The conversation about AI and jobs usually stops at “will I still have one?” A new RAND Corporation study asks a different question: if AI replaces workers, who pays for Social Security and Medicare?
The answer is uncomfortable.
The Study
In a working paper published November 2025, RAND researchers Carter C. Price and Akshaya Suresh modeled what happens to US federal revenue when AI systems replace human labor. Not in some distant future — they simulated a 10% workforce displacement starting in 2025, projecting outcomes through 2035.
The study, Federal Revenue When AI Replaces Labor, doesn’t ask whether AI will take jobs. It asks: given that some displacement will occur, what happens to the money that funds government?
The findings are stark.
84% of Federal Revenue Comes from Labor
Here’s the number that matters: in 2024, 84% of federal revenue came from individual income taxes and payroll taxes. Both are derived primarily from labor income.
- Personal income taxes: 47% of federal revenue. 60% of that comes from wages and salaries.
- Payroll taxes: 37% of federal revenue. 100% of that comes from wages and salaries.
- Corporate taxes: 10%. This could go up or down depending on who owns the AI.
- Everything else: 6%.
The punchline: roughly two-thirds of all federal revenue is directly tied to someone having a job. When the jobs go, the money goes.
Four Scenarios, Four Levels of Pain
RAND modeled four scenarios based on two variables: whether displaced workers find new jobs, and whether AI systems are priced by monopolists or at-cost.
Scenario 1 — Business as Usual (New jobs + Monopolistic AI pricing) AI replaces workers, but they find new industries. AI profits concentrate in big companies. Federal revenue actually increases slightly because corporate profits generate tax revenue. This is the optimistic case.
Scenario 2 — Hyper-concentration (No new jobs + Monopolistic AI pricing) Workers can’t retrain fast enough. AI profits flow to a handful of companies. Payroll taxes collapse. Personal income taxes drop. Even though corporate taxes rise, they can’t compensate — because corporate tax rates are roughly half individual rates. 65% of federal revenue is directly vulnerable.
Scenario 3 — Price Shocks (New jobs + At-cost AI pricing) Workers retrain, but AI is cheap (open-weight models). Services deflate. Nominal GDP drops by 26% — from $43.9T to $32.4T. Federal revenue drops by roughly a quarter. Deflation of 2% per year (comparable to US industrialization from 1866-1897). 20% of federal revenue is directly vulnerable, but the deflation ripple affects everything.
Scenario 4 — System Breaks (No new jobs + At-cost AI pricing) This is the worst case. Workers can’t retrain, and AI is cheap. Labor income collapses. Corporate profits evaporate. Deflation hits hard. 83% of federal revenue is directly vulnerable. The system as currently structured cannot sustain itself.
Why Social Security and Medicare Specifically?
Social Security and Medicare are funded primarily through payroll taxes — taxes that only exist when someone receives a paycheck.
If 10% of the workforce is displaced by AI:
- Payroll tax revenue drops by 10% immediately
- The Social Security trust fund, already projected to be depleted by the mid-2030s, gets hit from both sides: fewer contributions and more early retirements
- Medicare faces the same squeeze with rising healthcare costs
And these are the moderate scenarios. The 10% displacement model is a conservative starting point. If AI displaces 20% or more, the fiscal effects compound.
The 20% Unemployment Scenario
While RAND modeled 10% displacement, the study’s framework makes clear that scaling to higher displacement rates isn’t linear — it’s exponential. At 20% unemployment:
- Payroll tax revenue could drop by 20-40% depending on which income brackets are affected
- If high-earning jobs are automated first (which AI disproportionately targets), the top 10% of earners contribute 34% of federal revenue — a disproportionate hit
- Social Security’s already-fragile timeline shortens dramatically
This connects directly to our previous coverage of Goldman Sachs’ decade-long earnings damage findings and Musk’s universal income proposals. The RAND study provides the fiscal architecture that makes both the doom and the proposed solutions legible.
What RAND Suggests
The study doesn’t recommend a single policy. Instead, it outlines seven intervention categories:
- Reduce federal debt now — so the balance sheet can absorb future shocks
- Shift from labor taxes to capital/corporate taxes — reduce dependence on payroll revenue
- Intentional inflation — maintain nominal prices to preserve tax base (but hurt fixed-income households)
- AI excise taxes — tax compute, electricity, or AI queries directly
- Eliminate corporate tax avoidance — close loopholes that allow AI profits to shift offshore
- Universal income — replace lost labor income with guaranteed transfers
- Nationalize AI monopolies — the most extreme option, potentially requiring constitutional changes
Notably, RAND frames these not as ideological choices but as mathematical necessities. When 84% of your revenue depends on wages and wages are what AI displaces, something has to change.
Why This Matters Now
The RAND study isn’t speculative futurism. It’s a fiscal model built on current tax data and a 10% displacement scenario that’s conservative by 2026 standards. We’ve already documented 78,000 tech layoffs in Q1 2026 alone, and companies like Meta are preparing for workforce cuts of 10-20% specifically to fund AI investment.
The study’s key insight is that the AI jobs conversation has been looking at the wrong scale. Individual career disruption matters. But the structural risk — that AI could undermine the fiscal foundation of the social safety net — is a second-order effect that few are talking about.
Social Security and Medicare don’t fail because politicians vote to end them. They fail because the payroll taxes that fund them dry up when the paychecks disappear.