How Far Can an S-Curve See? Logistic and Gompertz Fits to Mobile and Internet Adoption in India and Five Peers, 1990–2025, and Why Early Fits Underestimate the Ceiling

S-shaped curves are the standard tool for describing and forecasting the spread of a technology. I fit logistic and Gompertz curves by nonlinear least squares to World Bank series on mobile subscriptions per 100 people and internet users as a share of the population for India, China, Indonesia, Bangladesh, Pakistan and Nigeria, 1990–2025. Over the full sample both curves describe India's mobile series well (root mean squared error 3.35 for the logistic and 3.13 for the Gompertz, both with a ceiling of 85.1 subscriptions per 100 people), and the Gompertz has the lower AIC in all six countries for mobile subscriptions. The forecasting record is different. Fitted to data up to 2015, the logistic put India's internet ceiling at 20.0% of the population and forecast 19.7% for 2025; the actual value was 70.0%. The Gompertz did better but still forecast only 31.2%. For mobile subscriptions the 2015 fits were close (out-of-sample errors of 8.23 and 5.00). I show why: before the inflection point the data identify the growth rate well but the ceiling poorly, and the logistic forces symmetry, so an early slowdown is read as saturation. India's largest annual gains were 17.6 subscriptions per 100 people in 2010 and 13.9 points of internet use in 2020.