Wright's Law: every doubling of cumulative production drops unit cost by a constant percentage. Rogers Diffusion:
adoption follows an S-curve as innovators give way to early adopters, the early majority, the late majority, the
laggards. Overlay the two and you get a usable theory of when a frontier technology stops being expensive enough
to talk about and starts being cheap enough to deploy.
Cost halving~24 mo
Inflection—
Adoption then—
How the curves are drawn
Cost follows C(t) = C₀ · exp(−λt) with λ chosen so cost halves roughly every 24 months. Adoption follows
a logistic A(t) = L / (1 + exp(−k(t − t₀))) calibrated so the inflection is reached when cost has dropped
by ≥ 90% from baseline. The intersection is found by linearising both segments and solving numerically.
These are illustrative curves, not forecasts. The point is the shape: the cost curve decays fastest while
adoption is slowest. By the time adoption flips into the early majority, unit cost is no longer the binding
constraint. This is why "we are waiting for AI to get cheaper" is almost always the wrong question.
Esc
↑↓ navigateEnter openEsc close
Welcome
Tsholofelo K. Setati
AI. Technology Economics. Governance. Strategy.
Translating emerging technology into compounding business advantage -
with field notes, working demos and a clear point of view.