Sometime in the next few weeks, U.S. federal government debt will reach $40 trillion. The United States now owes roughly as much as all the other major advanced countries combined. With Washington adding another $2 trillion, or six percent of national income, to the figure each year, it has long been apparent that the U.S. debt trajectory is unsustainable. Or is it?
According to some proponents of artificial intelligence, today’s remarkable technological advances could soon make the problem disappear. In this view, the spread of AI through the economy will generate such tremendous growth that it will generate huge new tax revenues that will help pay off the debt. But any reduction in the debt burden will require not only that those tax revenues actually materialize, but also that the government avoids overspending any bonanza it gets. This will be especially challenging given that the AI boom will likely drive up interest rates, making servicing the debt more expensive.
Of course, Washington’s debt troubles do not translate immediately into a crisis. Thus far the country has continued to have no trouble placing its debt, as long as it is willing to pay a higher price to do so. The United States, after all, is already a spectacularly rich country and can easily afford to pay the higher interest rates—if the government is willing to adjust long-term tax and spending trajectories accordingly. Yet investors have become increasingly attuned to signals that Washington is not eager to tackle the problem. A reasonable view is that until there is a debt crisis of some form—generally starting with either sharp or sustained** **rising interest rates—neither party is going to make a serious effort to trim deficits when it is in power, not if it wants to stay there.
Of course, if growth surpasses expectations, it will indeed help create surpluses, as long as favorable surprises keep rolling in. But if our impecunious government confidently anticipates getting sustained higher revenues for many years to come, there is every risk that tax cuts and spending hikes will get out in front of the boom, which in turn may never fully materialize. Over the long history of sovereign debt and inflation problems, the inability to pay has almost never been the root cause of a crisis, at least not for a middle-income or high-income country. That suggests important lessons for the United States as it enters the AI age.
“THIS TIME IS DIFFERENT”
The United States’ debt problems go back many decades. But to understand how the country got to the present situation, it is sufficient to start with the 2008–9 global financial crisis. Back then, governments around the world, in sensible Keynesian fashion, leaned into stimulus plans to mitigate the postcrisis recessions. At first, it was thought that over time, governments would want to allow growth to organically whittle down debt-to-income ratios, which is the ideal way to pay down massive peacetime spending and tax cut binges.
After World War II, the fact that growth consistently outstripped interest rate costs and helped bring down debt-to-income ratios. Even so, it is now understood that interest rates in the United States, the United Kingdom, and other advanced economies were held down by having highly regulated financial markets that favored government debt. So-called financial repression is an implicit tax on savers, especially on low- and middle-income individuals who hold a lot of their savings in bonds and bank accounts.
In any event, growth after 2010 was so weak that despite very low interest rates, debt-to-income ratios continued to climb and soon spiked even more than in the first year of the financial crisis. The good news, however, was that long-term government borrowing rates remained very low, so that the high and rising debt levels didn’t weigh heavily on government finances. Indeed, a great many economists began to assume—based on demographics and low productivity growth—that interest rates would remain ultralow indefinitely, a theory that seemed to imply that there were only minimal costs to higher debt.
Thus emerged the magical thinking of the 2010s: instead of worrying about debt, the argument went, policymakers should open their minds to the endless possibilities of using fresh mountains of borrowed funds to reshape government. Of course, proponents of greater borrowing acknowledged that there could be a future shock that once again demanded large government borrowing. But in their firm belief that long-term interest rates would always stay low, they regarded as old school any effort to rebuild fiscal ammunition by going through a few years of modest deficits or balanced budgets. The next war or financial crisis could always be dealt with by yet another massive stimulus, and overall debt levels would have to ratchet up a lot higher before they were worth thinking about, perhaps eons in the future. This view, encapsulated in former Treasury Secretary Lawrence Summers’ “secular stagnation” theory, even became orthodoxy among many policy economists.
The massive AI build-out means that there is less capital available for other investments.
Unfortunately, the longer history of interest rates and, in particular, interest rates adjusted for inflation expectations, made clear how misguided the “this time is different” argument was. In reality, it has not been at all unusual to have prolonged periods in which interest rates are very low—until they are not. For example, market-based interest rates were very low throughout the Great Depression, but during World War II and its aftermath, the government had to regulate and intervene heavily to keep long-term interest rates down. Underlying market-based interest rates can change relatively quickly; debt ratios cannot. Over the long run, reversion to mean in real interest rates—a return to the historical trend—gives a more sober prediction than just extrapolating from one decade.
Indeed, this is what in fact has played out over the past few years. The ten-year inflation-indexed U.S. Treasury yield, which had averaged around zero percent from 2012 through 2021, started rising in 2022; as of this writing, it stands at over 2.4 percent. Rates today look a lot like they did at the start of the twenty-first century and are arguably at more normal levels relative to historical trends. This is not to say that they are predictable, but it does mean that they may be just as likely to go up as go down. But in one sense, things are very different from the way they were 25 years ago: whereas interest rates may be similar, debt is much higher relative to U.S. income, and the costs of paying for the debt proportionately higher.
