Treasury Secretary Scott Bessent has few cards left to play to free the U.S. from its debt chokehold. He appears to be putting all his money on one: an AI-driven economic growth boom. It’s unlikely to be the ace he thinks it is.
“We’re going to have to grow our way out of this,” Bessent said last month when the U.S. national debt crossed the US$40-trillion mark.
The assumption here is straightforward: the AI revolution will generate a rip in U.S. productivity, wages, and tax revenues, which will close the budget deficit and ultimately reverse the upward debt-to-gross domestic product (GDP) trajectory. It will be a repeat of the internet-fueled boom of the 1990s, but on steroids.
The astronomical spending driving the AI rollout is certainly juicing strong economic and corporate revenue growth right now. The Federal Reserve Bank of Atlanta’s GDPNow current estimate for real growth in the third quarter is an annualized 5.0 per cent, while nominal GDP growth is probably running close to an eye-popping 9 per cent.
Meanwhile, corporate earnings skyrocketed by 35 per cent in the second quarter, and company profits as a share of GDP have never been higher. And with an unemployment rate of just 4.1 per cent, a historically high proportion of Americans are contributing to government revenue from income and payroll taxes.
The budget deficit is sure to shrink then, right? Probably not.
The problem isn’t growth. It’s the tax structure.
The statutory U.S. corporate tax rate of 21 per cent is now the lowest since 1940. President Donald Trump cut it in his first term from the Clinton-era rate of 35 per cent. As a result, corporate tax revenue as a share of the overall government tax take and GDP has rarely been lower. In the 1990s, it made up 10-12 per cent of all federal receipts. That figure is now down to 9 per cent and appears to be falling. Corporate tax revenues as a share of U.S. GDP are also the lowest among the Group of Seven (G7) rich nations.
This dynamic was exacerbated by last year’s One Big Beautiful Bill Act (OBBBA), which the Trump administration hoped would turbocharge growth and reduce the deficit. But non-tariff tax revenue has been falling this year, even as the economy has been expanding. The OBBBA will also bake in years of wide budget deficits and decades of rising debt-to-GDP levels, according to the non-partisan Congressional Budget Office (CBO), which estimates that net federal revenues will fall by almost US$5-trillion in the 2026-2035 decade.
That’s a bleak outlook.
So what are the options?
One would be tax hikes.
A large reason the dotcom boom had such a positive fiscal impact in the 1990s was then-President Bill Clinton’s 1993 tax hikes. Once the internet boom really took off, the higher tax revenues helped the country experience a brief, rare budget surplus. This, in turn, reduced the debt-to-GDP ratio by around 10 percentage points to around 55 per cent.
Of course, the chances of the current administration introducing similar deficit-reducing tax hikes are close to zero. Indeed, Trump recently floated the idea of giving every U.S. adult a US$5,000 cheque if the Republicans retain control of the House and Senate in the midterm elections, a measure that would cost around US$1.3-trillion.
If taxes aren’t going to be raised, increased tax revenue will hinge on a wider tax-paying base. Income and payroll taxes together account for around 85 per cent of the federal tax take, so a job boom could be a game-changer.
But U.S. job growth is tepid. Average monthly nonfarm payroll growth over the last two years is 53,000 jobs, and only 41,000 over the last 12 months. Moreover, the break-even rate of monthly job growth needed to keep the unemployment rate steady is near zero, largely thanks to the Trump administration’s tough immigration policies, which have sharply reduced labor supply. For comparison, monthly job growth in the mid-to-late 1990s was well over 200,000.
There is still some hope that a 1990s-style technology-driven productivity boom could help fill the gap. San Francisco Fed research published in May suggests “the current moment may resemble the early stages of the 1990s technology boom.” This gives the authors “cautious optimism” that a similar productivity miracle could unfold today. After all, the economy is growing strongly with barely any employment growth.
But that may speak to the potentially unique ability of AI to boost output without increasing jobs. If AI boosts productivity largely by replacing human activity across large swathes of the economy in the coming years, then the income and payroll tax take is more likely to shrink than grow. Tech firms’ profits would surely rise, but the low corporate tax rate would limit the benefits to the government.
And even if there is a 1990s-style productivity miracle that does boost the tax take, it may not show up in the data for years.
In the meantime, the fiscal picture is expected to continue deteriorating. Interest rates are also rising substantially, making it even harder for growth to prevent debt-to-GDP from marching higher.
“We cannot simply grow our way out of a $2-trillion annual deficit and $40-trillion national debt,” House Budget Chairman Jodey Arrington said last month, a view seemingly at odds with Bessent’s.
Strong growth and productivity will help. Steven Englander and Dan Pan at Standard Chartered estimate that the AI boom could generate additional revenue equivalent to 0.6 per cent of GDP relative to the CBO’s baseline forecasts.
But to cure America’s debt woes, stronger medicine is needed.