Not just startups, established companies as well. My employer (technically not itself large but a wholly owned subsidiary of a high-ranking Fortune 500 company)

Not just startups, established companies as well. My employer (technically not itself large but a wholly owned subsidiary of a high-ranking Fortune 500 company) has been undergoing a long process of moving almost everything into the cloud and deprecating on-prem resources. The way they've done this was almost comically silly at first: letting every one of our scores of data scientists provision 32 or 64 node clusters all day, every day, without (initially) any thought or guidance given as to how to control costs. At some point they decided this was too expensive and switched cloud providers, even though the cloud provider wasn't the problem. Now I have people breathing down my neck every time I need a decent amount of compute for something ... it pains me to think of what kind of on-prem setup they could have stood up for the tens of millions (or more?) they've spent moving to "the cloud".