The AI Tax Debate and Fiscal Reality
Senator Wyden recently proposed limiting tax benefits for new data centers and levying a new excise tax. Many lawmakers actively want to use taxes to address worries that AI could eliminate jobs and inflict environmental harm. However, a specific tax on AI would likely do far less than its supporters hope. A new National Bureau of Economic Research working paper examines how AI could affect the U.S. fiscal trajectory. At first, the news looks encouraging because faster growth means a bigger economy to carry the national debt. Yet, even with optimistic AI economic growth, the federal debt will continue to rise significantly relative to GDP over the next three decades.
The Ineffectiveness of Narrow AI Taxes
A specific tax on AI pulls policy objectives in opposite directions. For instance, a compute tax actively discourages the essential training that makes models better. Similarly, a token tax falls on all users and quickly spreads costs across the entire economy. Furthermore, taxing windfall profits remains difficult in today’s tax system, meaning actual revenue may be years away. Ultimately, none of these narrow levers are likely to raise enough revenue to meaningfully change the national debt trajectory. Instead, they risk creating a dangerous fiscal illusion that a narrow tax can solve a massive economic deficit.
Taxing Capital Income as a Solution
The deepest debt reduction actually occurs when AI shifts income from labor directly to capital. Therefore, lawmakers must focus heavily on how they tax capital income rather than singling out AI. There are plenty of options for strengthening capital income taxation today. These broad options include raising corporate and capital gains tax rates, taxing wealth, or implementing a broad-based consumption tax. Closing this specific taxation gap would significantly improve the long-term fiscal outlook. Finally, this broader approach would help fund government responses to potential AI-driven job displacement and future recessions.
Reference
Patel, E., & Gordon, T. (2026, 7 agosto). AI tax debate misses the threat that’s already here. Brookings. https://www.brookings.edu/articles/ai-tax-debate-misses-the-threat-thats-already-here/?utm_campaign=Brookings%20Brief&utm_medium=email&utm_content=432144006&utm_source=hs_email