Could AI help America out of its debt hole?

When US President Donald Trump unveiled his “big, beautiful bill” (BBB) last month, critics used data from Washington’s congressional budget office to attack it.

The reason? The office is tasked with projecting the long-term US fiscal outlook. And even before that BBB initiative, with its trillions of dollars of tax cuts, the watchdog’s baseline scenario – using unchanged policies and fundamentals – was that debt will jump from its current 100 per cent of GDP to 156 per cent by 2055.

However, there is a crucial, and little-noticed, caveat that matters enormously right now: while the congressional budget office’s (CBO) baseline projections grab all the attention, it recently produced eight other forecasts showing the possible impact of shifting fundamentals.

Some are more alarming: for instance, if interest rates rise five basis points a year more than the office’s baseline projection then debt will exceed 200 per cent. One, however, is not: if annual productivity grows by 0.5 percentage points more than CBO projections because of artificial intelligence, debt will flatline at “just” 113 per cent of GDP, even without austerity.

The question isn’t how Jim Gavin will run for President, it’s why he would want to

Justine McCarthy: Catherine Connolly’s continuing association with Clare Daly and Mick Wallace is troubling

Electric Picnic 2025 in pictures: The party kicks off at summer’s largest music festival

That is bad, but not a disaster. And while the CBO report does not attribute this just to AI, it has previously lauded the technology’s impact on productivity.

“AI could solve the US fiscal problem,” suggests Apollo, the private capital group. Is this too good to be true? Perhaps: there are at least three reasons to question the optimism.

First, history shows that innovation affects productivity in an unpredictable manner. The Brookings Institution, for example, notes that annual productivity rose around 3 per cent between 1995 and 2005, possibly because of digitisation. It was just 1.5 per cent between 2005 and 2022 – and equally low between 1973 and 1995.

Indeed the Nobel laureate Robert Solow lamented in 1987 that computers “were everywhere” except in the productivity statistics.

Second, this uneven impact has arisen in the past because there is wide variety in how companies embrace innovation. That could intensify with AI, as Oxford Economics suggests.

Third, insofar as productivity does rise, this could unleash social turmoil, given that groups such as JPMorgan and the IMF reckon that AI could displace half of US jobs by 2034.

Dislocation on this scale has happened before. Just look at the industrial or agricultural revolutions.

However, those shocks unfolded slowly (JPMorgan reckons it took 15, 32 and 61 years respectively for the internet, electricity and steam engines to have a discernible impact on productivity). Moreover, governments unveiled major policy reforms, such as the introduction of universal education and welfare state.

This time, however, JPMorgan thinks AI could change productivity in just seven years. And thus far the White House shows no sign of preparing a proactive set of sensible industrial and social policies to offset the human costs. T

he risk, then, is that AI unleashes political and social strife, which will sap growth and undermine fiscal reform.

Optimists also have a riposte to those three points. The dismal recent productivity data means that even modest shifts in this series could change the statistics markedly. JPMorgan, for example, projects that AI will “only” boost growth by around 10 per cent by 2034, while Goldman Sachs and PwC expect uplifts of 15 and 20 per cent respectively.

And the fact that companies are responding to innovation unevenly might not be such a disaster. A fascinating recent report from McKinsey suggests that what really determines whether countries grow is whether a few big influential companies embrace innovation (or not), rather than what happens to the mean.

Third, while the current Trump administration has hitherto failed to produce a coherent policy response to the growth of AI – of the sort that Singapore, for instance, is unveiling – future administrations might do. After all, techies are now more involved with Washington, and the Trump regime has shown how quickly the so-called Overton window – the range of publicly-debated policy ideas – can shift.

Nothing can be ruled out on the future policy front if social conflict explodes.

Don’t get me wrong: by noting those counterfactuals, I am not endorsing this optimistic AI-as-saviour thesis – or at least not yet. The AI shock is still so new that I assume that if anything changes the path of US debt, it will be inflation, financial repression or implicit default.

However, the CBO’s upbeat AI scenario matters for two reasons. First, it is a humbling reminder of the vagaries of economic predictions. Second, this helps to explain the actions and thinking of Trump’s policy team.

For while Trump’s critics (and most mainstream economists) fear that America is heading towards ever-rising debt and stagflation in the wake of the BBB, the White House economic team does not see it that way.

They believe in the CBO’s optimistic vision and think that deregulation and an AI productivity miracle will cause lower inflation, higher growth and falling debt. Crucially, they also want America to reap the most global benefits of that, at the expense of Europe, among others.

Critics might view this as a mirage – or, to use AI jargon, a “hallucination”. But it should not be ignored. All eyes, then, should be on that (lesser known) CBO projection – and not just investors in tech stocks, but holders of treasuries too. – Copyright The Financial Times Limited 2025