I once had a contract where they spent a decent chunk of money for me to get them running on AWS but set up in prep to do a zero downtime migration to Google, and then to Hetzner, in order to first milk AWS and Google credits... It worked well, but the savings from moving to Hetzner were so large they could've just set things up there from day one (they had massive egress relative to compute, and at the time the reduction in egress fees vs AWS alone was more than their total cost at Hetzner)