The AI Value Shift: Why the Biggest Winners Won’t Be Selling AI Models
Table of Contents
The Shift from Infrastructure to Implementation
In the early days of the social media explosion, a 27-year-old associate at Accel named Chi-Hua Chien identified a six-person startup out of Harvard called The Facebook. It wasn’t the underlying server architecture or the networking protocols that made the platform a global phenomenon; it was the specific, human-centric application of that technology to social networking. Today, as a co-founder of Goodwater Capital, Chien is applying that same pattern-recognition to the artificial intelligence era.
The current market obsession with LLM (Large Language Model) providers and GPU manufacturers mirrors the early days of the PC and web cycles. However, Chien argues that we are witnessing a rapid commoditization of the model layer. The thesis is simple: the companies selling the ‘picks and shovels’ (infrastructure) rarely capture the long-term value that the companies building the ‘cities’ (applications) do.
- The Infrastructure Trap: Historical data from the PC, web, and mobile eras shows that while infrastructure creates initial spikes, the vast majority of market capitalization eventually migrates to the application layer.
- The Price War: Recent moves by Google to lower subscription prices for AI products indicate that the industry is moving from a ‘discovery’ phase to a ‘price competition’ phase.
- The Application Edge: The real winners will be those who use AI not as the product, but as a silent engine for hyper-personalization and solving supply-side bottlenecks.
