On founder control and no secondary market When Jensen Huang bet on CUDA in 2006, he couldn’t possibly have known it would be useful for deep learning—because deep learning didn’t even exist at the time. What he bet on was just a vague proposition: parallel computing would always be useful for something. This means that NVIDIA’s most critical success factors were not a single strategic vision, but the combo of three things: 1. Betting on a sufficiently general direction (parallel computing) 2. Wait for a super-large and clear PMF application to come in (deep learning) 3. During the more than ten years he was waiting, the direction and he as the CEO wasn’t replaced In 2008 and 2018, the stock price was halved twice, analysts scolded him for not doing his proper job, and wall st gives selling rating. If it had been a professional CEO under pressure from short-term shareholders, CUDA would very likely have been cut in one of the cost reduction rounds, or slashed down to just a skeleton. As a founder, Huang had enough control that the board couldn’t touch him, so CUDA was able to endure until the AlexNet moment in 2012.​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​ 1. Maintain company control 2. Early IPO/TGE pre PMF is against long-term growth