Microsoft CEO Satya Nadella issued a stark warning: companies must own their artificial intelligence “learning loops” or risk surrendering all value to a handful of powerful frontier models. He argues that this platform shift is fundamentally different from any that came before it.
Nadella's thesis centers on the interplay between human capital and what he calls “token capital.” Human judgment, he contends, must guide AI systems that continuously learn and improve over time, creating a compounding effect that benefits the firm itself.
The executive emphasized the urgent need for businesses to build both types of capital. Without owning these loops, he claims, organizations will be unable to capture the economic upside of AI, instead funneling it entirely to model providers.
This perspective emerges as companies across industries race to integrate generative AI. Nadella's framing suggests that sustainable competitive advantage will depend not on access to models, but on proprietary data and human expertise that refine them.
Some industry observers note, however, that owning learning loops requires significant infrastructure and talent, potentially creating new barriers for smaller players who already rely on frontier models due to cost constraints.