The promise is older than the technology draped over it. Since Ronald Reagan, Republicans have sold tax cuts as self-financing engines of growth. The cuts arrive, the growth doesn't cover them, and the deficit widens. Scott Bessent's 3% growth target and Trump's claim that the economy is "growing at a faster rate than we've ever grown before" are the latest iteration — this time with AI as the magic ingredient. The pattern is identical; only the justification has been upgraded. The numbers are unforgiving. The Committee for a Responsible Federal Budget ran the scenarios: if temporary tax cuts are made permanent and tariff revenue losses are not recovered, hitting a 3% deficit-to-GDP ratio by 2036 would require 4.4% annual growth for a decade. A balanced budget would require 7.2%. The US has exceeded 3% growth only twice this century outside the post-Covid bounce. Trump's One Big Beautiful Bill Act alone is estimated to add $4.7 trillion to federal debt through 2035, and his proposed $5,000 per-adult "dividend" would dig the hole further. Bond markets are pricing the skepticism directly. The 10-year Treasury yield surged to its highest level in nearly 25 years, more than a full percentage point above levels when Trump launched the Iran conflict. Interest payments on federal debt now consume 3.3% of GDP, up from a 50-year average of 2.1%. Rising yields create a negative spiral: more expensive borrowing increases the deficit, which increases borrowing, which pushes yields higher. The foreign demand cushion that once absorbed Treasury issuance is eroding. Foreign central banks have cut back their exposure to US government debt. The Treasury now competes for private capital against the very AI hyperscalers whose productivity miracle is supposed to rescue government finances. Those companies are borrowing aggressively to fund datacenter buildouts, crowding the same capital markets the government needs. The AI growth thesis has a structural tax problem even if it delivers. An AI-supercharged economy would shift income from labor to capital. The US tax rate on capital is roughly half the rate on labor. So even spectacular GDP growth driven by AI would generate proportionally less tax revenue than equivalent growth driven by wage gains. Meanwhile, mass labor displacement would demand substantial new government spending on safety nets — the fiscal math cuts both directions. Stanford economist Hanno Lustig estimates datacenter owners need 45% average annual revenue growth for seven years just to break even on investments projected to reach $1.43 trillion this year. Jared Bernstein and Ryan Cummings calculate that six superscalers — Google, Meta, Microsoft, Oracle, SpaceX, and Amazon — need $13.1 to $18.7 trillion in additional revenue over a decade to cover AI capital expenditure, roughly matching their entire revenue of the past decade. If those targets are missed, the financial market consequences would compound the government's already precarious fiscal position. Stabilizing federal debt would require total factor productivity growth averaging 2.5% per year over the next decade. The US has hit that threshold once since 1959 — despite electrification, the interstate highway system, the telecom revolution, and the first wave of IT automation. Betting the nation's solvency on AI delivering what none of those transformations could is not fiscal policy. It is a fantasy with a spreadsheet attached.
