Most of the money in AI has gone upstream. Chipmakers gained about $9 trillion in value between late 2022 and September 2026, against $246 billion for the companies everyone said would win, and Who Makes Money in AI by Nitin Kumar maps the twelve positions in the AI stack and the margin, capital, cash timing and rebuild speed each one sets.
Pricing AI Products covers the decisions inside a company: the value metric, packaging and margin choices that decide how much of what an AI product creates ever reaches revenue.
Questions
Who makes money in AI?
So far, the companies upstream. Chipmakers gained about $9 trillion in value between late 2022 and September 2026, against $246 billion for the companies everyone said would win.
How should an AI product be priced?
Start from the value metric, then package and protect margin, because every model call carries a real cost. Pricing AI Products covers usage, token and outcome pricing, tiering and raising prices on existing customers.
What makes an AI application defensible?
Its position in the stack. A system of record keeps about 52 cents of each revenue dollar, and 49 after the capital charge, while a wrapper differs from it by more than fifty times on the compounding test.