AI saves the U.S. debt crisis, a productivity revolution or a beautiful fantasy?

Faced with high inflation and the expansion of U.S. debt, the U.S. government is pinning its hopes on the productivity miracle of AI to bring about lower inflation, higher growth, and declining debt levels.

The latest forecast from the Congressional Budget Office (CBO) indicates that if AI drives U.S. annual productivity growth beyond expectations by 0.5 percentage points, the ratio of U.S. debt to GDP will stabilize around 113%, rather than the current baseline forecast of soaring to 156% by 2055.

However, this optimistic expectation faces three challenges: history shows that the impact of innovation on productivity is difficult to predict, the degree of corporate adoption of innovation varies widely, and AI may trigger large-scale job displacement.

Employment Shock Risks Raise Concerns About Social Unrest

Pessimists are currently concerned that AI will bring about a massive employment displacement shock in the short term.

JP Morgan currently predicts that by 2034, AI could replace half of the jobs in the U.S., a pace of disruption that far exceeds the historical impacts of the steam engine, electricity, and the internet. Historically, the steam engine, electricity, and the internet took 61 years, 32 years, and 15 years, respectively, to have a discernible impact on productivity.

While historical transformations such as the Industrial Revolution and the Agricultural Revolution also led to large-scale labor displacement, their impacts unfolded relatively slowly, and the government implemented significant policy reforms such as universal education and welfare states to address these changes.

Currently, the White House has not shown signs of preparing a set of forward-looking and reasonable industrial and social policies to offset labor costs. At that time, AI could trigger political and social conflicts, thereby undermining growth and disrupting fiscal reforms.

Historical Experience Warns of Uncertainty in Innovation Impact

Meanwhile, there are doubts about AI's ability to enhance U.S. productivity.

Data from the Brookings Institution shows that U.S. annual productivity growth has been extremely unstable: from 1995 to 2005, driven by digitization, the annual growth rate of productivity was about 3%. However, from 2005 to 2022, it was only 1.5%, and it was similarly low from 1973 to 1995.

Nobel laureate Robert Solow lamented in 1987 that computers were "everywhere," yet difficult to find in productivity statistics. This phenomenon highlights the complexity and lag in the conversion of technological innovation into actual productivity improvements.

Despite these concerns, optimists have also put forth strong counterarguments. The recent sluggish productivity data means that even moderate improvements in this metric could significantly alter statistical outcomes. JP Morgan expects that by 2034, AI will "only" drive growth by about 10%, while Goldman Sachs and PwC anticipate increases of 15% and 20%, respectively.

McKinsey's latest research indicates that the key to whether a country grows lies in whether a few large influential companies embrace innovation, rather than the average level. This suggests that the uneven adoption of innovation among companies may not be catastrophic