According to some measures, interest payments on the federal debt now exceed defense spending. Despite the United States’ growing defense needs, it is by no means clear which is going to rise faster in the future. There is little apparent political will from either party to rein in deficits. At the same time, creditors are starting to demand higher returns in exchange for holding U.S. debt. Part of the problem is that investors no longer regard U.S. government debt as fully safe, certainly not from inflation and possibly not from other factors, such as restrictions on what assets investors are allowed to hold, or financial repression.
There are, of course, many dynamics that are affecting long-term interest rates on U.S. debt today. The Balkanization of global trade and finance hurts countries such as the United States that rely heavily on massive foreign inflows. Rising military expenditures across the world bid up interest rates by reducing the output available for savings. And the massive AI build-out means that there is less capital available for other investments, again bidding up interest rates.
FAST GROWTH, SLOW SAVINGS
As the AI boom has taken off, a new argument has emerged for how the United States can solve its debt problem. According to this view, promoted by the Trump administration but also by many others, including some economists, the spread of AI will soon create a surge in productivity. This in turn will swell government coffers with tax revenues, making deficit concerns a thing of the past. As evidence, many point to the 1990s, when the tech boom played a big role in helping the Clinton administration move toward a balanced budget and ward off market concerns about U.S. debt levels that had become quite widespread in the early 1990s.
At least as far as productivity is concerned, there is some merit to this thinking. Current high-side estimates of what AI could add to average long-term U.S. productivity growth are in the stratosphere, with three to four percent being a low-end estimate among the tech elite of Silicon Valley. But there are several reasons to be wary of assumptions that this will translate into the government windfall that proponents are promising.
The productivity benefits of AI could take much longer to play out.
For one, there is substantial evidence that capital’s share of income has been rising over time—that is, firms’ profits have been rising even as labor income remains stagnant. The fact is that partly because of reasons of political power and partly because of mobility, capital has become harder to tax than labor. So even as the economy grows, long-run tax revenues are unlikely to keep pace. Of course, this imbalance can and should be addressed in the future, but doing so while maintaining the same level of growth will be challenge. And whether a sustained reconfiguration of tax burdens is politically realistic is far from clear. The productivity benefits of AI could also take much longer to play out than most Silicon Valley investors seem to believe. Many economists would put the more likely trend change over the next decade closer to one to two percent, which is still significant. But that figure would be partially offset by the costs related to an aging population, rising populism, and a retreat from globalization. And no matter how brilliant the technology, the transition to AI could lead to a plethora of problems—whether having to do with military uses, cyberattacks, or rogue AI—that will force policymakers to dramatically slow down its implementation.
If growth does rise sharply as a consequence of AI, interest rates may rise just as much or more, so the government would still have to deal with higher debt payments. Among the forces that could push rates up are the huge investments needed to fund data centers. Consumers may also have less reason to save if they think they will be far better off in the future—whether because incomes will be higher or that the government will take care of them generously through universal basic income or other subsidies. Less saving and more investment mean more competitive bidding for funds and therefore higher interest rates. Finally, even if AI generates substantial new tax revenues, the United States will face myriad other demands on how to spend those funds. These include more defense spending, dealing with environmental catastrophes, and of course the political demands driven by populism. If people become convinced that AI will deliver abundance, they will expect policymakers to fast-track the future and give them more of the rewards immediately. Think, for example, of Social Security being effectively indexed to growth and not just inflation.
FUTURE SHOCK
For U.S. policymakers, navigating the country’s debt problems will be a matter of balancing risks, not just maximizing short-term growth. A debt crisis does not typically come out of the blue. Rather, it often arises when a big shock—say a cyberwar or a real war—catches a country unprepared, particularly one that is already facing high debt and high interest rates and that is politically stuck, either by polarization or incompetence. The war in Iran may be thought of as a mini dry run for a larger problem. A disastrous environmental shock, hardly unthinkable over the next decade or two, could similarly be one that simultaneously restricts output and pushes interest rates up, making it harder to respond with standard Keynesian stimulus. AI may seem like a surefire cure for the country’s runaway debt, but that is far from certain, and policymakers need to plan accordingly. Over the centuries, the world’s wealthiest countries, including France, Germany, Japan, and Spain, among many others, have repeatedly run into debt and inflation spirals—more times than one might imagine. In the case of the United States, AI shows incredible promise to make an already very rich country even richer, but it is unlikely to reduce the political pressures that continually push each administration to keep spending far more than the government’s tax income. And if the United States does experience a debt crisis, it will affect the entire world and ultimately the standing of the dollar. Unfortunately, it may well take such a crisis to wake up voters, and therefore politicians, to reality.
You are reading a free article
Subscribe to Foreign Affairs to get unlimited access.
- Paywall-free reading of new articles and over a century of archives
- Six issues a year in print and online, plus audio articles
- Unlock access to the Foreign Affairsapp for reading on the go
Already a subscriber? Sign In