Debt service, revenue, and borrowing costs are coupled through standard macro mechanisms: the Phillips curve, Fed reaction, fiscal response. Most shocks move them in offsetting directions, which is why fiscal cascades are rare. The question is what happens when a single shock moves them together instead.
A single ratio: federal debt service divided by available revenue, where the revenue itself shrinks as unemployment and inflation rise. Move the sliders to see how the ratio responds under any combination of the three variables, at any horizon.
AI displacement is the first shock that plausibly overrides those offsetting mechanisms and drives the three together. Adoption sits in single digits today. If it scales as the labor-exposure research suggests, the cascade the model describes moves on a timeline of years, not decades.
Under ordinary conditions, unemployment, inflation, and interest rates are coupled through standard macroeconomic mechanisms — the Phillips curve, the Federal Reserve's reaction function, and fiscal-policy feedback — and those mechanisms move the variables in the canonical inverse pattern. Recessions raise u and lower i; supply shocks raise i without raising u; rate cycles trail both with a lag. The fiscal arithmetic absorbs each shock in turn because the canonical mechanisms drive the variables in offsetting directions.
What changes the picture is a shock that overrides those canonical mechanisms — one that pushes all three variables in the same direction at the same time through a different coupling pathway. AI-driven cognitive displacement is the first plausible such shock in modern macroeconomic history. The model below shows what the fiscal math does if it fires.
The ratio fell year-over-year because revenue growth from customs duties (+$118B) and individual income tax (+$230B) outpaced the rise in interest costs. Roughly half of the customs duty growth came from tariffs imposed under the International Emergency Economic Powers Act, which the Supreme Court ruled unlawful in Learning Resources v. Trump (February 20, 2026). The refund liability is unresolved but potentially in the $100–200B range. The favorable revenue dynamic was therefore partly produced by a fiscal action subsequently ruled to exceed statutory authority. The cascade scenarios below describe what happens when the favorable dynamic reverses; the structural fragility this represents is exactly what the model is designed to make legible. Sources: Treasury MTS; Joint Economic Committee, U.S. Senate (October 2025); Supreme Court, Learning Resources, Inc. v. Trump, February 20, 2026; Penn Wharton Budget Model, "IEEPA Revenue and Potential Refunds" (February 2026).
Receipts in the first seven months of FY2026 are $200B higher than the same period of FY2025, despite the IEEPA ruling. The favorable dynamic is continuing in part because Section 301, Section 232, and other statutory tariffs remain in place, and because nominal growth continues to push payroll and individual income tax receipts upward. The ratio has now fallen three years in a row (0.213 → 0.198 → 0.190). The model's argument is that this trajectory is conditional on conditions that have not yet reversed. Sources: CBO Monthly Budget Review (May 2026); Bipartisan Policy Center Deficit Tracker; BLS Employment Situation, April 2026.
The Crisis Ratio is computed at a point in time using current inputs, but debt accumulates: every year that conditions persist, debt held by the public grows by roughly $2 trillion in baseline conditions and significantly more under stress. Use the horizon control to see how the same configuration evolves when conditions don't reverse. The static ratio shows the present state. The 3-year ratio shows the cascade.
Debt service consumes 21% of federal receipts — inside the sustainable band but approaching the IMF-style stress threshold of 0.30. AI displacement has not yet materially begun; the canonical macroeconomic mechanisms still govern.
Coefficients (α = 0.05, β = 0.04) are stylized sensitivities anchored to specific historical episodes and tested against an illustrative regression on 1990–2024 annual data. The regression implies a historical α near 0.03 and a β that is statistically insignificant; the model uses 0.05 deliberately, on the argument that AI displacement contracts revenue more than past cyclical episodes (high-wage cohort, no corporate-tax offset). The cascade conclusion is robust across the full plausible coefficient range — even at conservative historical estimates, the moderate, severe, and cascade scenarios all breach the 0.30 threshold. The coefficients govern how quickly, not whether. The 0.55 revenue floor encodes the worst peacetime federal-receipts contraction on record (1932–33); the 0.30 threshold is anchored to the IMF MAC-DSA interest-to-revenue stress band for advanced economies. See methodology for the full derivation, limitations, and citations (Auerbach & Feenberg 2000; Blanchard 2019).
The cascade described above is conditional. None of it has fully fired. But the conditions favorable to it have measurably accumulated over the past eighteen months. Three recent examples worth pointing to:
None of these is the cascade. Each is a piece of evidence that the conditions favorable to it are accumulating.
Each objection below is engaged on its strongest form, not its weakest. The point is not to dismiss the question — it is to show what the cascade scenario actually requires, and where the canonical answer does not yet reach.
The Phillips curve couples unemployment and inflation in offsetting directions through wage-driven price inflation: tight labor markets raise wages, wages raise prices, the Fed responds. The cascade described here does not run through that channel. It runs through asset inflation and a Fed policy trap, with goods inflation potentially muted by automation's downward pressure on costs. The Phillips curve is not wrong; it is silent on this configuration. That the curve has been empirically flat since roughly 2010 — with the 2021–2022 inflation arriving without prior labor-market tightening — only reinforces that it is not a binding constraint here.
Japan has run debt above 250% of GDP for decades with near-zero rates and no cascade. The objection is fair, and the answer clarifies what the cascade actually requires. Japan does not cascade because the conditions that drive the U.S. cascade are largely absent there.
The lesson of Japan is not that high debt is safe. It is that high debt is survivable under conditions — domestic financing, high savings, no coupling shock — that the U.S. does not meet. The U.S. configuration is the photographic negative of the Japanese one.
Capping yields does not escape the cascade; it relocates it. Holding yields below the level the market demands requires the central bank to buy unlimited quantities of debt with newly created money — and to do so in an already-inflationary environment, which accelerates the revenue erosion on the other side of the ratio.
Yield curve control is not an escape mechanism. It changes which wall the system hits, not whether it hits one.
Revenue contraction per percentage point of unemployment, applied to the inner bracket term. A disclosed stress value: historical central estimate ≈ 0.03; CBO cyclically-adjusted range 0.04 to 0.06. Calibrated to the 2008–09 episode (federal receipts fell ~17% against a 4.3-pt unemployment increase); the cascade conclusion holds even at 0.03.
Fiscal-capacity erosion per percentage point of inflation above the 2% target. Calibrated to the post-1985 indexed-bracket era — modern indexation substantially mutes the inflation-revenue channel relative to pre-1985 episodes.
Additional annual debt service per percentage point increase in the weighted-average rate on debt held by the public ($28.3T, FY24). Each percentage point now carries more fiscal weight than at any time since the early 1990s.
Three modeled futures, with the math worked out step by step.
View Scenarios →