AI Moats

Spending the most on AI compute is not a moat. It is a cost. In the last week of January 2025, a Chinese lab released a model that matched the Western frontier at a fraction of the cost, and Nvidia lost close to 600 billion dollars of market value in a single day. The theory that AI leadership belongs to whoever spends most on compute survived about seventy-two hours of real evidence.

So where does durable advantage in AI actually come from? A real moat compounds with use, stacks at least two of four currencies (data, compute, distribution, and trust), and can be specified mechanically. Anything else is a marketing claim. AI Moats gives you the four currencies, a six-question diligence framework for any AI strategy claim, and how to design a defensible stack and build defence in depth.

Written for founders, operators, and investors who have to decide where the defensibility really sits. Part of the Executive AI Mastery series, alongside TRANSFORM and Ethical AI.