Amazon said (to Congress, no less) that it does not let it's own team responsible to deciding on which products to buy & sell access confidential sales reports of third party sellers that use their platform. And then it was revealed that they regularly do get access to these information.
There's no reasonable expectation that Amazon will not use any data it can easily access to make decisions.
I don't find that your argument regarding "using US created computers" and "US-owned gTLD" holds water. It's very hard to build e.g. CPUs that will contain very advanced backdoors that are totally hidden but can easily be targeted remotely to gather information, and the same goes for secretly using a domain. Additionally, the cost of doing so is astronomically high: the total loss of trust in US chip makers and internet companies. They may very well do that during major wars, but not to steal some company secrets.
On the other hand, it's easy for Amazon to siphon off data from AWS and even just using meta data (e.g. looking at how and when a customer scales to what size) gives you valuable insight into their business. And if they get caught? A few politicians will be annoyed, a few media articles (but hey, they may overshadow reports on unionization, so it's not all bad) and potentially a few companies leaving their platform, while others will undoubtedly say "I have nothing to hide from Amazon, so what?"
In other words the threat model is dramatically different.