Skip to content
Briefings are running a touch slower this week while we rebuild the foundations.See roadmap
AI: Jobs, Power & Money
17MAR

AI could hollow out the US tax base

3 min read
13:50UTC

RAND and Brookings warn AI displacement will erode the tax base funding 84–85% of federal revenue. Anthropic's CEO and Andrew Yang agree: tax robots, not labour. The IRS has lost a quarter of its staff.

← Back to Meta cuts 20% while Big Tech spends $650bnJump to analysis ↓
EconomicAssessed
Key takeaway

AI deflation could erode the tax base precisely as government retraining costs surge. Amodei's endorsement transforms AI taxation from outsider advocacy into an intra-industry debate.

Roughly 84–85% of US federal revenue derives from labour income — individual income tax and payroll taxes combined — according to both the RAND Corporation and The Brookings Institution 1 2. Every displaced worker who moves from a $90,000 salary to unemployment insurance represents a lost revenue stream on both sides of the ledger.

RAND modelled a scenario in which AI is priced at marginal cost. The result was deflation, as the cost of AI-substitutable services collapses while displaced workers reduce spending 3. The US federal debt stands above $36 trillion. Even moderate deflation increases its real burden while shrinking the tax base that services it.

Brookings warned that "government revenues from payroll taxes as a fraction of GDP will decline just as needs for retraining programmes and transition support increase" 4. If AI displaces 2–3% of the labour force over five years — well within Goldman Sachs's estimate of 1–4 million US jobs annually — the annual payroll tax shortfall runs into tens of billions. Social Security and Medicare face accelerated insolvency timelines.

Anthropic CEO Dario Amodei urged governments to tax AI-generated wealth: "There is so much money to be made with AI — literally trillions of dollars per year" 5. Andrew Yang renewed his proposal in March to "stop taxing labour and start taxing AI," citing Amodei's support 6. Yang's example: replacing a $28-per-hour housekeeper with a $2-per-hour robot produces a tax gap no existing mechanism fills. Amodei's endorsement carries strategic logic — a uniform tax regime protects incumbents against competitors who externalise displacement costs onto public budgets.

The IRS has lost roughly 25% of its workforce since January 2025, according to the Treasury Inspector General 7. The agency tasked with collecting revenue is being hollowed out while the revenue base it collects from faces structural erosion.

Deep Analysis

In plain English

The US government funds most of its spending — Social Security, Medicare, defence, infrastructure — primarily through taxes on wages. When you earn money from work, a share funds these programmes directly. This differs from investment income, which is taxed at lower effective rates and is easier to shelter legally. If AI displaces workers at scale, the wage pool shrinks and so does that tax revenue. Simultaneously, more displaced workers need support: unemployment benefits, retraining, healthcare subsidies. RAND warns this could also trigger deflation — if AI makes goods and services dramatically cheaper, prices fall. Falling prices make government debt harder to repay, because debt is fixed in dollar terms while tax revenues decline. This is a compounding problem: less revenue coming in, more need going out, and harder debt to service simultaneously.

Deep Analysis
Synthesis

The fiscal risk is self-reinforcing in a way the body does not articulate. The $650–690 billion AI infrastructure investment generates corporate profits and capital gains — both taxed at substantially lower effective rates than wage income in the US system. The shift from labour to capital income simultaneously displaces workers and reduces the effective tax yield on the income replacing them. This creates a structural double shortfall: fewer workers paying high labour tax rates, and more capital income taxed at preferential rates — a dynamic that conventional fiscal projections treating AI investment as growth-positive routinely underweight.

Root Causes

The 84–85% labour income dependency reflects a deliberate post-WWII policy architecture: funding the welfare state through payroll taxes kept the system politically popular by creating visible contribution-benefit links. But it built structural fragility into the fiscal model if the labour income share of GDP ever declined materially. Capital's rising share of national income — documented consistently since the 1980s by Piketty and confirmed by post-2008 data — had already been eroding this foundation before AI arrived. AI accelerates a pre-existing structural trend rather than creating a novel fiscal vulnerability.

