Part Three of a Series on the Coming Labour Disruption
When the Money Runs Out: The Coming Collapse of the Global Tax Base
The Fiscal Foundation and Its Flaw
Modern states are, at their core, fiscal machines. They provide services—education, healthcare, infrastructure, defence, social support—in exchange for revenue extracted from economic activity. The architecture of that revenue system was designed around a specific and now-challenged assumption: that most economic value flows through human labour, and that labour can be taxed where it is performed.
Income tax, payroll tax, social security contributions, sales taxes paid by wage earners—the vast majority of government revenue in every OECD country is directly or indirectly tied to employment income (Cognizant, 2026). The system works when labour captures a significant share of economic value. It begins to fail when that share contracts structurally. And it faces something close to catastrophic stress when the contraction is simultaneous, global, and permanent rather than cyclical.
That is precisely the scenario that credible projections of AI-driven displacement now describe.
The Scale of What Is Coming
The displacement projections from major institutions are not fringe estimates. They are the considered outputs of the world’s leading economic research organizations, and they describe a labour market transformation of historic magnitude.
McKinsey estimates that between 400 and 800 million workers worldwide could be displaced by automation by 2030—equivalent to the entire current labour force of the United States and European Union combined (Facebook/EconomicsForTheMany, 2026). Goldman Sachs Research estimates 300 million jobs globally are exposed to AI-driven substitution over a ten-year horizon (Goldman Sachs, 2026). The IMF’s Managing Director Kristalina Georgieva has compared AI’s employment impact to a “tsunami,” estimating that 40% of jobs worldwide—and 60% in advanced economies—will be affected, either enhanced, replaced, or fundamentally reshaped (IMF via Times of Israel, 2025). The Boston Consulting Group estimates AI could displace up to 25 million jobs in the United States alone, with the newly appointed CEO of Verizon forecasting U.S. unemployment rates of 20% to 30% within two to five years—figures that would exceed the peak unemployment recorded during the Great Depression (Yahoo Finance, 2026).
Cognizant’s 2026 analysis found that 93% of jobs are currently exposed to some degree of AI-led automation, with exposure scores climbing at 4.5 times the previously anticipated rate and the share of jobs with minimal exposure collapsing from 31% to 7% (Cognizant, 2026). This is not a distant horizon problem. The displacement is accelerating now.
The upper bound of these projections approaches 900 million displaced workers globally when secondary job losses—the trades, public sector, and service roles that depend on a functioning middle-class consumer economy—are included alongside direct AI substitution. At that scale, the fiscal consequences are not a policy challenge. They are a civilizational one.
The Fiscal Scissors, Writ Global
The mechanism of fiscal collapse under mass displacement is not complicated. It is simply the simultaneous operation of two forces that have historically never converged at this scale or speed.
On the revenue side: the employment base that generates income tax, payroll tax, and consumption tax contracts sharply and permanently. Unlike a recession, in which displaced workers eventually find new employment and tax contributions resume, AI-driven displacement removes roles structurally. The revenue does not return when the “downturn” ends because there is no downturn—there is a permanent reduction in the number of humans whose labour generates taxable income.
On the expenditure side: the demand for public services rises sharply at exactly the moment revenues fall. Unemployment benefits, healthcare demand driven by economic stress and declining mental health, social housing pressure, retraining programs, child poverty interventions—all of these increase as mass displacement accelerates. And underneath both of these, immovable and non-negotiable, sits the existing debt load.
Most Western governments have operated in a state of quasi-austerity for nearly two decades. Despite years of spending restraint, debt-to-GDP ratios in major economies remain historically elevated. The United States currently spends more on debt interest payments than on national defence—and that interest bill is paid before any other obligation, in full, regardless of what is happening to the tax base. When revenues contract, debt service does not. It remains first in the queue. What compresses is everything else: education, healthcare, social support, infrastructure, public employment. The compression will be brutal, and it will be experienced everywhere simultaneously.
The Corporate Tax Minimization Layer
Even the economic value that continues to be generated—by the AI systems, by the corporations deploying them—will be aggressively shielded from the taxation needed to replace lost labour revenue. This is not speculation. It is the established and well-documented practice of the organizations that will own AI’s productivity gains.
