[Barrin92]

Yes, and I think that's the problem. The infrastructure is actually a cost for Uber. It's better to have less infrastructure. I have sometimes the feeling the infrastructure team at Uber is trying to build the tower of Babel to touch the heavens or something rather than considering that the end goal is giving people rides and not building more infrastructure.

I think people should recognise that the taxi market has an advantage. It's a distributed system and a market. Taxis organise themselves.

If you're going to replace an entire market with a centrally planned system and a giant electricity and compute eating machine then you better have something to show for it in terms of efficiency.

[wpietri]

Exactly. It's like measuring the success of building an airplane by how much it weights.

What's especially interesting to me here is how the landscape has changed since Uber launched 10 years ago. In 2010, you legitimate had to build a lot of your own stuff; at that point Amazon hadn't even launched SNS or Redshift. [1] Docker didn't exist. Etc, etc.

So the question for me isn't, "Can Uber justify their apparently large infrastructure?" It's more, "If somebody started an Uber competitor today, how much of the work could they get from open source, PaaS, and SaaS providers?"

[kavalg]

A really interesting question indeed. While Docker containers are a blessing from many perspectives, their orchestration is far from easy (yes k8s I am looking at you). Another consideration of a CTO, when choosing buy vs build, would be the cloud vendor lock-in and pricing consideration. Big cloud vendors may be a good offering for startups, when you need to move fast, growth is more important than margin and vendor lock-in is not an issue. However, as you grow, these things become increasingly important. Just to provoke some constructive thinking, I urge you to consider what businesses would become possible if running cost of IT & online payments equaled zero :)