This is an amazing anecdote. Spinning up a side company to solve a need you have as a company that isn’t addressed by the market is the winning lottery ticket of entrepreneurship (you have validated the product/service because you need the product/service to conduct your original business, you are intimately familiar with what features you need to offer, how much you can afford to charge for it, etc)
Also, "scratching your own itch" is always the best startup method because you know intimately if your proposed solution really is a viable solution in the first place. It's then just a matter of pricing and cost, rinse and repeat.
This is the story of Interactive Intelligence, a company I worked for long ago. They were going to build children's software, but had trouble setting up a phone system. So they decided to build phone systems, and later sold for 1.3B
You have it backwards. AWS didn't solve anything for Amazon internally (it's actually kinda funny how tricky moving internal systems onto AWS is). FBA and AWS are Amazon taking systems that they already had to build for its own use and selling them as a product/service to external companies. FBA is just a way to use a portion of Amazon's existing distribution system. AWS is a way to use Amazon's existing tech platform(s).
Fulfillment solves warehousing, inventory, and delivery of Amazon packages. Amazon built out their own network. FBA is Amazon renting out that network to third parties - whether they sell on Amazon or on their own ecommerce site.
One possibility related to commingled SKUs would be that they can control coverage of items in warehouses by controlling which warehouse(s) sellers and their product to.
To bring it back to the article, it's a bit like how CannaSOS (think a mix of Amazon & Rotten Tomatoes for weed) are stepping up and developing a crypto token & platform specifically for the purpose of buying/selling marijuana.