Escalation

The IRS losing 25% of its workforce since January 2025 creates a compounding enforcement gap this event's body does not address. Reduced audit capacity, declining voluntary compliance, and a structurally eroding payroll tax base simultaneously undermine federal revenue from three distinct directions. This combination — structural AI displacement, cyclical economic slowdown, and institutional enforcement capacity loss — is analytically more severe than any single cause implies, and more urgent than the medium-term framing of RAND and Brookings suggests.

What could happen next?
2 risk1 consequence1 meaning1 precedent
  • Risk

    AI-driven deflation could compound federal debt-service costs by compressing the nominal GDP base against fixed nominal debt obligations, replicating Japan's Lost Decade dynamic at greater scale.

    Medium term · Suggested
  • Consequence

    Accelerated AI capital expenditure depreciation reduces near-term corporate tax revenue simultaneously with payroll tax base erosion — a fiscal double hit that standard growth models underweight.

    Short term · Assessed
  • Risk

    Compounding IRS enforcement capacity loss with a structurally eroding tax base creates a revenue shortfall more severe than either cause independently implies.

    Short term · Assessed
  • Meaning

    The 84–85% labour income dependency reveals the US fiscal model was not designed for an economy where capital replaces labour at scale — adequate reform is structural and multi-decade, not incremental.

    Long term · Assessed
  • Precedent

    Japan's Lost Decade demonstrates that deflation combined with high nominal debt creates fiscal traps that conventional stimulus cannot resolve — a structurally analogous dynamic the US may enter if AI displacement accelerates.

    Long term · Suggested
Sources:RAND Corporation·RAND Corporation·Brookings
Mentions:Dario Amodei →Andrew Yang →Goldman Sachs →Brookings Institution →
First Reported In

Update #1 · Meta cuts 20% while Big Tech spends $650bn

RAND Corporation· 17 Mar 2026
Read original →
Different Perspectives
UK financial regulators (BoE FPC / FCA)
UK financial regulators (BoE FPC / FCA)
The Bank of England's April FPC directive on agentic AI in payments was scoped around one frontier model; AISI confirmed a second model cleared the same 32-step threshold on 1 May. The supervisory architecture is one model behind the capability it was built to contain.
Indian IT sector workers (TCS, Infosys, Wipro)
Indian IT sector workers (TCS, Infosys, Wipro)
TCS posted its first annual revenue decline in the modern era, Infosys shed 8,400 workers in a quarter, and Wipro hit its zero-fresher target. Western Big Tech's AI automation is cannibalising the offshored-services model that employs roughly five million Indian IT workers.
Chinese workers (Hangzhou and Beijing plaintiffs)
Chinese workers (Hangzhou and Beijing plaintiffs)
Workers Zhou and Liu won cases that established a two-court doctrinal chain: AI adoption is the employer's deliberate strategy, placing the cost of displacement on the employer rather than the worker. Any Chinese employee facing AI-driven dismissal now has a citable legal route that American, British, and European counterparts do not.
Chinese government, courts, and domestic employers
Chinese government, courts, and domestic employers
The Hangzhou rulings were released on Workers' Day eve alongside the Ministry of Human Resources' recognition of 42 new AI occupations. Domestic firms now face mandatory retraining obligations; the Orgvue estimate of 8-14 months added to displacement timelines will feature in employer compliance briefings throughout 2026.
EU regulators and European Parliament
EU regulators and European Parliament
The second Digital Omnibus trilogue collapsed without agreement on 28 April; the third is scheduled for 13 May with the binding employer AI-literacy obligation still contested. Brussels is arguing over a non-binding encouragement clause while Beijing's courts have already bound employers.
US legislators (Warner, Rounds, Hawley, Sanders)
US legislators (Warner, Rounds, Hawley, Sanders)
Warner and Rounds produced the Economy of the Future Commission Act, the most concrete federal vehicle still moving, endorsed by the companies it would notionally regulate. The Sanders-AOC moratorium was killed by Democratic senators; the Hawley-Warner disclosure bill remains in committee with no floor date.