Dimitri Spanos talks with Molly Rocket about why AI companies keep pushing scale across data centers, users, tokens, and revenue. The discussion separates ordinary market mania from AI-specific pressure such as logarithmic scaling claims, limited fresh training data, subsidized user input, and the need to justify very large capital spending. It also gets into rougher edges like workplace monitoring for training data, military and intelligence demand, copyright litigation, and AI-generated code that creates more AI maintenance work.