Reforming the Fiscal Architecture for an Automated World

Taxing the Machine: What Must Change in the Tax Code to Fund Aging Societies When AI Does the Work

Three-quarters of U.S. federal revenue comes from taxes on labor. AI doesn't pay payroll taxes. As automation transfers income from workers to capital owners, the fiscal foundations of every developed welfare state begin to crack. This is what a coherent solution actually requires — and why it is politically harder than it sounds.

Analytical Report | The Epoch Times Research & Policy Desk | March 19, 2026 | Sources: Brookings, IMF, CEPR, OECD, Texas A&M, Senate Finance Committee

▶  Bottom Line Up Front (BLUF)

• The core structural problem is simple: About 75% of U.S. federal revenue and roughly 50% of OECD country revenue comes from taxes on labor — payroll taxes and income taxes on wages. AI and automation shift value creation from labor to capital. Capital income is taxed at lower rates, with more opportunities for deferral, avoidance, and offshore sheltering. The welfare state was designed for a labor-income economy. It is being disrupted by a capital-income economy.
• The fiscal gap is already visible: America's billionaires effectively pay approximately 8% on realized and unrealized income. Workers pay 22–37% plus 15.3% payroll taxes. This is not a loophole — it is the intended architecture of a tax code built before AI existed. It is now dangerously misaligned with fiscal sustainability.
• Seven specific structural reforms — each with significant institutional backing — can close the gap without strangling the AI-driven productivity growth that represents the only genuine long-run offset to demographic decline. The reforms must be designed carefully to tax the rents from AI, not the investment in AI.
• The OECD's 15% Global Minimum Corporate Tax (Pillar Two) is already partially implemented in 65 countries. It raises an estimated $155–192 billion annually. It is a necessary but wholly insufficient first step — a floor, not a roof, and subject to ongoing U.S. political resistance under the Trump administration's January 2025 withdrawal.
• A "robot tax" as conventionally imagined — a direct levy on machines — is opposed by the IMF, most OECD economists, and serious tax scholars. The reason is not ideological: it would slow the productivity growth that is the only way aging societies can remain solvent. The right approach taxes AI's profits and rents, not its existence.
• The U.S. faces a unique structural vulnerability: unlike most advanced economies, it has no federal VAT. When labor income falls, VAT revenue in European welfare states partially compensates. The U.S. has no such automatic stabilizer — making its pension and healthcare systems uniquely exposed to a shift from labor to capital income.
• International coordination is essential and currently fragile. Without it, any nation that taxes AI capital more heavily simply exports that capital to lower-tax jurisdictions — the same problem that drove three decades of declining corporate tax rates before Pillar Two. The Trump administration's withdrawal from OECD coordination in January 2025 is the single most dangerous current obstacle to a viable fiscal solution.

Bill Gates articulated the core problem with characteristic directness in 2017: "Right now, the human worker who does $50,000 worth of work in a factory — that income is taxed. You get income tax, social security tax, all those things. If a robot comes in to do the same thing, you'd think that we'd tax the robot at the same level." [1] Gates was not proposing a specific tax mechanism. He was identifying a fiscal architecture failure: the welfare state was funded on the assumption that economic value would continue to flow primarily through human labor and be taxed at the point of employment. That assumption is now being systematically violated.

The Brookings Institution's January 2026 framework paper on AI and tax policy made the structural diagnosis explicit: "The modern tax system in the U.S. rests on two pillars: labor income and, to a lesser extent, consumption. According to 2023 data from a Congressional report, about three-quarters of all U.S. federal tax revenue comes from labor." [2] In the European welfare states — which are even more fiscally dependent on labor through both income taxes and payroll contributions — the equivalent figure is roughly 50%, with VAT providing another 30% that partially tracks consumption rather than labor. Neither architecture was designed for an economy in which a meaningful and growing fraction of value is generated by machines owned by corporations and wealthy individuals.

The CEPR's analysis by economists Bastani and Waldenström identified the three ways AI is reshaping the distribution of income that are most relevant to tax design: a changing wage distribution (some workers becoming dramatically more productive while others fall behind or become unemployed); an increasing share of total income consisting of capital income; and greater dispersion of capital income — meaning the capital gains from AI are concentrated among a relatively small number of asset owners. [3] Each of these trends erodes the existing tax base in a different way, and a coherent fiscal response must address all three.

