The value migration machine
Drag the price of intelligence down three orders of magnitude and watch where the value goes. A toy model with its assumptions showing.
try the slider, it's oddly satisfying
trying to answer →Where does value move when intelligence becomes commoditized?What becomes scarce when intelligence becomes cheap?
This is a deliberately small model. It splits the value created by AI products into six layers and asks one question: as the cost of a unit of intelligence falls, how do the layers’ shares change? Drag across it.
At today’s price, intelligence itself keeps 52% of the value (it keeps 52% today)
How it works
Each layer’s share is proportional to a · ce. Here c is the cost of intelligence relative to today, a is the layer’s share today, and e says how the layer responds when intelligence gets cheaper. A positive e means the layer is intelligence and deflates with it. A negative e means the layer is a complement and gains.
| Layer | Share today (a) | Response (e) | Reasoning |
|---|---|---|---|
| Chips & energy | 0.18 | −0.08 | Cheaper intelligence means vastly more of it; the physical inputs gain slowly. |
| Frontier models | 0.30 | +0.35 | The thing being commoditized. |
| Thin applications | 0.22 | +0.25 | Wrappers are made of the commodity. |
| Proprietary context | 0.10 | −0.15 | Data and context that the model can’t get elsewhere. |
| Distribution | 0.12 | −0.18 | When supply is infinite, reaching people is the constraint. |
| Trust & accountability | 0.08 | −0.25 | Somebody has to sign. |
What would change my mind
If frontier models kept a durable quality lead, their exponent would be close to zero, and most of this picture would collapse back toward today. So the model is really a bet that commoditization happens. The numbers in the table are there so that bet is explicit.