I am often told that the AI cycle is overestimated. But Wall Street has very often been wrong about previous technology cycles. Four examples where market size exceeded initial forecasts: 1. PC --- 225 million users expected in 2000, against 354 million observed, 57% above expectations. 2. Internet --- 152 million users expected in 2000, against 361 million observed. +138%. The final market represented more than double what was anticipated. 3. Mobile --- 657 million smartphones shipped expected in 2013, against 1.019 billion observed, +55% above forecasts. 4. Cloud --- $90.2 billion in revenues expected for the top three providers in 2020, against $115.6 billion observed. +28%. In these four examples, achievements exceed initial forecasts by about 70% on average. We may still underestimate the size of the final market. If costs decrease, usage multiplies, and adoption widens, demand may exceed what current models project; they are, in fact, completely traumatized by the dot-com bubble. (translated)