Should artificial intelligence be taxed?

Artificial intelligence promises to make economies more productive, but governments are beginning to confront an uncomfortable possibility: What happens to public finances when companies can generate more output with fewer workers, and who should benefit from the wealth that follows?

The concern stems partly from governments' reliance on taxing employment. In the United States, individual income and payroll taxes account for approximately 85% of federal tax receipts, according to the Tax Policy Center. As AI replaces employees' tasks, economic activity could shift toward corporate profits and away from traditionally taxed wages.

To address this potential shift, policymakers are considering proposals to tax the technology, its use, or the profits it generates. Supporters argue that governments need new revenue to compensate displaced workers and prevent the benefits of automation from concentrating among a small number of companies. Critics warn that taxing technological progress could discourage investment without necessarily protecting employment.

The debate exposes a central tension in modern taxation. If machines assume a growing share of productive work, should tax systems continue treating human labor and automated production differently?

The economic case

The economic consequences of taxing artificial intelligence depend heavily on which activities governments target and what they hope to achieve.

An October 2026 Brookings Institution analysis identifies three arguments for AI-specific taxes. These concern declining tax revenues as automation shifts income toward corporate profits, infrastructure costs imposed by AI facilities, and the possibility of economically excessive AI investment.

These concerns do not necessarily justify the same policy response. According to Brookings, two-thirds of data centers built since 2022 are located in water-stressed regions, raising questions about who should finance their electricity and water infrastructure. Taxing data center electricity consumption could recover some of these costs, although distinguishing AI-related activity from other computing services would be difficult.

Taxes on AI usage present another trade-off. Charging companies according to the amount of information their models process could generate revenue as adoption expands, but would increase costs for applications that complement human workers as well as those that replace them.

Taxing extraordinary corporate profits could avoid some of these consequences, theoretically raising revenue without discouraging production. Yet identifying profits exceeding normal investment returns is difficult, particularly when AI development carries substantial financial risks.

Taxing the machines

These competing approaches are beginning to enter legislative debates, as policymakers consider how to translate the economic arguments for AI taxation into concrete measures.

In the United States, legislation introduced in August 2026 by Representatives Greg Casar, Valerie Foushee, and Sara Jacobs would impose an excise tax, a levy on a specific product or activity, on major AI companies. The tax would be calculated using either the value of AI tokens, the small units of information that AI systems process and generate, or qualifying AI revenues, whichever produces the larger liability.

The initial rates would be 2% and 3%, respectively, rising if unemployment increases. The proceeds would finance employment programs in sectors including infrastructure, housing construction. and care services. The mechanism is unusual because the tax burden would respond to labor-market conditions rather than remain fixed.

In June 2026, Senator Bernie Sanders proposed requiring major AI developers to give the government ownership shares, allowing the public to benefit directly from their profits. Rather than merely collecting revenue, the proposal seeks to give Americans a direct financial interest in companies developing powerful AI systems.

Neither proposal has become federal law.

An international debate

The debate is particularly consequential for countries already experiencing the economic effects of widespread automation.

In May 2026, South Korea's National Assembly Futures Institute proposed an AI social security tax to redistribute automation profits. President Lee Jae Myung has also expressed openness to a robot tax.

Potential approaches include taxing wages theoretically replaced by automation, profits attributable to AI, or computational activity such as processing tokens and using graphics processors.

South Korea has already experimented with a less direct intervention. In 2018, it reduced tax incentives for automation investment, effectively making some machinery purchases more expensive.

For developing economies, taxing automation presents a particular trade-off, as governments must weigh the need to protect workers against the risk of discouraging investment that could improve productivity and competitiveness.

The difficulty of taxing multinational technology companies is already evident in Europe, where disagreements over how to tax digital revenues have complicated efforts to establish a common approach. The European Commission proposed a 3% tax on certain digital revenues in 2018, although the measure was not adopted across the bloc. 

Who should pay?

The economic costs of an AI tax may ultimately fall on businesses and workers far beyond the companies directly targeted.

The International Monetary Fund has cautioned against taxing labor-replacing technologies, warning that such measures could discourage innovation, prove difficult to administer and penalize productive AI applications alongside those that displace workers.

Instead, it recommends reviewing corporate tax incentives that favor replacing workers with machines, strengthening taxation of capital income, considering excess-profits taxes and improving social protection.

A central difficulty remains identifying who ultimately bears the burden. Companies facing higher AI taxes could pass costs to customers, reduce investment, or relocate operations to jurisdictions with more favorable treatment. Developing economies face particular difficulties, as new levies could deter technology investment even as governments need revenue to support displaced workers.

As AI generates new sources of wealth while threatening established sources of employment, the choices governments make over taxation will carry consequences beyond public finances. Governments must decide how to capture the wealth created by automation without undermining the investment that produces it, and how much of that wealth should belong to the societies whose workers and economies are being transformed.