Category: Taxation · Horizon: Short to medium · Consensus: Contested · Feasibility: Medium; hard to define the base
What it is
This family of ideas taxes AI and automation directly, so that some of the value they create flows back to the public. Versions include a “robot tax” on automation equipment, a tax on the computing power used to train and run AI, a tax on “tokens” (the units of text AI models process and are billed by), levies on data centres, or a broader shift of the tax system from wages towards capital and profits.
Why it matters in the age of AI
Most governments raise much of their money from taxes on work: income tax and payroll contributions. If AI does more of the work, that base shrinks just as demand for support rises. Economists writing for the Brookings Institution in 2026 argued that the main burden of taxation will have to move away from labour. The question is whether to tax AI itself or to tax capital and profits more generally.
Evidence and proposals
- South Korea’s “quasi-robot tax”. In 2017 South Korea cut the tax credits that larger firms could claim for investing in automation. A 2025 study found that afterwards, Korean industries installed about 28% fewer robots than comparable Japanese ones, with no statistically significant effect on employment.
- Europe said no. The European Parliament rejected a robot tax proposal in 2017, with lawmakers arguing it would hurt business.
- A wave of 2026 proposals. In the US, Representative Greg Casar has proposed a tax on AI tokens, Senator Elizabeth Warren an excise tax on data centre energy, and Senator Ron Wyden a gross receipts tax on data centre operators. Bridgewater Associates’ leaders called for a consumption tax on tokens to lower taxes on labour and buy stakes in AI companies for all citizens. In the UK, the think tank IPPR proposed an international token tax.
- Bill Gates first suggested a robot tax in 2017 and has returned to the idea with a token tax.
Arguments for
- Protects public finances as wage-based taxes shrink.
- Ensures those who profit most from automation help those it displaces.
- Could fund other ideas on this site, such as a basic income, retraining or a citizens’ wealth fund.
- Token-based taxes are easy to measure, since AI providers already bill by the token.
Arguments against
- Taxing the tools of productivity could slow the growth that makes everyone richer.
- AI development and data centres could move to countries without the tax, unless many countries act together.
- It is hard to define what counts as a “robot” or as AI, especially as automation moves into ordinary software.
- Some economists, from groups such as the Tax Foundation and ITIF, argue mass unemployment is unlikely and that broader taxes on capital income or VAT are better tools.
Open questions
- Should we tax AI specifically, or shift the whole tax system towards capital and consumption?
- Can countries agree on an international approach, as they did with the global minimum corporate tax?
- Should the money go into general budgets, or straight to citizens?
Leave a Reply