AINEWS
Companies pay a tax for every worker they hire, then get a write off for the robots and software that replace them. RAND has a fix, and it is not the House bill being called a "robot tax."
Companies pay a tax for every worker they hire, then get a write-off for the robots and software that replace them. RAND has a fix, and it is not the House bill being called a "robot tax."
Roughly two-thirds of every dollar the U.S. government collects comes from taxes on human labor, most of it the employer-side payroll tax that hits each new hire. At the same time, the same Internal Revenue Code lets companies depreciate the industrial robots, software, and GPU clusters that replace those workers on an accelerated schedule, writing the cost off faster than the equipment actually wears out. That two-sided asymmetry is what makes the current AI-displacement wave a fiscal event, not just a labor story. A new proposal from the RAND Corporation, sketched by mathematician Carter Price in a September 2026 Q&A, is designed to close that gap.
Price's idea, drawn from a longer RAND research report published August 25, 2026, is not a tax on machines. It is a progressive corporate bracket keyed to profit per worker. A company that earns $1 million in profit and employs 100 people has $10,000 of profit per worker; replace 90 of those workers with AI, and the same $1 million of profit now sits on 10 employees, or $100,000 per worker. Under Price's design, that 10x jump in profit density pushes the firm into a higher corporate tax bracket, without changing its total profit or making the automation decision uneconomic. The point is not to punish efficiency. It is to make the federal government whole for the payroll taxes that disappear when the workers do.
Hiring a human today triggers a 7.65% employer payroll tax, while buying a robot triggers an accelerated depreciation schedule that lets the same firm recover the equipment's cost over five years, often front-loaded. The asymmetry compounds: each laid-off worker is a recurring loss to the Treasury, and each robot is a one-time deduction. Price's bracket is one specific counterweight, and it does not pretend to stop displacement where AI is genuinely cheaper or better. The source itself is candid about that limit. The tax makes displacement more expensive, not impossible.
Whether the math holds at scale is the open question. RAND's August report finds that if AI replaces labor across a meaningful slice of the economy and displaced workers do not find new jobs, federal revenue could fall sharply, with corporate AI profits failing to fully offset the lost labor taxes. Price is explicit that the proposal is meant to preserve revenue and slow displacement where AI is not a clear win, not to pick winners and losers in any specific industry.
H.R. 9501, the AI Tax Integrity Act of 2026, was introduced June 29, 2026 by Rep. Vern Buchanan (R-FL) and ordered reported by the House Committee on Ways and Means on July 1, 2026. The Joint Committee on Taxation's description of the bill shows its core is a narrow AI fraud-detection pilot for the IRS, the kind of enforcement tool the agency has asked for in past budget requests. It is not the profit-per-worker corporate bracket, and it is not the mechanism Price is sketching in the RAND Q&A. Calling the two ideas the same thing flatters neither.
The reason the conflation matters is that the two proposals answer different questions. H.R. 9501 asks whether the IRS can use AI to catch tax fraud. The RAND proposal asks what happens to the federal balance sheet when the payroll-tax base, roughly two-thirds of revenue, starts shrinking because the workers behind it are being replaced by depreciable capital. The first is a procurement question. The second is a structural one, and the wire-level "robot tax" shorthand hides it.
The next CBO long-term budget outlook, with its payroll-tax assumptions, will be the first public test of whether the bracket idea gains any traction in Washington. The arithmetic, in the meantime, has already been done: the current code picks a side in the AI displacement race every time a company files its 1120.