Historically, when someone tried to emulate a company like Google, the response was that they are not Google so they should choose a solution that fits their scale and worry about scaling when it becomes a problem. The reason being, at Google scale you make a different set of trade-offs that you don’t need to make at regular scale. Those trade-offs tend to be costly at lower scales without benefit. For the most part, this still seems to be true at the application layer.
Is there something different about the infrastructure that makes this view false? Is the view actually false for the application layer as well, these days?