Comment by peterwwillis

Because all a benchmark tells you is the result of one use case. It's not an estimate of what your use will result in. It's basically a sales pitch. And sadly, upper management falls for it all the time, often without an evaluation period or even spitballing it for a week with the teams that would be using it. Tons of factors will change the result for a particular use case, regardless of how "normalized" the test tries to be (especially considering most are designed for high-load high-performance high-scale scenarios, which isn't the average real-world use case; most people just have unoptimized queries, or a shit legacy stack, or they expect to eventually have a big use case and were told to find either the fastest or cheapest solution, which may not even be in AWS to begin with, and so on...). The AWS cost is just another benchmark, so it's just as much a lie.