>The higher quality of the CUDA toolkit compared to other SIMD languages, combined with it's increasing relevance in compute (gaming, followed by blockchain, followed by ML, followed by GPT) would have made this an NVDA an easy pick for anyone (of the increasing number of people) that worked in parallel computing from 2006-2023.

Hindsight is 20/20. I highly doubt people in parallel computing, unless they already worked at Nvidia, have done better than anyone else with their portfolios. Other than the standard delta you'd assume since those people are probably savvier investors in general.