The Arithmetic of the Current Failure

The scale of the existing fiscal asymmetry between labor and capital taxation is documented and is not in serious dispute. Under current U.S. law, wages are subject to both income taxes (up to 37%) and payroll taxes (15.3% split between employer and employee), for an effective combined rate that can exceed 50% for high earners. Long-term capital gains — the primary form of income for the ultra-wealthy — are taxed at a maximum of 23.8% when realized, and at 0% until realization. Inherited assets receive a "step-up in basis" — meaning capital gains accrued over an entire lifetime are simply forgiven at death. [4]

The White House estimates that America's billionaires effectively pay approximately 8% on their total realized and unrealized income. In some years — as documented by ProPublica's 2021 analysis of leaked IRS data — individuals including Jeff Bezos and Elon Musk paid zero federal income taxes. [5] This is not fraud or evasion. It is the legal architecture of a tax code that taxes transactions (selling assets) rather than accumulation (owning and pledging appreciated assets as collateral for loans). Larry Ellison of Oracle holds over 300 million shares as collateral for personal credit lines, accessing billions in purchasing power without triggering a taxable event. The loan's interest is paid from further borrowing, and the eventual tax may be extinguished by death and step-up basis.

For the Swiss Institute of Artificial Intelligence's January 2026 analysis, the robot-tax-is-no-longer-a-joke assessment quantified the consumption-side exposure: "A 5% decrease in employee compensation relative to GDP over a ten-year compression era could lead to a 1.65 percentage point decrease in spending as a proportion of GDP in steady state." Since VAT revenue is proportional to consumption, that translates directly into a VAT shortfall of comparable magnitude. [6] In the United States — which relies on state-level sales taxes rather than a federal VAT — the equivalent structural shock hits state budgets directly, at precisely the moment when federal programs face their own shortfalls.

Three-quarters of U.S. federal revenue comes from labor. AI does not pay payroll taxes, does not file a W-2, and does not contribute to Social Security. When it replaces a worker, the fiscal foundation cracks.

The Seven Structural Reforms: What Is Needed and Why

The following reforms represent the spectrum from near-term implementable improvements to more fundamental architectural changes. They are organized by urgency and political tractability — not by ideological preference. Each has institutional backing from credible sources; each also has credible objections. The objections are addressed honestly.

Reform 1

Close the
Step-Up Basis
Loophole

The problem: Under current law, when a taxpayer dies holding appreciated assets (stocks, AI companies, real estate), heirs receive the assets at current fair market value with no tax owed on lifetime gains. This "step-up in basis" allows multi-generational wealth compounding entirely outside the income tax system.

The fix: Carryover basis — heirs adopt the decedent's original purchase price, so gains eventually face taxation when the heirs sell. This is how it works in Canada, Australia, and most OECD countries. The Congressional Budget Office estimates eliminating step-up basis raises approximately $105–115 billion over ten years.

Objection: Illiquid assets (family farms, private businesses) may force sales to pay the tax. Addressed by installment payment provisions and exemptions for operating businesses below a size threshold. Most OECD countries manage this.

Status: Proposed in Biden FY2025 budget. Rejected in current political climate. Widely supported by tax economists across the political spectrum as the most obvious technical correction to an acknowledged distortion.

~$110B
over
10 years

Reform 2

Mark-to-Market
Taxation of
Billionaire Income

The problem: Billionaires compound wealth in unrealized capital gains indefinitely, accessing that wealth through collateralized loans without triggering tax. This is not avoidance of a specific rule — it is a feature of a code that taxes realization, not accumulation.

The fix: The Billionaires Income Tax Act (reintroduced September 2025, Senators Wyden, Cohen, Beyer): taxpayers with over $1 billion in assets would pay annual tax on unrealized gains in publicly traded assets, with a 9-year catch-up period for previously untaxed gains. For illiquid assets, a "deferral charge" accrues interest until realization. Applies to fewer than 1,000 taxpayers; estimated revenue: over $500 billion.

Objection: Constitutionality under the "realization" doctrine, administrative complexity in valuing illiquid assets, behavioral avoidance. The bill's architects have built in specific anti-avoidance provisions. The Supreme Court's 2024 Moore v. United States ruling upheld mandatory repatriation without realization, suggesting constitutional headroom exists.

Status: Bicameral legislation introduced September 2025. No prospect of passage in current Senate. Supported by polling (72% of Americans, including majority of Republicans in some surveys).

$500B+
over
10 years

Reform 3

Raise Capital
Gains Rates
Toward Wage
Parity

The problem: Long-term capital gains are taxed at 23.8% maximum; wages at up to 37% plus 15.3% payroll tax. The tax code explicitly rewards passive wealth accumulation over active work. In an economy where AI concentrates value in capital, this asymmetry compounds.

