Key Takeaways
- Foundational AI models will likely become commoditized infrastructure, much like electricity or cloud compute (AWS), limiting provider margins.
- Sam Altman's vision of selling AI intelligence “on a meter like water or electricity” misreads the inherently thin margin structure of utility industries, Evans says.
- The significant value opportunity lies in the application layer, where companies build diverse products and services on top of these commoditized models, similar to mobile apps.
- Unlike operating systems with strong lock-in, AI models currently lack inherent network effects, suggesting perpetual competition and no single winner-take-all scenario.
- Don't confuse complex science and Nobel Prize-level innovation (like flat-panel screens) with guaranteed high margins; many such technologies become low-margin commodities.
Your AI "Utility" Has Bad Margins
Many founders assume the biggest value in AI will accrue to the companies building the foundational models. Benedict Evans, a long-time tech analyst, throws cold water on that idea. He argues that this thinking fundamentally misunderstands how value is captured in new tech stacks, drawing parallels to industries far older than Silicon Valley.
Evans points to Sam Altman's vision for OpenAI: “we're going to be selling electricity, we're going to be selling AI intelligence on a meter like water or electricity.” Evans's blunt response? “My dear sweet child, you need me to explain the margin structure of the utility industry to you.” Utilities, by their nature, are low-margin businesses. They provide a vital, commoditized resource, and their pricing power is heavily constrained.
The core reason, Evans explains, is a lack of network effects. “The models companies crucially what I said is the models don't seem to have network effects. So there doesn't seem to be a winner takes all effect where one of these will run away ahead of the others. So you should have competition indefinitely.” Without strong network effects or proprietary lock-in, different models will constantly compete on performance and price, driving margins down. Think about mobile screen manufacturers – incredible science, yet a low-margin commodity.
The Application Layer: Where the Real Value Lives
If the models become commoditized, where does the value go? Evans predicts it shifts “further up the stack” to the application layer. This is where companies build actual products and services that users interact with daily, powered by the underlying AI models.
He clarifies this distinction with an example: “if you're I don't know an engineering company or a law firm buying a piece of software, you don't care which cloud it runs on and you don't have to like standardize on AWS because that's where all the software is... That's not how it works. That's how Windows OS works, but that's not how how works.” Users care about the software's functionality, not the cloud infrastructure it uses. Similarly, they won't care which foundational model powers their favorite AI app, as long as the app delivers value.
Evans asks, “if the chatbot isn't the UX and it needs to be apps and the model companies aren't going to build that and the models themselves are basically commodities as at least as you can see them as users then why would the model companies have pricing power and wouldn't all the value be further up the stack?” The answer is clear: the application layer is where proprietary data, unique user experiences, distribution channels, and specific product features create defensible moats and pricing power.
What to Do With This
If you're building in AI, stop thinking of foundational models as the ultimate goldmine. Instead, shift your focus to the application layer. Identify a specific user problem, build a unique product experience, and leverage proprietary data or distribution to create defensibility. Your strategy should be less about building the best raw intelligence and more about building the best product using that intelligence.