Machine-Capital: Accumulable Capacity, the Optimisation-Belief Wedge: and the Efficiency-Fragility Tension

Economics cast the machine as a substitute for labour. This paper casts it differently. Machine-capital is the artificial twin of human capital, an accumulable, ownable stock of artificial thinking capacity. More of it sharpens everyone's choices and raises efficiency. But machines optimise given beliefs; they do not make beliefs correct, so where learning is unstable more capacity can entrench the wrong model. Because everyone leans on the same machine judgement, inference aligns and dependence concentrates onto one channel that can carry a self-sustaining shock. Efficiency and fragility then grow from a single source and cannot be prised apart by policy, and the safeguard meant to govern machine-driven action lapses as fragility peaks. The paper proves this tension and warns; it offers no remedy, its data illustrating the pattern without testing the mechanism.