I've seen this over and over with cloud projects. People absolutely get promoted/raises based on optimising/cutting cloud budget. So there's no incentive in doing things efficiently from the start. You're much better off to build a vastly (but maybe not obviously) inefficient first version and then go back later and show how you "saved" the company tons of money by turning off a bunch of unused services etc. It's harder to hide racks of powered off servers which is why this is so prevalent in the cloud.
For many businesses the faster time to market is worth the additional starting costs. Doing it quickly and optimizing cost later has been an explicit desire for a lot of organizations I have been a part of.