The fix: Brookings economists Clausing and Sarin propose raising long-term capital gains and dividend rates by 5 percentage points — to approximately 28.8% at the top, moving toward but not reaching full wage parity. Combined with the NIIT (Net Investment Income Tax) reform to close the SECA gap, this raises substantial revenue while preserving investment incentives.

The CEPR position: Capital income taxation should be based on "flows (capital income) rather than stocks (wealth)" — taxing capital gains as they accrue rather than levying an annual wealth tax. This is the approach Bastani and Waldenström recommend as most efficient and least distorting to investment.

Revenue: Raising capital gains rates by 5 pp: estimated $200–300B over 10 years. Closing the NIIT gap: additional ~$150B.

Status: Proposed in Biden FY2025 budget. Opposed by current administration.

$200–300B
over
10 years

Reform 4

OECD Pillar Two:
Strengthen &
Extend the
15% Minimum

The problem: Multinational AI companies — Google, Microsoft, Meta, Amazon, Nvidia — book profits in low-tax jurisdictions while their productive activity and customers are in high-tax countries. Before Pillar Two, effective corporate tax rates for the most profitable tech companies were routinely in the single digits.

The status: The OECD's Pillar Two global minimum corporate tax of 15% is now law in 65 countries and took effect in 2024. The OECD estimates it raises $155–192 billion annually in additional corporate tax revenue, a 6.5–8.1% increase in global corporate income tax collection. Approximately 90% of multinationals in scope will be subject to the 15% minimum by 2025.

The problem with the current version: 15% is a floor, not a solution. AI hyperscalers generating extraordinary returns on their IP and compute infrastructure regularly earn effective returns far above what a 15% tax captures. "Excess profit" or "super-normal return" taxation — taxing profits above a risk-free return — is needed to capture the rents from AI market dominance.

The U.S. complication: The Trump administration withdrew from the OECD Pillar Two project via executive order on Day 1 of the new term. A "Side-by-Side" package was agreed January 5, 2026, but the U.S. remains structurally outside the framework, creating competitive pressure that could unravel the entire arrangement.

Status: Live in 65 jurisdictions; U.S. non-participant; fragile international consensus.

$155–192B
annually
globally

Reform 5

Digital Services
Tax / AI Compute
Usage Fee

The problem: AI services consume enormous public infrastructure — electricity grids, broadband networks, publicly funded research, trained workforces. The companies capturing the value from AI were often built on publicly funded foundational research (DARPA, NSF, NIH) without commensurate contribution to the public systems their profitability depends on.

The option: A digital services tax (DST) on revenues from AI-powered services — not on the machines themselves, but on the commercial use of AI in consumer-facing products. Maryland, Texas, and several EU jurisdictions have implemented variants. The Brookings framework notes that "digital services taxes on AI-provided services to consumers represent consumption taxation" — taxing the output, not the investment, thus not slowing the productive deployment of AI.

Alternative framing — compute tax: Tax API calls or GPU compute hours for commercial AI inference. Administratively trackable; proportional to actual productive use. Estimated revenue from a modest $0.001/token equivalent tax on large-scale commercial deployments: $5–15B annually in the U.S. alone at current usage levels, scaling rapidly.

Objection: Risk of dampening AI adoption in jurisdictions with DSTs, putting them at competitive disadvantage. Requires coordination — otherwise capital and services migrate.

Status: State-level implementation in U.S. (Maryland 3% tech services tax, Texas SaaS clarification 2025). Multiple EU DSTs. No federal U.S. equivalent.

$5–15B
U.S. alone
(current; scaling)

Reform 6

Depreciation
Reform: Slow
Accelerated
AI Write-offs

The problem: Current U.S. tax law allows companies to fully and immediately deduct capital expenditures on AI hardware and software — accelerating the cost recovery of investments that generate income over many years. This is an enormous and largely unnoticed subsidy to AI deployment that simultaneously erodes corporate income tax revenue.

The fix: The Swiss Institute of Artificial Intelligence (January 2026) identifies "a small fee on accelerated capital allowances" as a targeted measure that "could increase revenue without halting investment in supporting capital that widens job opportunities." Texas A&M's Bogenschneider proposes a specific robot tax mechanism: reducing the accelerated depreciation available for labor-replacing automation, making the tax deduction for a robot less valuable than for an equivalent wage bill. This doesn't tax robots directly — it removes their preferential treatment relative to workers.

Revenue estimate: Restoring normal depreciation timelines for AI hardware (5-7 year straight-line rather than 100% immediate expensing) raises an estimated $80–120B over 10 years in the U.S., while still allowing full deductibility — just timed to match actual asset life.

