Taxpayers Must Not Be on the Hook for Bailing Out the AI Industry

The U.S. economy is heading toward uncharted territory. Artificial Intelligence (AI) is transforming business operations and all areas of finance, the national debt is on an unsustainable fiscal path, and most middle-income Americans cannot keep up with rising costs. Moreover, as hundreds of billions of dollars are flooding into artificial intelligence, there are serious concerns that an AI bubble could trigger another bailout with taxpayer money.

In the interview that follows, leading progressive economist Gerald Epstein, who initiated and directs the Game Changers project at the Political Economy Research Institute at the University of Massachusetts Amherst, talks about AI’s impact on finance and the bailout problem and how a transformative public finance strategy can revive democracy, promote equity and social justice, and build strong and sustainable communities. Epstein is professor of economics and founding co-director of the Political Economy Research Institute.

C.J. Polychroniou: AI is possibly the fastest-growing technology in history and is already reshaping the U.S. economy in innumerable ways as organizations of all kinds are deploying AI tools across every functional area. But there are many public concerns surrounding AI, including whether it is becoming “too big to fail.” If so, taxpayer money may be used yet again to bail out private companies and institutions from collapse. What are the dangers here?

Gerald Epstein: There are indeed many public concerns about the economic and financial implications of AI. But let me focus on AI’s impact on finance as there are legitimate fears that the government may be asked at some point to bail out the AI industry, especially given the close ties of the Trump administration to AI firms.

As Marc Jarsulic and I argue in our Game Changers policy analysis “No More Bailouts,” the AI investment boom is increasing the level of risk for AI-related firms. The reason is that for these investments to be profitable, AI revenue will need to grow substantially. But some financial analysts doubt that the revenue will materialize. For example, in Bain’s Technology Report 2025, it is estimated that capital expenditures of $500 billion per year would be required to cover anticipated AI demand. To fund this investment, annual AI revenue would need to increase to $2 trillion. However, after accounting for likely cost reductions at the AI firms, Bain’s research concluded that their annual revenues would fall $800 billion short of that mark.

In that context, some of the risks of the AI-led investment boom have been shifted to financial markets as the AI boom is relying increasingly on debt. The revenues of the tech monopolies are extraordinary, but they are insufficient to pay all the costs of data centers. In 2025, the capital expenditures of the top five hyperscalers were more than 30 percent of their total revenue. They are projected to rise to more than 40 percent by the end of 2026. This has forced these companies to turn to debt and equity markets for finance. In 2025, major tech hyperscalers alone issued $120 billion in debt, but the figure was pushed toward $175 billion in 2026 and is expected to rise to $300 billion annually in the coming years.

A distorted government response to an AI-bust is in the cards. A rational policy response to an AI bust would not include bailouts of AI-related firms or their creditors. The investment boom, and the debt that helps finance it, have been engineered by sophisticated actors aiming to dominate a potentially lucrative innovation. There is no ex ante (“before the event”) government safety net for financial bets of this kind. Caveat emptor (“let the buyer beware”) applies to the firms playing the AI domination game if it ever applied to anyone. Unfortunately, we cannot count on a rational response from the current administration. Extensive conflicts of interest and cronyism mean that government bailouts for insiders would be on the table. Hyperscalers and other AI-related firms have received extraordinary government support. The Trump administration has taken over 90 federal actions to help the AI industry, including executive orders easing permitting for massive data centers, federal support for AI exports and reducing regulatory barriers for the industry. The Trumpian transactional quid pro quo is apparent. Nvidia and Intel have given the government equity to curry favor. A number of major AI companies contributed several million dollars to Trump’s White House demolition — a.k.a. his “ballroom” — fund. And AI firms continue to finance MAGA. This is business as usual for Trump.

If AI goes south, we should expect major pressures from the Trump administration for a debt- or taxpayer-funded bailout.

Bailout operations are a trademark of neoliberalism. That being said, what is the actual problem with bailouts, and how do we put an end to the bailout problem?