Large multinationals already pay effective corporate tax rates substantially below statutory rates across OECD countries. The average statutory corporate tax rate across OECD countries is 20.2%, but large multinationals regularly pay effective rates well below this through profit-shifting, transfer pricing, IP holding structures, and treaty shopping (AGI Social Contract, 2026). Apple and Google currently pay roughly 15.6% in effective corporate tax rates globally—below the headline rate and, in the case of most countries where they operate, contributing almost nothing at all. Google pays over 80% of its corporate taxes in the United States despite generating less than 50% of its revenue there, contributing vanishingly little to the fiscal systems of the countries where most of its customers live (AGI Social Contract, 2026).
In 2025, corporations earning over $105 billion in collective pretax income not only paid no federal income tax but received $4.7 billion in tax rebates—a total tax break of $26.7 billion in a single year (Thomson Reuters, 2026). These are not marginal actors. They are profitable, large, and sophisticated organizations deploying mature legal strategies to minimize contributions to the public finances.
The companies that will own AI’s productivity gains are precisely these organizations—or their successors—with the largest legal departments, the most sophisticated tax minimization infrastructure, and the deepest experience navigating international tax arbitrage. The value generated by replacing human labour with AI will not automatically flow into the fiscal systems that need it. It will flow to shareholders, to retained earnings, and to the jurisdictions and structures best positioned to hold it at lowest tax cost.
The Proposed Solutions and Their Limits
The policy discussion around AI’s fiscal consequences is maturing, but it remains far behind the pace of the problem. The main proposals in circulation each face serious structural objections.
Automation or robot taxes, most prominently proposed by Bill Gates in 2017, would levy a charge on companies for deploying AI or robotics in place of human workers (McGill Law Journal, 2019). The appeal is intuitive: if a machine replaces a worker, the machine should contribute what the worker would have contributed in taxes. The practical difficulties are severe. Defining “a robot” in an era of software-based AI is extraordinarily difficult—is a customer service chatbot a robot? Is a code-generation tool? The definitional boundary is arbitrary and gameable. More fundamentally, automation taxes create incentives to deploy AI through cloud services from low-tax jurisdictions, converting a domestic tax problem into an international arbitrage problem. Most economists conclude that automation taxes are probably not the best funding mechanism, even if the underlying redistributive goal is sound (McGill Law Journal, 2019).
Wealth taxes face capital flight, valuation challenges for illiquid assets, and the political reality that the people most able to shape tax policy are precisely those most affected by wealth taxes. They have been attempted in several European countries and repeatedly abandoned or scaled back.
Universal basic income at meaningful scale requires a tax base capable of funding it—the very base that is contracting. Billionaire investor Vinod Khosla proposed that if 125 million Americans are displaced by AI, they should be relieved of income tax obligations, with the revenue gap covered by higher capital gains taxes (Forbes, 2026). The arithmetic is not unreasonable in a scenario where capital gains are rising sharply as labour income falls—but the political conditions required to implement a large-scale capital gains tax increase are, to put it gently, challenging in most Western democracies.
Sovereign AI funds, modelled on Norway’s oil fund, would require governments to have acquired significant ownership stakes in AI infrastructure before the value accrued. The window for most governments to do this has likely already closed.
International tax coordination is the only mechanism that could close the arbitrage gap that allows AI profits to be sheltered across borders. The OECD’s global minimum corporate tax—the most significant international tax coordination effort in decades—took years to negotiate and is being partially undermined by jurisdictional non-compliance almost from the moment of implementation. International coordination on a problem as complex and fast-moving as AI taxation operates on decade-scale timescales. AI capability is advancing on month-scale timescales. A progressive global corporate tax for the age of AI has been proposed (AGI Social Contract, 2026), but its implementation would require the kind of international political will that has not been demonstrated even on simpler problems.
The Developing World: A Different and Darker Problem
The fiscal crisis described above assumes a government with at least some existing tax infrastructure, some institutional capacity to respond, and some buffer of existing public services to compress. For much of the developing world, those assumptions do not hold.
AI displacement in the developing world presents a paradox. On one hand, only 7% to 14% of workers across Latin America and the Caribbean hold jobs that can directly benefit from GenAI adoption (Brookings Institution, 2026). The infrastructure for AI-assisted productivity gains simply does not exist for most workers in lower-income economies. On the other hand, 1% to 6% of jobs in those same economies face a high risk of direct automation—disproportionately in banking, finance, public sector administration, and customer support: the very middle-class roles that represent economic progress for women and young workers (Brookings Institution, 2026).