Status: Modest provisions were included in pre-TCJA law. The One Big Beautiful Bill Act (2025) restored and extended full bonus depreciation, moving in the opposite direction.

$80–120B
over
10 years

Reform 7

Federal VAT or
Automated-
Economy
Dividend

The problem: The United States is the only major developed economy without a federal value-added tax. In Europe, the 20%+ VAT functions as an automatic stabilizer: even if labor income falls because a worker is replaced by AI, the consumption of the products and services AI produces still generates VAT revenue. The U.S. has no equivalent federal consumption tax that captures value regardless of whether a human or a machine generated it.

The option: A federal VAT — even at a modest 5% rate — would generate approximately $800B–1T annually, completely changing the fiscal math for Social Security and Medicare. Alternatively, a Universal Basic Income (UBI) or "automation dividend" funded by AI productivity taxes could both maintain consumer demand (sustaining the income side of VAT) and provide income floor protection for displaced workers — simultaneously addressing fiscal sustainability and distributional consequences.

The Andrew Yang connection: Yang's 2020 campaign proposal for a 10% VAT plus $1,000/month UBI represented the most coherent synthesis of these ideas to reach a major political platform. Variants are currently being studied in Finland (basic income pilots), Germany (Bürgergeld reforms), and by the OECD.

Objection: VAT is regressive (higher share of income for low-income households) unless accompanied by progressive rebates. Politically near-impossible in the U.S. without framing as an "automation dividend" or equivalent positive-sum narrative.

Status: No current U.S. legislative proposal. Long-run necessity according to Brookings, IMF, and multiple fiscal economists.

$800B–1T
annually
(5% VAT alone)

What the IMF Actually Says: Don't Tax Robots — Tax Their Rents

The IMF's Staff Discussion Note SDN 2024/002, "Broadening the Gains from Generative AI: The Role of Fiscal Policies," arrived at a nuanced position that cuts through much of the popular debate. Its core finding: "The direct policy implication is that there should be no special tax on gen AI, robots, or other forms of labor-replacing technology." [7]

This sounds like a rejection of the robot tax idea. It is not. The IMF's reasoning is that a direct tax on capital investment in automation would slow the productivity gains that are the only genuine long-run path to fiscal sustainability. You cannot afford pensions by making workers less productive. The same report continues: "Yet there can be both efficiency and equity reasons to deviate from this principle" — specifically when AI generates "excess profits" or "super-normal returns" that represent economic rents rather than returns to productive investment.

The distinction matters enormously. A tax on buying a GPU discourages investment and slows productivity. A tax on the excess profits generated when a handful of AI companies achieve market dominance that allows them to charge prices far above their cost of production — that is taxing a rent, not an investment. The latter does not slow AI development. It redistributes the gains from it.

The Tokyo Foundation's analysis of the IMF's 2021 automation working paper identified an additional instrument: a "markup tax on excess profit" that "could enhance economic efficiency" by discouraging the monopolistic market structures that generate those rents in the first place. [8] This is not science fiction — the European Union has deployed windfall profit taxes on energy companies since 2022 using exactly this logic.

The Arguments For and Against a Robot Tax Directly

✓ Arguments For (Direct Robot Tax)

• Creates fiscal symmetry: if a human worker generating $50K is taxed, a machine generating equivalent value should face equivalent contribution to social insurance
• Slows displacement pace, allowing worker retraining to match automation speed
• Revenue is proportional to the very substitution that causes the fiscal damage
• Bill Gates, Xavier Oberson (Taxing Robots, 2019), Ryan Abbott and Bret Bogenschneider (Harvard, 2018), South Korea's partial implementation (2017) all endorse variants
• Signals that society, not just shareholders, must benefit from automation
• Administratively simpler than taxing unrealized capital gains

✗ Arguments Against (Direct Robot Tax)

• IMF, OECD, most academic economists: slows productivity gains that are the only long-run fiscal solution
• Definitional nightmare: what is a "robot"? Is AI software a robot? Is a spreadsheet? Where is the line drawn?
• Capital flight: companies re-domicile or route automation investment through low-tax jurisdictions without coordination
• Lawrence Summers: "profoundly misguided" — discourages the technology that raises living standards
• CEBR (2017): robots contributed 10% of GDP growth in OECD countries 1993–2016; taxing them suppresses that contribution
• Texas A&M's Bogenschneider (2025): even those who proposed it acknowledge "the levy of a robot tax is best viewed as inevitable; the question is only how it will be implemented" — not WHETHER it is good economics

The synthesis position, endorsed by the IMF, CEPR, and Brookings in their most recent analyses, is the following: don't tax the machines; tax the excess profits they generate; tax the capital gains of their owners at rates approaching wage-tax parity; close the inheritance loophole that allows AI wealth to compound dynastically tax-free; implement a global minimum corporate rate high enough to eliminate the incentive to book AI profits in tax havens; and — eventually — implement a broad consumption tax that captures economic activity regardless of whether a human or a machine generated it.