Charles Kindleberger, the famed MIT economic historian, summed it best when he said that international financial crises are “a hardy perennial” of capitalism. Going back to the 16^th century, bubbles and crashes have occurred roughly every 10 years. And government bailouts have not been far behind. As Andrew Haldane, former chief economist of the Bank of England, put it, “Historically, the link between the state and the banking system has been umbilical. Starting with the first Italian banking houses in the 13th century, banks were financiers of the sovereign.”

Sovereign defaults were frequent and dangerous for banks. But, as Haldane noted, “For the past two centuries, the tables have progressively turned. The state has instead become the last-resort financier of the banks” through bailouts and other supports.

Ominously, writing in 2009, Haldane noted that “Today, perhaps the biggest risk to the sovereign comes from the banks. Causality has been reversed.” In other words, these days, the pressure for massive bailouts can get so powerful that the resulting government debt can be destabilizing for the entire economy. The extreme financial deregulation cemented by Alan Greenspan, Bill Clinton, Robert Rubin, and others, best exemplified by the overthrow of the New Deal-era Glass-Steagall Act, paved the way for the Great Financial Crisis of 2007-2009. The Dodd-Frank Act of 2010 tried to put new guardrails into place but began to be eroded by both Democrats and Republicans even before the ink was dry. It has now practically been torn up by the Trump administration in its push for all-out war on financial regulation.

Government bailouts are in the cards not only for AI. Crypto might be in line as well. Crypto currencies are privately created, speculative, and highly volatile digital “assets.” They are traded on online private exchanges. As with war, one might ask: what is crypto good for? Despite crypto’s common reference as a “digital currency,” the fact of the matter is that it satisfies virtually none of the qualifications to serve as a currency. It is not widely accepted as a means of exchange; it is highly subject to fraud; the “blockchain” technology that underlies it is inefficient and error-prone; and, perhaps most important, its value relative to goods and services is highly unstable.

There are “use cases” for crypto, but they are mainly for money laundering, tax evasion, and other criminal activities. According to the 2026 Crypto Crime Report, crypto laundering activities skyrocketed in 2025, with a record $158 billion of crypto tokens were sent to criminal wallets, representing a 145 percent increase from the previous year.

Like AI, the growth of the crypto industry creates a raft of problems for the financial sector. A run on stablecoins backed by a significant amount of short-term Treasury bills could create a market shock.

As Andrew Haldane emphasized, an important factor that leads to financial crises is the financier expectation of government bailouts. These bailouts mean that financiers do not have to pay the full costs of their risky behavior and, as a result, they take on more risk than is warranted, risks that increase the chances their banks will get into financial trouble. The insurance term for this dynamic is “moral hazard.”

A no-bailout rule would help interrupt this dynamic. Perhaps the best-known early advocate for such a rule was the 19th century journalist and economic commentator, Walter Bagehot. Bagehot’s prescription for central bank bailouts dealing with a financial panic was clear: the central bank should lend freely, only to solvent banks, against good collateral and at a high (penalty) rate of interest. In other words, no bailouts for the bankrupt. A commitment to a no-bailout rule would help prevent the development of financial doom loops of excessive risk-taking, crisis and bailout. But historically, as Kindleberger puts it, “the only rule is that there is no rule.” Central banks and governments have felt compelled to bail out banks and/or financial markets when panic strikes, it seems, no matter what preconceived strictures they have tried to put on their behavior.

Progressive economics also mandates that we address the problem of inequality. How can this be done in a manner that is consistent with sound public finance, especially since, as you have pointed out, “the way Washington finances its spending today is itself a driver of inequality.”

As Aaron Medlin and I show in our Game Changers analysis of tax policy, the United States has a massive inequality problem. The top 1 percent of households hold about a third of the nation’s household wealth and roughly half of the stock market. In terms of income, the top 1 percent takes home twice as much total income as the entire bottom half of the country combined.