Previous waves of automation allowed developing economies to absorb manufacturing and services work as it was priced out of higher-wage markets. That relief valve is closing. AI-native competitors can deploy in any jurisdiction at minimal marginal cost, undermining the labour cost advantage that has historically driven development. The developing world will not benefit from a “next wave” of work flowing downmarket. It will face the same displacement as advanced economies, with less fiscal capacity to absorb the consequences, less institutional resilience to manage the transition, and less political stability to contain the social fallout.
The Timing Problem: Why Policy Cannot Keep Up
The cruelest dimension of the coming fiscal crisis is not its scale but the mismatch between the speed of fiscal collapse and the speed of any credible policy response.
Tax policy operates on multi-year legislative cycles. International coordination operates on decade-scale timescales. AI capability is advancing on month-scale timescales. By the time governments accurately understand what is happening to their revenue base, acknowledge it politically, build a legislative coalition to respond, design a new fiscal architecture, negotiate international agreement to prevent arbitrage, and implement the resulting framework—the displacement wave will be substantially complete. The fiscal regime change will not be managed. It will be absorbed, chaotically, by the institutions and individuals least equipped to absorb it.
This is not pessimism for its own sake. It is a structural observation about the speed mismatch between technological change and institutional response. Democratic governments are designed to move slowly, to build consensus, to avoid dramatic reversals. These features are virtues in stable conditions. They are liabilities in conditions of rapid structural disruption. There is no version of this transition in which policy arrives on time.
What This Means for the Individual
The fiscal consequences described in this article are not abstract. They will manifest in the lived experience of every person in every country in specific and concrete ways: public services cut to honour debt obligations, healthcare systems under funding pressure, education budgets constrained, social supports strained beyond capacity, and the social contract between citizens and states under unprecedented stress.
None of this is inevitable in its worst form—but avoiding the worst outcomes requires action that is not currently occurring at the necessary scale or speed. And in the absence of systemic solutions, the burden falls, as it always does, on individuals.
The individuals best positioned to navigate this transition are not those with the most credentials or the most technical skills—both of which can be rapidly obsoleted. They are those who have invested in the capabilities that function across contexts: the ability to analyze complex and rapidly changing situations, to frame problems that have not yet been named, to adapt under uncertainty, to create value in circumstances that no prior training specifically addressed. These are not job skills. They are human capabilities that remain relevant regardless of what the fiscal and labour market landscape looks like—because they are the capabilities that help people reshape their circumstances rather than merely endure them. These are capabilities that are in dire short supply, both at a societal level and for individuals.
The tax base may collapse. The organizational promises may prove hollow. The policy responses may arrive too late. None of those outcomes removes the agency of an individual who has invested, deliberately and in advance, in becoming the kind of person that disruption of any kind cannot make redundant.
That investment is the only one that cannot be taxed away, automated away, or restructured away by a board of directors that has discovered a cheaper alternative.
References
AGI Social Contract. (2026, February 11). A progressive global corporate tax for the age of AI. https://www.agisocialcontract.org/anthology/a-progressive-global-corporate-tax-for-the-age-of-ai
Brookings Institution. (2026, March 1). Artificial intelligence is transforming middle-class jobs. Can it also help the poor? https://www.brookings.edu/articles/ai-transforming-middle-class-jobs-can-it-help-the-poor/
Cognizant. (2026, April 8). AI job displacement: When capital can think, who pays? https://www.cognizant.com/us/en/insights/insights-blog/ai-impact-on-jobs-and-tax-rebalancing
Forbes. (2026, February 17). If 125 million are unemployed by AI they shouldn’t pay taxes. https://www.forbes.com/sites/maryroeloffs/2026/02/17/billionaire-khosla-if-125-million-are-unemployed-by-ai-they-shouldnt-pay-taxes
Goldman Sachs. (2026, March 17). How will AI affect the U.S. labor market? https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-us-labor-market
McGill Law Journal. (2019). I robot: U tax? Considering the tax policy implications of automation. https://lawjournal.mcgill.ca/article/i-robot-u-tax-considering-the-tax-policy-implications-of-automation/
Thomson Reuters. (2026, April 15). Experts find profitable corporations paid little or no federal income tax in 2025. https://tax.thomsonreuters.com/news/experts-find-profitable-corporations-paid-little-or-no-federal-income-tax-in-2025/
Yahoo Finance. (2026, April 20). AI could destroy 25 million U.S. jobs. https://finance.yahoo.com/economy/policy/articles/ai-could-destroy-25-million-141527754.html
World Economic Forum. (2025). The future of jobs report 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/