When Uber started, they basically reinvented half of Google/Amazon since the cloud was no where near as mature as it was now. They were running their own data

When Uber started, they basically reinvented half of Google/Amazon since the cloud was no where near as mature as it was now. They were running their own data centers. It takes a lot of people to rebuild robust NoSql DBs, stream processing, container management, monitoring and the like from scratch. Could Uber be built with a lot less now days? Absolutely, but it's also an interesting balance of even if they could, if their cloud bill for X service is 10m a year, and they think their own team of 4 could build X internally for less then that...why not do that?