The AI boom rests on one load-bearing assumption: that somewhere at the end of the chain sits a final customer — someone who pays for compute out of earnings rather than borrowed or recirculated capital.
Follow the money, though, and that customer proves remarkably hard to find.
Neoclouds pay with debt. Labs pay with venture capital and hyperscaler equity. Hyperscalers pay with free cash flow that the sheer scale of the buildout is already squeezing. Revenue gets counted afresh at every stop along the chain; the capital that funded it does not multiply along with it.
Then comes the demand assumption.
The industry increasingly behaves as though nearly everyone wants to become a vibe coder.
But how many people outside the AI ecosystem — ordinary customers spending real, earned money — are actually willing to pay meaningful sums for that?
Step outside the Ouroboros of AI companies funding, supplying, investing in, and selling to one another, and the landscape looks very different.
You find plenty of enthusiasm.
Plenty of demos.
Plenty of subsidized usage.
But remarkably few vibe coders reaching into their own pockets deeply enough to carry the weight of the infrastructure built for them.
In short: no. Coding is not for everyone, nor is it the only domain in which productivity gains matter. We are living through a transition period in which coding happens to be unusually exposed to AI — and that conveniently paints a much rosier picture of broad, durable AI demand than may actually exist.
A few years from now, the vibe-coding boom may look less like the beginning of a permanent mass-market transformation and more like a landmark receding in the rear-view mirror.