Technology S-curves
New technologies diffuse along an S-shaped curve — slow, then steep, then saturating. Unit costs fall with cumulative production. The cost side is forecastable; the timing of the knee and the ceiling are not.
The single most robust stylized fact in the study of technological change is the S-curve. Adoption starts slow, accelerates past a tipping point, then decelerates toward a saturation ceiling. Everett Rogers gave the vocabulary — innovators, early adopters, early majority, late majority, laggards — in Diffusion of Innovations. The Bass diffusion model makes it quantitative: coefficient p captures innovators who adopt independently; coefficient q captures imitators who follow how many already have; market potential M sets the ceiling.
What makes some curves steeper than others is feedback. Wright’s Law says unit cost falls a roughly constant percentage per doubling of cumulative production — the learning rate that drives the cost engine underneath adoption. Curve steepness parameter k rises when learning curves, economies of scale, network effects, and lock-in compound. Modern curves, the wiki says, are getting steeper.
Cost crosses first
The canonical chain runs: unit cost falls predictably; once price-performance crosses a tipping point the technology beats the incumbent; adoption then goes non-linear along the S-curve; linear-extrapolating forecasters systematically underestimate speed and scale. Tony Seba and RethinkX push a stronger version — technology convergence, where several independently improving cost curves intersect at once and open a “vast new possibility space” no single trajectory would predict. Clayton Christensen’s disruptive-innovation frame supplies the incumbent-blindspot story: entrants foothold where leaders ignore, then move up-market. It is influential directionally and genuinely contested as a predictive tool.
The asymmetry that matters for investing: technological cost progress is genuinely forecastable. Wright’s Law beat five rival laws across 62 technologies, with forecast error growing predictably. Adoption timing and the eventual ceiling are far less forecastable. Detecting an inflection in real time, judging whether a market is underestimating or overhyping a curve, and mapping S-curve position into stock evaluation are all open questions on the wiki — question pages without settled answers.