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

we're a startup currently handling around 5k requests per second, and we're debating our next scaling move. our current service is running on 2 medium-sized instances, and we could horizontally scale to 8 similar instances, or vertically scale by upgrading to a single, much larger instance. there's a strong bias towards horizontal scaling in general, but for our current workload, the service is mostly cpu-bound, stateless, and doesn't have complex distributed coordination needs. when is vertical scaling actually the right, simpler, and potentially cheaper choice, especially in a startup context where minimizing operational overhead is key? are there specific thresholds or conditions where horizontal scaling introduces more complexity than it solves initially?