The title isn't bad: "ignoring" implies they were never on AWS and just went their own way.

The savings implies they saved $2B over AWS. I can't read the article since I don't have a BI account, but unless they saved $2B over AWS (rather than $2B over their previous infrastructure) it's definitely misleading.

It is a bad title. The inclusion of "ignoring Amazon" implies that the difference between the using AWS and not using AWS is $2B. Otherwise, it makes as much sense to title it "Bank of America saved $2B per year ignoring shark attacks, building own cloud".

And it's still cheaper than public cloud. Look at cloud provider margins, that profit is your opportunity at this scale (and BoA, JPM, etc are big enough to run their own technology teams, and do, having worked with them). Fat public cloud margins come from somewhere.

This isn't a religious argument. Some workloads are fine on prem (known load profiles, shed infrequent traffic spikes instead of bursting), some need cloud (Netflix.com, remember all of their video content streams from their own CDN hardware), it's ensuring you're making the decision in an objective way. Sort of, but not quite, similar to making software budget and timeline estimates.

If software is eating the world, and every company is turning into a software company, why would you be surprised that orgs above a certain size wouldn't also turn into an infrastructure company? You don't have to compete against Amazon, you just need to satisfy your engineering needs for your vertical and the software that drives it. Amazon needs to be everything to everyone.

but why aren't those fat profit margins being competed away by competitors?

Because of the capital and talent required to build a competitive offering. Even large companies like IBM and Oracle have failed to keep up.

The big 3 clouds have built a nice moat and can charge whatever they want. They still compete on prices but it's clear they're colluding on network fees at least.

IBM and Oracle are really not in the same class as AWS, Azure and GCP, when it comes to investing in cloud infrastructure. They’re bit players at best: http://www.platformonomics.com/wp-content/uploads/2018/05/cl...

From this article: https://www.platformonomics.com/2018/05/follow-the-capex-sep...

Yea, that's the point. If they can't keep up then nobody else can.

It's not just capital or talent. I have seen first hand non-tech executive decision makers pick a specific cloud provider because of the old adage "No one got fired for buying IBM". That momentum is incredibly hard to compete against. Google Cloud is fairly solid from a technology perspective, not to mention Google's deep pockets, and still has an uphill climb against MS and AWS.

GCP may be just as good technologically, but their customer support and especially enterprise support is piss poor.

Not directly a response to you, but GCP is actually phenomenal in some areas over AWS: BigQuery, Data Studio (I know it's not really "GCP" but vs AWS' offering), and Pubsub come to mind.

Oddly enough, they falter in other areas: no equivalent to Kinesis Firehose, and true to your point, I was told by customer service just to build my own DataFlow job to take a Pubsub topic and make files.

It's like some of their product managers have a very customer-first mindset (BigQuery, Pubsub) and some could not care less.

Some would say, non-existent.

Their model is third party partners.

There's not much competition because of enormous lockin effects. Once a cloud provider is chosen, you're pretty wedded to it.

Inside your dev team, every institutional bit of knowledge about how AWS works, how to configure it, how to work around their mostly bad docs, and how to navigate their hot garbage UI.

Every line of code that touches an AWS api.

If you need data centers in certain regions for eg localization laws, AWS is afaik the only choice.

Who has the resources and interest to compete in that market who isn't trying to already?

There aren’t that many competitors and they like profit as well.

Besides some edge cases where you actually need elastic scaling, public cloud is always more expensive.