Comparison: Current System vs. Required Reform Landscape

Tax Instrument	Current Rate (U.S.)	Reform Proposal	Estimated Revenue Gain	Political Status
Capital gains (long-term, top)	23.8%	28–33% (+5–10 pp); mark-to-market for $1B+ wealth	$200–800B / 10 yrs	Blocked; Democrat-only support
Step-up in basis at death	Eliminated (0%)	Carryover basis (most OECD standard)	~$110B / 10 yrs	Blocked; proposed Biden FY2025 budget
Corporate minimum tax (global)	15% (OECD Pillar Two, 65 countries); U.S. GILTI imperfect substitute	Raise to 20–25%; include excess-profit tier for AI rents	$155–192B annually (current 15%); more at higher rate	Fragile; U.S. outside framework
AI depreciation (bonus expensing)	100% immediate write-off (One Big Beautiful Bill, 2025)	5–7 year straight-line for labor-replacing automation	$80–120B / 10 yrs	Moving in opposite direction
Digital services / AI compute tax	None (federal); state-level variants	3–5% on AI service revenues; or per-compute-unit fee	$5–15B / yr initially; scaling rapidly	Proposed in 4 states, 2 EU DSTs; U.S. federal opposed
Federal VAT / consumption tax	None (U.S. unique among OECD)	5–10% federal VAT with progressive rebate; or automation dividend	$800B–1T+ annually	No current legislative proposal; long-run necessity
Estate / gift tax	$13.6M exemption; 40% rate above	Return to 2009 parameters: $3.5M exemption, 45% rate	$300–500B / 10 yrs	Exemption doubled by TCJA extensions; moving opposite direction

The Coordination Problem: Why This Cannot Be Done Nation-by-Nation

Every element of the reform agenda above faces a common structural obstacle: capital mobility. If the United States raises capital gains taxes to 33% while the UK remains at 28%, the EU at 26%, and Singapore at 0%, sophisticated investors and the companies they own will route transactions through the lowest-cost jurisdiction. This is not speculation — it is what happened when corporate tax rates diverged from the 1980s through the 2010s, driving effective corporate tax rates in OECD countries from an average of 46% in 1985 to approximately 23% by 2019.

The Pillar Two global minimum corporate tax was the first serious attempt to stop this race to the bottom. Its $155–192 billion in estimated annual revenue gains is meaningful but represents approximately 6.5–8.1% of global corporate income tax revenues — a fraction of what is needed to close demographic fiscal gaps. [9] More fundamentally, the 15% minimum rate was set at a political compromise far below where most economists believe corporate taxation is optimally set for large, highly profitable multinationals generating substantial rents from market power.

The Trump administration's January 2025 withdrawal from the OECD Pillar Two framework — via executive order on Day 1 of the new term — is the single most consequential recent development in global tax policy for the purposes of this analysis. The U.S. hosts the world's most valuable AI companies. Without U.S. participation, the global minimum tax architecture loses its most important participant and creates the conditions for competitive erosion of every other country's implementation. [10]

The January 5, 2026 "Side-by-Side" package agreed by the OECD Inclusive Framework represents an attempt to preserve the architecture while accommodating U.S. concerns about the UTPR (undertaxed profits rule) applying to U.S.-based companies. Whether this holds over a multi-year horizon — given the combination of U.S. political dynamics, tech industry lobbying, and competing national interests — is the most consequential open question in global fiscal policy.

Without international coordination, any country that taxes AI capital more heavily simply exports it. The global minimum tax was the first answer. The Trump administration's withdrawal may be the beginning of its unraveling.

The Distributional Dimension: Taxing AI Must Fund the People It Displaces

The fiscal problem and the distributional problem are the same problem viewed from different angles. AI's productivity gains accrue primarily to capital owners. The workers displaced by AI — or whose wages are suppressed by AI competition — are disproportionately concentrated in the middle of the wage distribution: administrative support, financial operations, paralegal work, software coding at the entry level. These are exactly the workers who have historically funded Social Security through decades of payroll tax contributions and who are now facing displacement before they reach retirement.

