If everyone can build and sell, who is buying?

I get this question all the time these days, so here is my two cents:

When the supply of something explodes, its price heads toward marginal cost, and value moves to its complements, whatever is still scarce. For example, cheaper development increases demand for the models and compute used to build and run software. That is why frontier labs and coding-tool vendors are commoditizing application layer, which is a complement to their models and compute.

At the same time, profit does not vanish when one layer commoditizes. It moves to adjacent layers that become the new bottleneck. In 2026 there are three such layers:

1. Below the app: models, compute and chips and generally infrastructure
2. Beside the app: distribution, systems of record and proprietary data, brand and trust, big market share in a niche or a particular regulatory standing
3. Above the app: the outcome itself, sold as a service. e.g. FDE, productivity and efficiency. In general, the “Services-as-Software” thesis where the target market is labor budgets

Those layers point to an emerging hybrid building-buying future where for example, a business can decide to build and maintain its own internal applications while paying for compute, relying on an established system of record or purchasing a service/app that takes full responsibility for the finished work.

So who’s buying? Pretty much the same customers as before. They are just paying different people for different things.