It depends on what level of "run your own infra" were talking about here. Like, purchasing hardware, racking and connecting, coming up with a nice way to PXE bu

It depends on what level of "run your own infra" were talking about here. Like, purchasing hardware, racking and connecting, coming up with a nice way to PXE build your fleet and a way to manage the OS and config (chef/ansible/puppet/nix?), before implementing a complete monitoring stack that works on network equipment, raid controllers etc, and then bootstrapping your own Kubernetes cluster or hashistack (or pretending that you don't need complexity, but over time cobbling together your own unique undocumented mess). It's a huge amount of work that startups usually can't afford. The time and effort to do it properly is monstrous. You need to be a company of a certain size to devote the required engineering and support staff to it.