[emteycz]

What's lacking in Google Collab, AWS, GCP and Azure? In my non-ML programmer point of view what you're saying is already available. Why do you think it's not?

Isn't it true that the most high end hardware is available only in these clouds anyways?

[dbish]

+1. I'm not sure exactly what the issue is here being described, but pay-per-use, running of large scale ML training techniques across high end instances can be done today. Maybe it's not cheap enough for this person, but if that's the case, I don't really expect a startup to find ways to make it cheaper (unless they subsidize with VC money, but that's not sustainable).