Data as a Moat — Mostly Cope

Every AI pitch claims a data moat. Almost none have one. A real moat is data that is proprietary (competitors can't get it), self-refreshing (using the product generates more), and load-bearing (more of it measurably improves the product).

Static datasets — a warehouse of past deals, a scraped corpus, a decade of exports — depreciate. They're assets, not moats: someone else can buy, scrape, or generate their way to the same pile.