Rapid AI adoption risks a demand‑driven financial crisis: if AI replaces paid cognitive work faster than income is redistributed or new paid demand appears, falling labour income and compressed intermediary margins can set off a self‑reinforcing collapse in consumption and credit values.

Main Finding

Rapid AI adoption can create a macro-financial stress scenario not primarily through productivity collapse or existential risk but via a distribution-and-contract mismatch: AI-generated abundance reduces the need for human cognitive labor while institutions (wage contracts, credit, consumption patterns, financial intermediation) remain anchored to the scarcity of human cognition. That mismatch can produce a self-reinforcing downward spiral in labor income, demand, and intermediary margins—potentially tipping a stable adjustment into an explosive crisis—unless offset by sufficiently fast reinstatement of human-paid demand or deliberate policy/market responses.

Key Points

• Three formal mechanisms transmit AI adoption into macro-financial stress:
• Displacement spiral with reinstatement effects - Firms rationally substitute AI for labor; aggregate labor income falls, lowering demand and accelerating further AI substitution. - The model identifies key parameters: AI capability growth rate, diffusion speed across firms/sectors, and the "reinstatement rate" (rate at which new demand or paid human roles reappear). - Depending on these parameters the net feedback is either self-limiting (convergent) or explosive (runaway adoption + demand collapse).
• Ghost GDP - When AI output replaces labor-intensive output without proportionate transfers to consumers, monetary velocity falls as the labor share declines. - That produces a wedge between measured GDP (which may rise) and consumption-relevant income (which can fall), creating hidden demand shortfalls—“ghost” output that does not translate into effective purchasing power.

• Intermediation collapse - AI agents reduce information frictions and automate value-added advice and coordination tasks, compressing intermediary margins toward logistics and execution costs. - This reprices business models across SaaS, payments, consulting, insurance, and financial advisory, with knock-on effects for firm valuations and the collateral value underpinning credit markets.

• Distributional concentration amplifies fragility:

  • Top-quintile households account for about 47–65% of U.S. consumption and are also the cohort with the highest measured AI exposure.
  • Therefore shocks to top-income labor/income risk disproportionately affect aggregate consumption and, through that channel, private credit and mortgage markets (estimated exposures cited: ~$2.5 trillion global private credit, ~$13 trillion mortgages).

• Eleven testable predictions are derived, each with explicit falsification conditions (examples below).

Data & Methods

• Model structure

  • A macro-financial stress-test framework couples: (i) a firm-level substitution decision rule (AI vs human labor), (ii) an aggregate demand mapping from labor income and household distributional structure, and (iii) financial-sector balance-sheet propagation channels (credit, mortgages, intermediary margins).
  • Analytical derivations yield conditions (inequalities) on capability growth, diffusion speed, and reinstatement rate that separate convergent from explosive regimes.
  • Ghost GDP is formalized by linking labor share → monetary velocity → consumption-relevant income, showing a monotone relationship absent transfers.
  • Intermediation effects are captured by modelling how reductions in information rents compress intermediary markups and change profitability and asset valuations.

• Calibration and simulation

  • Empirical inputs: FRED macro time series (GDP, labor share, monetary aggregates/velocity proxies, credit aggregates), BLS occupation-level employment and wages, and occupation-level AI-exposure indices.
  • Calibrated parameters include AI capability growth speed, diffusion/diffusivity across firms and sectors, reinstatement elasticity, consumption propensities by income quantile, and leverage/collateral parameters for private credit and mortgage sectors.
  • Numerical simulations produce phase diagrams and time-paths showing when gradual adjustment transitions into explosive demand collapse and financial stress under different parameter combinations.
  • The paper reports quantitative scenarios consistent with observed U.S. consumption concentration (47–65% by top quintile) and maps plausible exposures to the cited credit/mortgage aggregates.

• Falsifiability

  • The authors provide eleven predictions with clear empirical tests. Representative falsification conditions include:
    • If monetary velocity does not decline (or rises) as the labor share falls, the Ghost GDP channel is unsupported.
    • If top-quintile consumption is not concentrated or top incomes are not highly AI-exposed, the transmission into affluent-driven credit markets is overstated.
    • If intermediary margins remain stable despite measurable declines in information frictions, the intermediation-collapse mechanism is falsified.

Implications for AI Economics

• Measurement and monitoring

  • Standard GDP can mask demand shortfalls; central banks and statistical agencies should monitor labor-share–velocity links, distributional income measures, and consumption by income quantile alongside GDP.
  • Financial regulators and macroprudential authorities should incorporate AI-exposure metrics into stress-test scenarios, especially for concentrated exposures in wealthy households and in sectors providing collateral for credit and mortgages.

• Policy responses

  • Compensation/transfers: Policies that translate AI-generated gains into broad-based purchasing power (universal transfers, wage top-ups, targeted subsidies) can raise the reinstatement rate and prevent explosive feedback.
  • Automatic stabilizers and active fiscal policy: Faster or larger fiscal support to displaced workers/households can offset the Ghost GDP wedge and stabilize velocity.
  • Repricing and regulation of intermediaries: Anticipate margin compression across service and financial intermediation; regulation and transitional support (retraining, new licensing) may be needed to avoid sudden asset repricing and credit shocks.
  • Credit-market interventions: Targeted backstops, adjusted capital requirements for exposures with high AI-exposure collateral, and stress-test scenarios that explicitly model distributional demand shocks.

• Research agenda

  • Improve measurement of AI exposure at firm, occupation, and household levels; track diffusion speed empirically.
  • Empirically estimate reinstatement elasticities—how quickly and through what channels displaced workers or firms generate new paid demand.
  • Sectoral studies of intermediary margin compression and the pass-through to credit spreads, valuations, and defaults.
  • Design operational stress-test modules for central banks and bank regulators that incorporate the three mechanisms and the calibrated phase-space results.

Summary takeaway: Rapid AI adoption can create a novel macro-financial fragility rooted in distributional and institutional frictions. The threat is not inevitable, but it is parametrically driven: fast capability growth and diffusion combined with low reinstatement will produce an explosive demand-driven crisis unless countervailing transfers, institutional adaptation, or regulatory measures intervene.