Horizontal vs vertical scaling: when is vertical scaling actually the right choice?

We're a small startup hitting about 5,000 requests per second on our main API. Right now, we're running on a couple of `m5.large` instances on AWS. The natural inclination is always to horizontally scale, so spin up more `m5.large`s. But honestly, managing eight `m5.large` instances vs. just two `m5.xlarge`s or even one `m5.2xlarge` initially feels a lot simpler from an operational perspective. Our current bottlenecks are mostly CPU and memory within individual service instances, not really network I/O or inter-service communication. For our current traffic, the larger instances would easily handle the load with plenty of headroom. I'm trying to understand when vertical scaling is actually the *right* choice. It seems cheaper in the short term, less setup, and simpler to manage at our current stage. Is there a specific threshold where the overhead of more instances (even if smaller) starts to outweigh the benefits of a single, more powerful box? Or is horizontal always the ultimate goal, even if it means more complexity upfront?