First, as you say Amazon is a middleman charging from 3-15% per sale, making many small 3rd party sellers unable to compete (if they pay shipping themselves etc

First, as you say Amazon is a middleman charging from 3-15% per sale, making many small 3rd party sellers unable to compete (if they pay shipping themselves etc.) Moreover, they often abuse their position by insta-banning sellers e.g. upon very first listing etc. due to faulty machine learning fraud detection, so using their platform is increasingly risky even for quality sellers. The main advantages of Amazon are centralization of payment information without the need to enter CC etc. at different shops for every single purchase (which blockchain has no issues with by design), fast shipping times with Prime (this might be tricky to reproduce; here is the supply chain efficiency advantage you mention; nevertheless a drop-shipping idea with generic warehouse robotics would commodize it) and returns for any reason (this is people-intense unfortunately and not so easy to automate right now). For us programmers, they also supply state-of-art merchant API services with some interesting things like real-time repricing possibility, observing top sellers for any category to duplicate etc. yet there are quite a few flaws and their API feels dated and could be made much better. Amazon as a brand name is also trustworthy; here trust-less approach in DApps and observing any seller/buyer's transaction history can obliterate this advantage, though some implications are quite scary. To replace Amazon, you'd need a marketplace, which seems like the easiest thing for a DApp; another DApp for background checks on sellers/buyers; for Prime you'd need additional robotics (which is no longer so difficult to be honest due to advancements in the past 3 years, especially in confined areas with known environment map), cheap land with pre-fab warehouses that could be owned and ran by individual communities. Obviously, human nature would be biggest risk here (trading illegal goods and services via chains of fully-automated companies, referencing each other), and you might argue that orchestrating all of this might require something like Amazon to exist anyway; I believe though that it only takes for Amazon to make a few hundred capable single programmers to become upset to solve many of coordination problems in automated fashion. Legal and political requirements might be much higher hurdle.