Wealth Concentration Is Manufacturing a Population Without Income

Wealth concentration has stopped being a distribution problem and become a production problem. Brookings describes a global elite of roughly 500,000 people, each holding more than $30 million, that controls about $60 trillion, roughly one-eighth of a global wealth stock that UBS places near $450 trillion. This same group dominates ownership of the firms building the AI economy. When investment sits in that few hands, it flows toward automation-heavy, capital-light sectors rather than labor-intensive ones. Each dollar of investment now generates fewer jobs, more productivity, and no corresponding wage gain.

The consequence is a rising share of workers who cannot secure stable income. This is not the familiar story of inequality as a fairness issue. It is a change in how economies generate jobs at all. AI-driven productivity grows fastest in the sectors the ultra-wealthy own, which raises a question Brookings poses directly: will there be enough demand for goods and services when wealth is this concentrated?

The feedback loop closes fast. Lower wages cut demand. Weaker demand cuts hiring. Less hiring pushes more people to zero income. Economists describe this as demand-side collapse, and unequal economies carry it as a standing risk.

No single source publishes a 2050 headcount of people without income, so we assemble it from the pieces that do exist. UN projections put global population at about 9.7 billion in 2050. ILO data place the share of workers in routine or low-skill roles, the roles automation displaces first, at roughly 45 to 55 percent of the labor force. Economic models of AI-driven automation show that workers below a skill threshold absorb absolute welfare losses even as the economy grows, and IMF work finds that AI adoption widens wealth inequality because capital owners capture most of the gains. NBER research reaches the same conclusion from the wage side: automation leaves wages and incomes stagnant at the bottom of the distribution.

Combine those pieces and the estimate emerges. If 15 to 20 percent of the global labor force ends up displaced or earning stagnant, unreliable income, the affected population is 9.7 billion multiplied by 0.15 to 0.20, a band between 1.5 and 2.0 billion people. This is a conservative inference, not a forecast. It counts workers displaced by automation, workers who cannot transition to higher-skill roles, workers in economies with thin safety nets, and informal-sector workers whose income collapses under a global shock. Some models imply the displacement runs higher.

The variable to watch is consumption inequality, not income inequality. Global consumption inequality has fallen over the past two decades, but within-country inequality keeps rising. The distinction matters because people without income cannot consume, and weak consumption undermines the whole economy. Governments then absorb a rising social-welfare burden at the same moment their tax bases narrow.

This is how a rich economy produces a poor labor market. The end state is a dual economy: a small asset-owning class benefits from AI while a large population works sporadically or not at all, chronic underemployment becomes normal, and governments intervene not to grow but to prevent collapse. The wealth-concentration and labor-market data already point in that direction, not speculation.

The strongest opposing read holds that this is another round of a familiar cycle. Every wave of automation destroyed specific jobs and created more in the aggregate. The loom, the assembly line, the spreadsheet: each one raised total employment, eventually.

The evidence says this round differs in three ways. First, the labor share of income has been falling for decades across advanced economies, which means productivity gains have stopped translating into wages the way they once did. Second, the displacement is not a slow generational churn; AI compresses the window in which a displaced worker can retrain. Third, the jobs that automation creates skew toward capital and toward a small number of highly skilled roles, not toward the routine work it eliminates. The “eventually” that rescued earlier transitions is exactly what has narrowed.

For one person the strategy compresses to a sequence. Earn portable income first: remote, digital, or gig work that does not depend on a single local employer. Compress the cost floor second: relocate or restructure spending toward the $500 baseline so the income target shrinks. Build a runway third: six to twelve months of expenses converts a job loss from an emergency into a decision. Reskill continuously fourth, into roles automation complements rather than replaces. None of this substitutes for the policy levers; it is what a person does while waiting for them.

“No income source” is a polycrisis amplifier. It compounds with climate displacement, fiscal strain, and political instability: a population with nothing to lose does not absorb a shock, it transmits it. For development economics the implication is a shift in the dependent variable. The question stops being how to create jobs and becomes how to keep a population solvent when jobs stop being the default. The D-coefficient, how much disruption a system can absorb before degrading, applies directly: a society’s capacity to absorb a labor shock now depends on how cheaply and how portably its people can earn.