The Wrapper Economy
I've been watching the AI tool landscape evolve, and something fascinating is happening beneath the surface. Companies like Cursor are raising at $9 billion valuations while reportedly making just ten cents for every dollar they spend. The economics are wild, and they reveal something important about where we're headed.
Most of today's beloved AI startups aren't really product companies—they're distribution layers for someone else's intelligence.
(Note: This mirrors earlier platform dynamics, but the unit economics are more extreme because compute is expensive and models improve rapidly.)The New Food Chain
The structure is becoming clear:
The Labs (OpenAI, Anthropic, Meta) develop the core models—the actual intelligence.
The Wrappers (Cursor, Bolt, Lovable) package that intelligence into user-friendly experiences.
The Users get powerful tools without thinking about the underlying complexity.
It's a classic supply chain, but with a twist: the suppliers keep shipping your best features natively.
The Cursor Paradox
Cursor built an incredible AI coding experience. They raised nearly a billion dollars and created something developers genuinely love. But underneath, they're essentially a very expensive interface to Claude and GPT.
(Note: Their value isn't just the wrapper—it's the UX insights, the context management, and the user relationship. But those advantages can erode quickly.)The paradox: the better they prove the market, the more attractive it becomes for their suppliers to compete directly.
The UX Factor
This dynamic reveals something interesting about user experience. The wrapper companies are often better at UX than the labs. They understand specific workflows, they build for particular use cases, they smooth the rough edges.
But they're also fronting massive inference costs while training users to expect a certain level of capability. They're essentially market research for the companies that could replace them.
Two Possible Futures
Path 1: Direct Integration The labs get better at UX and go direct. Why pay Cursor when Claude can code just as well in any interface?
Path 2: Commoditization Models become interchangeable, like generic aspirin. The wrapper companies win through better design, faster iteration, and deeper understanding of specific workflows.
(Note: Path 2 requires either proprietary models, unique data, or such superior UX that switching costs outweigh model improvements.)I suspect we'll see both paths play out in different categories.
What This Means
For users, this competition is fantastic. We get powerful tools at venture-subsidized prices while companies figure out sustainable economics.
For builders, it's a reminder that unless you own the core technology, you're always vulnerable to your supplier becoming your competitor.
For the industry, it's a preview of how AI capabilities will eventually be distributed. Not everyone will run their own models, but the companies that control the models will have enormous leverage.
The current AI boom feels like the early mobile app store days—lots of experimentation, venture-funded user acquisition, and uncertainty about who captures value long-term.
The difference is that in mobile, Apple and Google controlled distribution but not content creation. In AI, the model providers control both the intelligence and, increasingly, the interface.
The wrapper economy is teaching us what users want. Whether the wrappers survive to deliver it is another question entirely.