The Brookings framework paper identifies the most extreme scenario that current tax architecture fails to handle: if AI systems "become sufficiently intelligent, they might be able to run AI companies by themselves while reinvesting most surplus and therefore generating little taxable profit and, all in all, little tax revenue under our current tax system." [2] This is not a prediction for 2026. It is a stress-test of the architecture — and current architecture fails it completely.

The distributionally coherent solution requires that revenue from AI capital taxation flow specifically toward the systems being fiscally undermined: Social Security, Medicare, pension systems, and transitional support for displaced workers. This means earmarking — something economists generally oppose because it reduces budget flexibility — but which may be politically necessary to build the coalition for reform. Workers will accept a robot tax or a capital gains hike more readily if they can see it sustaining their retirement, not disappearing into a general fund.

The Political Economy: Who Blocks Reform and Why

The reforms described above have broad public support. They face opposition from two concentrated, well-organized interests: technology companies and their investors, and the broader capital-owning class whose effective tax rates would rise. These interests are not identical — a nurse who owns a 401(k) would see capital gains tax changes differently from Elon Musk — but they share an interest in blocking the structural reforms that would redirect AI's gains toward public fiscal systems.

The paradox the Brookings framework paper identifies is that AI itself could help close the political space for resistance: "AI and digital technologies are transforming tax administration, making it easier to track and tax a broader range of economic activity." The IRS's 2026 deployment of AI for audit targeting, fraud detection, and compliance monitoring — even while its human workforce was cut 25% — represents a preview of how tax administration itself can be automated. A well-funded, AI-equipped revenue service is harder to evade than a resource-starved one. [11]

The OECD's September 2025 report on AI in tax administration documented tax authorities using "satellite imagery to identify undeclared property or assets" and AI models to "detect fraud, unreported transactions, and inconsistencies between purchasing and sales ledgers." The gap between what AI can now detect and what tax codes currently require reporting on is closing — creating new enforcement capacity before the legislative reforms are passed.

The Synthesis Verdict: What a Sustainable Fiscal Architecture Looks Like

A tax system capable of funding aging societies in an AI economy requires not one reform but a coherent package that addresses the three simultaneous failures of the current architecture: the labor/capital asymmetry, the internationalization of profit, and the absence of a consumption-based automatic stabilizer. The package, assembled from the most credible institutional proposals, looks like this:

Near-term (2026–2030): Close step-up basis loophole (broad bipartisan technical merit, ~$110B/decade); restore normal depreciation schedules for automation capital (~$100B/decade); strengthen Pillar Two enforcement and raise minimum rate toward 21% (international negotiation); implement modest digital services tax at federal level ($10–20B/yr).

Medium-term (2030–2040): Mark-to-market taxation for billionaire wealth above a threshold (the Wyden-Cohen-Beyer framework or equivalent); raise capital gains rates toward wage parity (5–10 pp); earmark incremental capital income tax revenues to Social Security and Medicare trust funds explicitly.

Long-term (2040+): Federal VAT at 5–10% with progressive household rebate (automatic stabilizer that captures AI-generated consumption regardless of labor displacement); potentially an "automation dividend" — a universal partial basic income funded by the VAT and excess-profits tax, providing a consumption floor for displaced workers and retired citizens alike.

What none of this replaces: Structural reforms to retirement systems — raising retirement ages toward actuarial reality, rationalizing benefit formulas, and eventually means-testing a portion of Social Security for the wealthy — remain necessary. Tax reform can narrow the fiscal gap. It cannot close it alone. The demographic math is too large. The honest answer is that sustainable aging societies in an AI era require all three simultaneously: structural reform of entitlements, tax reform to capture AI's gains, and the productivity growth that AI itself delivers. Any one of these alone is insufficient. All three together may be just enough.

This article is Part IV of a four-part Epoch Times special series on the demographic crisis and technological response. Part I: Immigration cannot solve the crisis. Part II: Country-by-country collapse timelines. Part III: Can AI and automation outrun demographic collapse? Part IV (this article): Tax reform for an AI economy. All analysis draws on primary and peer-reviewed sources as cited. This represents analytical journalism, not tax, legal, or investment advice. The Epoch Times' editorial position is that citizens deserve honest engagement with the fiscal challenges their governments face.

