How Big Tech Funds Its Own Replacement

Abstract. The self-undermining theorem shows that concentrated investment in centralized AI infrastructure accelerates the learning curves that make distributed alternatives viable — the more hyperscalers invest, the faster they get displaced.