This is not irrational FOMO from investors, it is the reality that to get up and running with a competitively trained model from scratch you need to spend milli

This is not irrational FOMO from investors, it is the reality that to get up and running with a competitively trained model from scratch you need to spend millions. Software startups have traditionally not been very capital intensive, but clearly that doesn't apply to ML startups. Yes, you could have given a smaller seed and asked them to develop a POC, but with their credentials, I don't think their ability to deliver on that is really in doubt. The real question is whether they can create something on the scale of competing with OpenAI. I think that is a pretty good investment thesis, even if it is risky.