Verified Sources & Formal Citations

Gates, B. "The Robot That Takes Your Job Should Pay Taxes." Interview with Quartz, February 17, 2017. Cited in: Abbott, R. & Bogenschneider, B. "Should Robots Pay Taxes? Tax Policy in the Age of Automation." Harvard Law & Policy Review 12(1), 2018.
https://conferences.law.stanford.edu/werobot/wp-content/uploads/sites/47/2018/02/Should-Robots-Pay-Taxes.pdf

Korinek, A. & Lockwood, L. "The Future of Tax Policy: A Public Finance Framework for the Age of AI." Brookings Institution, January 8, 2026. Based on Brookings working paper of same date.
https://www.brookings.edu/articles/future-tax-policy-a-public-finance-framework-for-the-age-of-ai/

Bastani, S. & Waldenström, D. "AI, Automation and Taxation." CEPR Discussion Paper 19045, 2024. Summarized in: CEPR/VoxEU. "Future Tax Challenges in an AI-Driven Economy."
https://cepr.org/voxeu/columns/future-tax-challenges-ai-driven-economy

Center on Budget and Policy Priorities. "Everything You Need to Know About Proposals to Better Tax Billionaires." March 6, 2024. Detailed analysis of step-up basis, mark-to-market options, and tax treatment of unrealized gains.
https://itep.org/everything-you-need-to-know-about-billionaire-tax-proposals/
See also: Tax Foundation analysis of Biden FY2025 budget proposals.
https://taxfoundation.org/research/all/federal/biden-budget-2025-tax-proposals/

Wyden, R., Cohen, S. & Beyer, D. Billionaires Income Tax Act, reintroduced September 17, 2025. Senate Finance Committee. Citing White House estimate of 8% effective tax rate for billionaires.
https://www.finance.senate.gov/ranking-members-news/wyden-cohen-beyer-introduce-the-billionaires-income-tax-act

Swiss Institute of Artificial Intelligence. "Robot Tax Is No Longer a Joke: Why Revenue Design Must Catch Up with Automation." January 13, 2026. Quantitative analysis of labor income share decline and consumption tax implications.
https://economy.ac/review/2026/01/202601286696

Brollo, F. et al. (led by Dabla-Norris, E. & de Mooij, R.). "Broadening the Gains from Generative AI: The Role of Fiscal Policies." IMF Staff Discussion Note SDN2024/002. International Monetary Fund, 2024.
https://www.imf.org/-/media/Files/Publications/SDN/2024/English/SDNEA2024002.ashx

Tokyo Foundation for Policy Research. "Can a Robot Tax Help Narrow the Social Divide?" Citing: IMF Working Paper 2021/187. "For the Benefit of All: Fiscal Policies and Equity-Efficiency Trade-offs in the Age of Automation." Oberson, X. Taxing Robots. Edward Elgar Publishing, 2019.
https://www.tokyofoundation.org/research/detail.php?id=899

OECD. "Economic Impact Assessment of the Global Minimum Tax." January 2024. Summary: $155–192B additional CIT revenue annually; 6.5–8.1% of global CIT revenues.
https://www.oecd.org/tax/beps/summary-economic-impact-assessment-global-minimum-tax-january-2024.pdf
See also: EY. "OECD Releases Updated Estimates of the Economic Impact of Pillar Two." January 2024.
https://www.ey.com/en_gl/technical/tax-alerts/oecd-releases-updated-estimates-of-the-economic-impact-of-pillar

A&O Shearman. "The Side-by-Side Package and the Global Minimum Tax: What You Need to Know." January 28, 2026. Analysis of U.S. Day-1 executive order withdrawal and January 5, 2026 Side-by-Side package.
https://www.aoshearman.com/en/insights/the-side-by-side-package-and-the-global-minimum-tax-what-you-need-to-know

EisnerAmper. "AI at the IRS: Transforming Tax Enforcement and the Future of Taxpayer Services." January 27, 2026. Covering IRS Agentforce deployment (late 2025), 25% workforce reduction, AI audit targeting.
https://www.eisneramper.com/insights/tax/ai-irs-transforming-0126/
OECD. "Governing with Artificial Intelligence: The State of Play and Way Forward in Core Government Functions." September 18, 2025. https://www.oecd.org

Bogenschneider, B.N. "Robot Tax Proposals for Legislative Review." Tax Notes Federal, November 21, 2025. Comprehensive overview of all current robot tax legislative proposals.
https://www.taxnotes.com/featured-analysis/robot-tax-proposals-legislative-review/2025/11/20/7t92q

Clausing, K. & Sarin, N. "The Coming Fiscal Cliff: A Blueprint for Tax Reform in 2025." Brookings Institution, June 3, 2024. Capital gains rate increase (+5 pp), carryover basis, carried interest closure, FTT, SECA/NIIT reform.
https://www.brookings.edu/articles/the-coming-fiscal-cliff-a-blueprint-for-tax-reform-in-2025/

