Satya Nadella told companies that picking the fanciest AI model is not a strategy. The Microsoft CEO outlined a framework for competitive advantage on X. The core argument: organizations must build proprietary "learning loops" instead of outsourcing their intelligence.
Nadella introduced two key concepts. "Human capital" covers employee expertise and institutional workflows. "Token capital" refers to the proprietary AI assets a company develops and improves over time.
He coined "token-maxing" to describe what companies should avoid. That is throwing more compute at problems without building feedback infrastructure that makes AI useful in context.
Nadella explicitly warned against a scenario where a handful of AI models absorb most industry expertise. He labeled that outcome "economically, politically, and socially unstable."
The timing aligns with Microsoft's messaging at Build 2026, its annual developer conference. Microsoft wants to be the platform enabling companies to build their own learning loops.
The token capital framework has implications for AI-adjacent crypto tokens. Projects demonstrating genuine learning loops, where usage improves the AI models it serves, have a different value proposition than tokens providing access to a static model.
For investors, Nadella's memo offers a filter. Ask whether a project is building a learning loop or just wrapping an API in a token.