The way Washington finances its spending today is itself a driver of inequality. Working Americans cover much of the cost through taxes on their labor income, while the wealthy and corporations pay far less than their share. What these regressive arrangements leave uncollected, the government borrows, largely from those same wealthy households, paying them interest year after year and deepening the concentration of income and wealth. Our proposal flips the script by funding the future in a completely different way: it asks the wealthy and corporations to pay their fair share, creating more equality in the process.

Republican administrations especially — beginning with Ronald Reagan, then George W. Bush, and continuing through both Trump terms — have made these problems much worse, enacting large tax cuts that primarily benefit the wealthy and greatly widen federal deficits. In July 2025, the Trump administration signed the “One Big Beautiful Bill Act,” which permanently extended and expanded the 2017 tax cuts and delivered its largest gains to the top. The richest 1 percent of households are projected to receive tax cuts worth more than $50,000 a year, and the richest 10 percent more than $14,700 a year. According to the Congressional Budget Office, the law will reduce federal revenue by trillions of dollars over the next decade and widen the deficit by roughly $3 trillion through 2034, and, in fact, more than $4 trillion once added interest is counted, while increasing spending on the military and immigration enforcement. These trends have been compounded by a steep decline in corporate tax revenue, from about 7 percent of GDP in 1944 to about 1 percent in 2019.

The unifying idea behind the Game Changers’ progressive revenue package is to reach the income at the top that the current system lets slip. For decades the tax code has reached wages easily and rentier income barely at all. A nurse’s salary is taxed in full and on time, while a lifetime of gains on an inherited portfolio can go untaxed entirely. The package corrects that imbalance, drawing its revenue from capital income, corporate rents, dynastic wealth, and speculative finance rather than from the paychecks and purchases of working families. It uses the proceeds to make those families’ lives more affordable and secure. Two design principles follow. First, the package taxes high incomes on both wages and capital, where most plans reach only one. High salaries face graduated contributions, and the preferential rate on long-term capital gains ends, so that income from selling assets is taxed at the same rates as the wages of someone who works for a living. There is no longer a discount for income earned by owning rather than working. Second, working people are insulated: every wage-based levy is graduated from the first dollar, with nothing below $20,000 of earnings taxed. Households below $200,000 pay less on net once lower health costs, the child allowance, and federally funded child care are counted. The burden is concentrated at the top, not shifted onto the flat payroll and consumption taxes that fall hardest on working families.

What would a transformative public finance script look like for funding the future? What sort of progressive policies would be pursued?

First, let’s be clear about how both Republican and Democratic administrations preside over public finance. For Republican administrations, it is typical to massively increase government deficits and run up the debt to pay for big tax cuts for corporations and the wealthy, and then for the Democrats to come into office as the party of austerity in order to try to undo the damage by making workers and the middle class pay more. This is both bad economics and bad politics.

A transformative public finance script calls for the super rich and the corporations to pay their share again so that the bribe system of paying the rich to pay their share stops. The key aim is to rewrite the rules for the purpose of building an economy that prioritizes the health, security, and true prosperity of working people and their families.

The Game Changers policies to help create an economy that works for working Americans include universal health care, universal child care, social housing, fossil fuel phasedown, rigorous financial regulation, raising wages, defending reproductive rights, immigrant integration initiatives, and shortening the work week to four days. Most of these policies can more than pay for themselves over time. The short-term budgetary costs can be met by slashing a bloated and dangerous military budget, reallocating funds now spent on immigration enforcement to immigrant integration, and increases in tax revenue from corporations and higher taxation of great fortunes, as I discuss above.

Although difficult to quantify on balance sheets, the stakes are more than budgetary. When income and wealth concentrate at the very top, political power does so as well. Fortunes that compound untaxed across generations do not stay out of politics; they fund campaigns, lobbying, and media, and they bend public policy toward the interests of those who already have the most. A tax system that reaches rentier income, corporate rents, and dynastic wealth does more than raise revenue. It loosens the grip of concentrated wealth on public life and returns a measure of influence to everyone else. Funding the future, reducing inequality, and strengthening our democracy are not competing goals. They are the same project.