Intelligence Scales, Matter Does Not: Physical, Monetary and External Constraints on AI Funded Universal Income

Technology leaders argue that artificial intelligence and robotics will make goods radically cheap, and that a universal basic or "high" income can replace wages as the demand base of a post-labour economy . This paper tests that proposition in its limit case, full automation with a zero labour share, which is the hardest case for its monetary claims and the most favourable for its promises on physical goods. It sets Kaleckian income accounting, a biophysical constraint on physical output, standard incidence analysis and a balance of payments condition side by side in a common notation, and derives accounting results and bounds rather than solving a closed dynamic model.The monetary side of the proposal is coherent in nominal terms. With a balanced budget and no household saving, after-tax profits equal owners' own investment and consumption, so the circuit closes at any transfer level and no "new value" is missing. The binding constraints are real rather than monetary.First, the maximum feasible growth of physical consumption is bounded by the growth of physical inputs plus dematerialisation, independently of the growth of machine intelligence. Once automation has removed labour costs, sustaining a halving of physical goods prices every two years would require material efficiency to improve far faster than any economy-wide rate on record. Second, where physical supply is inelastic over the policy horizon, additional spending on physical goods is capitalised into scarcity rents; housing-benefit evidence from France (50–80% absorbed by rents) and average mine lead times of about 14–16 years indicate that this horizon can be long. Third, an open economy must pay for imported inputs with exports, foreign asset income or foreigners' willingness to hold its claims, and an economy with neither physical inputs nor margins on its exports f aces the tightest bound.Ownership-based variants, such as citizen wealth funds, change who receives the surplus and the bottleneck rents, but not the physical bound on output. Technological step changes, such as new materials discovered by AI, relax the bound, where they are embodied in new capital, only as fast as that capital is rebuilt; a gap may therefore open between potentially f aster cognitive displacement and slow physical provisioning.The AI-abundance narrative is thus monetarily consistent but overoptimistic about physical delivery at historically observed rates. The framework identifies the variables that decide what a transfer can buy , namely physical supply and its lead times, the ownership and taxation of bottleneck rents and access to imported inputs, and yields falsifiable predictions for prices, factor shares, terms of trade, lead times and diffusion.