A Stanford Lecture Explains Why AI Value Gets Trapped In Chips

A Stanford course led by investor Apoorv Agrawal argues that generative AI flips the usual software economics: expensive GPU compute captures most of the value.

In this model, semiconductors and infrastructure providers such as Nvidia benefit more than AI application companies, which may face thin margins.

The brief says user growth alone will not create software-style profits, so AI products may need advertising to improve monetization.

This helps explain why AI revenue is flowing disproportionately to chipmakers and cloud infrastructure, shaping how investors value the sector.