benefited, as did millions of other Americans, through pension funds and retirement accounts.
That isn’t the case with A.I. Anthropic and OpenAI, the two best-known A.I. companies, raised over $150 billion, mostly from venture capitalists, private equity firms and foreign sovereign wealth funds — funds mostly inaccessible to the vast majority of investors (let alone ordinary Americans).
With ownership of these firms concentrated in so few hands, any wealth they produce widens the gap between the richest households and everyone else. Also consider the fact that today’s A.I. firms employ far fewer people than established tech companies. OpenAI and Anthropic, which are already operating globally, employ only a few thousand people. Microsoft employs more than 200,000, and Amazon employs 1.5 million. The picture that emerges isn’t of just a deepening of the current divide. The A.I. story is one of more extreme concentration of wealth — at most likely not more than 3 percent of households, the very few who hold ownership in these A.I. companies or in the mostly private firms financing them.
Perhaps that doesn’t matter. So what if A.I. boosts inequality to new heights? Isn’t the more important question whether everyone has enough?
Start with economic growth itself. When inequality gets extreme enough, it starts to exact a cost on the whole system. Economists have shown that if the income share of the top 20 percent rises, growth in the gross domestic product actually declines over the medium term. That’s partly because lower- and middle-income households, which spend a far higher share of their earnings than wealthier ones do, have less money to put back into the economy. It’s also because inequality tends to make the most valuable skills more expensive and harder to access. (Think of the price of a house in a top-rated school district.)
This brings me to the fact that growing inequality worsens the affordability crisis. In 2025, for the first time since data collection on the statistic began regularly in 1989, the top 10 percent of earnings supplied nearly half of consumer spending, and those earners’ expensive tastes are already warping markets for essentials, from housing and health care to cars.
In San Francisco, where both OpenAI and Anthropic are headquartered, the median home price has surpassed $1.5 million amid soaring valuations. One home was recently listed for around $3 million and sold almost immediately for almost $5 million. As a small handful of winners supply an ever larger share of consumer spending, companies will mold their offerings to cater to their preferences, placing many essentials further out of reach for most Americans.
Next to fall are the fiscal dominoes.
Well-meaning policymakers often turn to federal spending to prop up our labor markets or address the affordability crisis. But they don’t factor in the tremendous debt load our government is currently servicing nor the negative impact A.I. is poised to have on the government’s coffers.
Because investment income is taxed at lower rates than wages — and because the wealthiest often find ways to defer or avoid those taxes altogether — A.I. will significantly shrink the tax base. Economists estimate that as $1 of value creation shifts from workers to owners, total tax revenue falls on the order of 10 to 15 cents. You don’t need to squint to see the resulting cuts to safety net programs like work-force training and Head Start that low- and middle-income families rely on — cuts that will, in turn, also worsen inequality.
None of what I have just described is a foregone conclusion. As M.I.T.’s Daron Acemoglu, Simon Johnson and David Autor note, A.I. could be designed to increase the value of human expertise, by building tools that make workers more capable, rather than replacing them outright, or even by creating whole new kinds of tasks requiring human involvement. But as they readily admit, so far A.I. is doing the opposite.
So what can we do?
A growing roster of economists suggest simply taxing investment profits more on par with labor income. Right now, the fruit of our labor is taxed at rates up to 37 percent, yet the tax rates for investment income top out at 23.8 percent.
Or we could follow the lead of other countries and make an effort to more evenly distribute investment profits. The sovereign wealth fund of Singapore, for example, which manages the country’s foreign currency reserves and uses the revenues for crisis funding and affordable housing, happens to be invested in Anthropic, which is why you are more likely to reap the upside of A.I. if you are in Singapore City than in Syracuse.
Another idea, so far still confined to think tank circles, proposes innovative tax structures to create public equity stakes in large A.I. firms; these stakes could then fund a better safety net or simply put money in workers’ pockets. After all, the “intelligence” in A.I. was ours to begin with. One especially promising fix is to incentivize more firms to convert into worker-owned cooperatives, building on modest federal support passed in 2022. If we put more workers in charge of the firms deciding how to use A.I., the odds climb that they will figure out how to use A.I. so as to increase their own value.
All of these fixes are made harder as the wealthiest parlay their economic clout into political sway. Billionaire contributions more than doubled as a share of total contributions from the 2020 election to the one in 2024. The largest A.I. companies are spending exponentially more to influence state level races this year compared with previous years.
The flywheel that turns wealth inequality into democratic backsliding is already churning. If we want to keep the “democratic” in democratic capitalism as A.I. takes hold, we’d better get to work. Fast.
Jennifer M. Harris directs the Economy and Society Initiative at the William and Flora Hewlett Foundation. She previously served as an economic official in the Biden White House.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our emA.I.l: [email protected].
Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.
The post We Are Witnessing the Rise of a New Aristocracy appeared first on New York Times.
