Timing, luck, and path dependence

The mechanism

The uncomfortable part of many great outcomes is that the decisive event arrived from outside and on its own schedule, so the bet only looks brilliant once the world happened to turn toward it. NVIDIA shipped CUDA in 2006 and then waited: the breakthrough came in 2012 when a few graduate students in Toronto trained their neural network on two off-the-shelf GeForce gaming cards and won an image-recognition contest, running on about a thousand dollars of consumer hardware NVIDIA had built for gamers, not researchers, and the full payoff waited until ChatGPT in late 2022 made every cloud need GPUs at once. TSMC's founder is even blunter about it: Morris Chang says of the fabless industry that made his foundry possible, "I only hoped for it," admitting that the macro shift his whole company depended on was something he bet on rather than foresaw. The skill in both cases was being built and ready when the window opened, but the window itself was not theirs to open.

The tension

The trouble with timing is that you cannot manufacture the trigger, so the same patient, correct-looking bet pays off enormously or sits dead for years depending on an event the strategist does not control. This is harder than it looks from the outside, because the stories we study are the ones where the window eventually opened, and that survivorship bias quietly edits out every equally smart bet that simply never got its AlexNet or its fabless boom. Path dependence sharpens the trap: an advantage that grew out of one lucky sequence often cannot be reproduced on purpose, either by the company that holds it or by anyone trying to copy it. The honest reading of these stories is that preparation was necessary and the timing was decisive, and the two are very hard to separate until long after the fact.