LEO satellites have potential for very low latency over large distances which could be very valuable for multi-zone customers and Amazon itself. Lasers in a vacuum.
But also low bandwidth/area. Which means that this won't be useful for any significant amount of backhaul, only for last mile service in low density areas. Their groundstations will certainly have good connections to AWS, no doubt, but I don't see a huge market for good connection to AWS from sparse area.s
There are some niches, for example oil/gas extraction like the article mentioned, but nothing huge. Maybe Amazon plans on making a play for in-car services providing both the internet connection, and back-end in AWS. But car internet connections have been shifting towards bluetooth as everyone already has phones and service, and it is wasteful to have another service for the car.
You don't need high bandwidth density for this kind of usage. A single downlink in an area connecting to some datacenter in LEO could be useful for latency sensitive stuff. Then on top of that there's a huge market hasn't been cracked yet which is automated planting of crops. It needs 1-2 cm of accuracy to be useful and the best in the market currently is at 6cm. Plus all kinds of IoT stuff and long-range latency sensitive backhaul( in which case LEO constellations win because the distance is actually smaller due to the low number of trans-oceanic cables and the high number of switches along the way).
There are some niches, for example oil/gas extraction like the article mentioned, but nothing huge. Maybe Amazon plans on making a play for in-car services providing both the internet connection, and back-end in AWS. But car internet connections have been shifting towards bluetooth as everyone already has phones and service, and it is wasteful to have another service for the car.