I think the big gamble is the bet on recursive self improvement (RSI) and getting there first is assumed to be the moat. At least I think what is being sold is this to justify the huge investment on future returns. I personally think that the probability of that happening is pretty low, but investors are swayed by the enormous returns it implies. I guess we will see how it plays out in a few years time. Until then it is important to keep a cool head, experiment with the new tools, and keep your reasoning skills sharp. As someone who can not personally invest on RSI, I feel my only play is to bet on AI becoming a commodity and ready for that future. If that happens, a lot of these datacenter bets will pay a fraction of the expected returns, but we will have cheap cloud compute and local models for inference, an age of personal AI that mirrors the advance of personal computing.