I agree. VMs rely on old technologies, and are reliable in that way. By contrast, the move to Docker then necessitated additional technologies, such as Kubernetes, and Kubernetes brought an avalanche of new technologies to help manage Docker/Kubernetes. I am wary of any technology that in theory should make things simpler but in fact draws you down a path that requires you to learn a dozen new technologies. The Docker/Kubernetes path also drove up costs, especially the cost associated with the time needed to set up the devops correctly. Anything that takes time costs money. When I was at Averon the CEO insisted on absolutely perfect reliability and therefore flawless devops, so we hired a great devops guy to help us get setup, but he needed several weeks to set everything up, and his hourly rate was expensive. We could have just "push some code to a server" and we would have saved $40,000. When I consult with early stage startups, and they worry about the cost of devops, I point out that we can start simply, by pushing some code to a server, as if this was still 2001, and we can proceed slowly and incrementally from there. While Docker/Kubernetes offers infinite scalability, I warn entrepreneurs that their first concern should be keeping things simple and therefore low cost. And then the next step is to introduce VMs, and then use something like Packer to enable the VMs to be uses as AMIs and so allow the devops to develop to the point of using Terraform -- but all of that can wait till the product actually gains some traction.