The rapid expansion of artificial intelligence could give a significant boost to economic growth, but it may also create a new challenge for governments already struggling with high levels of public debt. While increased productivity and corporate earnings from AI could enlarge economies, governments may find it increasingly difficult to collect enough tax revenue if a greater share of national income shifts away from workers and towards capital.
The issue is already becoming visible in the United States, where heavy investment in AI infrastructure has contributed to a sharp rise in corporate profitability. American companies recorded a $1.2 trillion, or roughly 30%, increase in pre-tax corporate earnings over the past year. However, the rise in profits has not translated into a comparable increase in corporate tax collections. One reason is that companies have been able to deduct a larger portion of their investment spending from taxable income, while technology companies have made particularly large investments in AI-related infrastructure.
Economists have long argued that AI could improve productivity and accelerate economic growth. A larger economy would normally be expected to make it easier for governments to manage debt because tax revenues would increase alongside economic activity. Research cited in the article suggests that stronger growth driven by AI could substantially reduce America’s debt-to-GDP ratio over several decades compared with a scenario without such growth. However, even under a strong AI-driven expansion, the US could still have a considerably higher debt burden by the 2050s than it does today.
The bigger concern is the composition of economic growth. Governments in wealthy countries remain heavily dependent on taxes linked to employment, including personal income taxes and social-security contributions. In the United States, labour accounts for about three-quarters of federal tax revenue. If AI replaces some jobs, reduces wages in certain occupations or shifts income from employees to company owners, this traditional source of government revenue could weaken.
The potential fiscal impact could be significant. An analysis cited in the article considers a scenario in which labour’s share of national income falls by 10 percentage points. Countries that rely heavily on taxation of workers could face substantial revenue losses. Italy could experience a particularly sharp deterioration in its budget position, while France and Germany could also face considerable pressure. The United States would be affected as well, although to a lesser extent under the scenario examined.
At the same time, AI could increase demands on governments. Workers displaced by automation may require unemployment assistance, retraining and other forms of social support. Governments could also face pressure to introduce measures such as universal basic income if technological change significantly affects employment. In addition, rapid wage increases in highly productive industries such as software could raise the cost of employing skilled workers in sectors where productivity improves more slowly, including parts of the public sector.
AI could nevertheless create opportunities for governments to provide better public services. Its use in healthcare, for example, could improve scheduling, diagnosis and drug development while potentially lowering some costs. If productivity gains are successfully translated into more efficient public services, governments could obtain greater value from existing resources.
One possible response would be to introduce taxes directly on AI. Proposals have included a tax based on the computing power used to train AI systems or a charge on the number of times an AI model is used. But such measures could generate relatively limited revenue compared with taxes on labour. Current spending by American consumers and businesses on developing and operating AI models is estimated at around $700 billion annually, considerably smaller than the country’s approximately $16 trillion labour bill. Moreover, future AI investment and spending patterns remain uncertain.
Another option is to increase taxation on wealth and the gains generated by AI. However, wealth taxes have faced difficulties in several countries because wealthy individuals can move assets or relocate, while privately held companies can be difficult to value accurately. Of the 11 wealthy countries that imposed taxes on fortunes in 1990, only four still do so today, according to the article.
Economists have increasingly focused instead on modifying existing tax systems, particularly taxes on capital income. This could include bringing capital-gains tax rates closer to those applied to wages, improving taxation of actual investment returns and increasing corporate taxation where companies earn unusually high profits.
Consumption taxes could also become more important as machines assume a larger role in generating income. Such taxes are difficult to avoid and can generate substantial revenues, although governments may need to compensate lower-income households because consumption taxes can represent a larger share of their income.
The central fiscal challenge posed by AI, therefore, may not simply be whether the technology makes economies richer. The more important question could be whether existing tax systems are capable of capturing enough of that additional wealth. If AI shifts a growing portion of economic income from workers towards machines, software and capital owners, governments may have to rethink how they raise revenue to finance public services and social protection.