Frontiers in Artificial Intelligence. "Inclusive Growth in the Era of Automation and AI: How Can Taxation Help?" May 3, 2022. Literature review of robot tax arguments and counter-arguments including CEBR (2017) robot contribution to GDP growth.
https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2022.867832/full

A&O Shearman. "What Does 2025 Hold for the Global Minimum Tax (Pillar Two)?" Covering 90% multinational coverage, U.S. complications, UTPR implementation timeline.
https://www.aoshearman.com/en/insights/what-does-2025-hold-for-the-global-minimum-tax-pillar-two

Sales Tax Institute. "How States Are Expanding Digital Goods & Services Taxation in 2025." November 25, 2025. Maryland 3% tech services tax, Texas SaaS guidance, state-level DST momentum.
https://www.salestaxinstitute.com/resources/expanding-digital-tax-net-digital-goods-services-2025

The Labor-Tax Dependence Problem

U.S. federal revenue from labor taxes~75%

OECD average from labor (income + payroll)~50%

U.S. maximum wage tax rate (income + payroll)~52%

U.S. maximum long-term capital gains rate23.8%

Effective tax rate for U.S. billionaires (White House estimate)~8%

AI companies in scope for Pillar Two 15% minimum90%

Annual global revenue from Pillar Two (OECD est.)$155–192B

U.S. federal VAT rate0%

The U.S. is the only major developed economy with no federal VAT — uniquely exposed when labor income shifts to capital.

The Robot Tax Debate — Key Voices

Bill Gates (2017): Robots replacing $50K factory workers should be taxed like those workers. Sparked current debate.

Ryan Abbott & Bret Bogenschneider (2018): "Should Robots Pay Taxes?" — Yes. Harvard Law & Policy Review. Now the foundational academic proposal.

Lawrence Summers (2017): "Profoundly misguided." Taxing technology that raises living standards contradicts economic logic.

IMF SDN 2024/002: "No special tax on AI/robots" — but taxing excess profits IS appropriate. The nuance that moves the debate.

South Korea (2017): Reduced tax incentive for automation investment — the world's only implemented "robot tax" variant. Not a direct levy; a depreciation reduction.

CEPR (Bastani-Waldenström 2024): Tax capital income flows, not wealth stocks. Raise capital income taxes as capital share of national income rises.

Bogenschneider (Nov. 2025): "The levy of a robot tax is best viewed as inevitable; the question is only how."

Revenue Potential Summary

Step-up basis closure: ~$110B over 10 years

Mark-to-market (billionaires): $500B+ over 10 years

Capital gains rate (+5pp): $200–300B over 10 years

Depreciation reform: $80–120B over 10 years

Pillar Two (current 15%): $155–192B annually globally

Digital services / compute tax: $5–15B/yr U.S. (growing)

Federal VAT at 5%: ~$800B–$1T annually

---

For comparison: U.S. Social Security annual cash deficit by 2035: ~$478B. The step-up + billionaire mark-to-market alone, if enacted, would exceed this figure over a decade.

The Coordination Imperative

Without international coordination, any country raising AI capital taxes faces:

• Corporate re-domiciling to lower-tax jurisdictions

• Transfer pricing manipulation routing AI IP profits offshore

• Competitive disadvantage in attracting AI investment

This is the same dynamic that drove corporate taxes from 46% (OECD average 1985) to 23% (2019) before Pillar Two.

Current status: 65 countries implementing Pillar Two. U.S. withdrawn via Day-1 executive order (Jan. 2025). OECD Side-by-Side package agreed Jan. 5, 2026 — fragile compromise.

Stakes: Without U.S. participation, tech companies headquartered in the U.S. — generating the majority of global AI profits — remain outside the framework.

What Other Countries Have Done

South Korea (2017): Reduced tax deduction for automation investment from 7% to 2%. Only implemented variant of robot tax anywhere in the world. Modest revenue; high symbolic significance.

EU (2022–present): Windfall profits tax on energy companies at above-normal returns. Exact model for AI excess-profit taxation. Raised ~€100B.

EU ViDA (2025–26): VAT in the Digital Age — mandatory e-invoicing, digital reporting requirements across EU. Real-time VAT tracking for AI transactions.

Maryland (2021/2025): Digital advertising tax (3%); tech services tax expansion. Multiple legal challenges.

UK: Digital Services Tax at 2% on revenues of large tech platforms. Generating ~£700M annually.

France: 3% DST on digital service revenues. ~€400M/yr; subject to U.S. trade pressure.