How America’s tax code subsidises artificial intelligence

The article argues that AI-driven substitution of software for workers could erode the US federal tax base, which derived 84% of fiscal-2025 revenue from individual income and payroll taxes, versus less than 9% from corporate income taxes. It cites roughly $13 trillion in wages and salaries against $4.1 trillion in corporate profits in 2025, while federal corporate-tax receipts were about $452 billion compared with $4.4 trillion from individual taxes and social-insurance contributions. The author contends that the 7.65% employer payroll-tax charge, within a 15.3% combined Social Security and Medicare levy, makes human labor structurally more expensive than AI and calls for lower labor taxes and a broader, more effective tax base for corporate profits and capital income.

Analysis

The investable implication is not an imminent “AI tax,” but a widening political-risk discount between labor-saving AI beneficiaries and firms whose earnings depend on labor intensity. Hyperscalers (MSFT, AMZN, GOOGL, META) and enterprise software vendors (NOW, CRM, PLTR) retain the near-term operating-leverage upside from automation, while staffing and outsourcing models (RHI, ASGN, KFY, G) face a more durable volume risk if customers redesign workflows rather than merely pause hiring. The second-order winner is data-center infrastructure—NVDA, AVGO, VRT, ETN, CEG—because a shift from payroll to compute raises capital and power intensity even where aggregate employment remains stable.

Over 1-3 months, this is primarily a policy-headline and valuation risk, not an earnings event: broad corporate-tax reform requires legislative alignment and is unlikely to be priced efficiently from an academic-policy debate alone. Over 6-18 months, rising deficits alongside resilient nominal GDP would increase the probability of revenue measures targeting corporate effective tax rates, buybacks, foreign profit treatment, or accelerated depreciation. The most exposed stocks are those priced on distant cash flows and carrying unusually low cash-tax rates; the relevant diligence metric is cash taxes paid as a percentage of pretax income, not statutory tax rates.

Consensus may overstate the direct threat to AI leaders. Policymakers have stronger incentives to preserve domestic compute investment and strategic competitiveness than to penalize AI deployment directly; any eventual reform is more likely to broaden corporate taxation than impose a per-worker automation levy. That would compress after-tax FCF across sectors, but labor-saving adopters could still outperform labor-intensive peers if their pre-tax margin expansion exceeds the tax drag. The thesis is falsified if employment-intensive sectors reaccelerate hiring without margin pressure, or if fiscal consolidation reduces revenue-raising urgency.

Key Decisions for Investors

• Maintain a 6-12 month long basket of AI infrastructure (NVDA, AVGO, VRT, ETN) versus a short basket of professional staffing/outsourcing exposure (RHI, ASGN, KFY): the pair targets structural substitution rather than a directional technology beta. Reassess if staffing firms show two consecutive quarters of organic revenue acceleration without gross-margin erosion.
• Avoid adding to high-multiple enterprise AI software solely on automation narratives until cash-tax-rate exposure is mapped. Create an alert for companies where cash taxes are below 15% of pretax income or where accelerated-depreciation benefits drive a material share of FCF; these names have asymmetric downside under corporate-base broadening.
• For diversified equity exposure over the next 1-3 months, favor an equal-weight AI-capex basket over labor-intensive services rather than a broad long technology trade. The risk/reward is better because it captures rising compute and electrification spend while reducing sensitivity to a corporate-tax multiple reset.
• Monitor federal revenue legislation, Treasury proposals, and deficit projections as 6-18 month catalysts. A credible proposal to raise the effective corporate burden or curtail depreciation should trigger profit-taking in capital-intensive AI beneficiaries first, while retaining relative shorts in staffing and outsourced-services models.