The infrastructure that lets AI agents trade with each other is being built while the physical infrastructure to power them remains politically contested. Anthropic ran agents through a real marketplace making actual purchases, a functional demonstration of autonomous economic behavior that required no new regulation to exist. Maine's governor vetoed a data center moratorium that would have frozen new facilities through 2027, rejecting the first serious attempt at state-level infrastructure constraint. Meanwhile, Cohere acquired Aleph Alpha with backing from a European retail conglomerate specifically to create a non-American alternative, revealing that governments now see AI capability as infrastructure worth protecting from foreign dominance the way they once guarded telecommunications and energy. The tension is sharp: builders have working agents conducting commerce while politicians and established players are still negotiating whether the power plants should exist and who gets to own them. OpenAI's apology to a Canadian town about a mass shooting suspect signals the company recognizes it operates in physical communities with real consequences, yet the infrastructure race itself remains abstracted from local impact. Google, Amazon, and Microsoft are competing for cloud dominance through AI chips and models rather than through price or service breadth, which means the question of computational power is now inseparable from the question of which company's AI you run. The market structure is clarifying: whoever controls the chips, the data centers, and the agent platforms controls the economic layer that sits on top of them.
Sloane Duvall