The Case for an AI Token Tax

What It Covers

A growing policy debate around taxing AI usage at the token level gains momentum, with US Senate candidate Mallory McMorrow, Senator Elizabeth Warren, Mark Cuban, DuckDuckGo's Gabriel Weinberg, and Anthropic's Dario Amodei all proposing variations of a per-token fee to fund displaced worker programs and public goods.

Key Questions Answered

• •Tax Base Erosion Risk: When humans perform work, income and payroll taxes capture roughly 35.1% of labor costs across OECD nations. When AI agents perform equivalent tasks, value surfaces as lower costs or capital gains, taxed at lower rates or not at all. The IMF flagged this labor-substitution tax base erosion risk explicitly in 2024, making structural reform increasingly unavoidable.
• •Token Tax Proposals on the Table: Mark Cuban proposes under 50¢ per million tokens at the provider level, projecting $10 billion annually scaling 30–100x over a decade. DuckDuckGo's Gabriel Weinberg suggests a 10% surcharge matching employer payroll tax rates. Anthropic's Dario Amodei floated 3% of inference revenue redirected to government redistribution, acknowledging it works against his own economic interest.
• •Tokenizer Endogeneity Problem: A flat per-token tax discriminates arbitrarily across providers because tokenization rates vary dramatically — Mandarin runs 2–3x more tokens than English, source code 1.5–2x more, and some low-resource languages up to 15x more. Since providers control their own tokenizers, taxing tokens creates perverse incentives for providers to game tokenization efficiency to minimize tax liability.
• •Token Price Deflation Makes Fixed Rates Unworkable: Per-token prices have declined roughly 200x annually for two years. A fixed 50¢-per-million-token tax representing 5% of frontier pricing in year one becomes effectively 1,000% of that same price by year three. Congress must either index the rate downward, collapsing revenue projections, or leave it fixed, making the tax confiscatory and pushing users toward foreign or open-source providers.
• •Intermediate vs. Final Use Distinction Is Critical: A Brookings-sponsored January 2025 paper on public finance in the AI age argues token taxes applied to business-to-business inference distort productive investment by taxing intermediate production. Their recommended approach: apply consumption-based token taxes only at the point of final human use, integrated into existing VAT and sales tax infrastructure, with B2B exemptions to prevent cascading economic distortion.