Re: 2025
It has become increasingly clear, year by year, that Amazon has NEVER had the right people in the room to design for resilience. Their contract says "best effort", and that's all anyone can get. "If it's not broke, don't fix it." When it inevitably DOES break, scramble to get it working, and then take the most obvious steps to prevent a repeat, but never bother to think about the class of failures represented by this case and prevent THOSE.
On the security side, they clearly are (or were) trying to do the right thing, but again, are comically inadequate.
Microsoft is simply genetically incapable of producing a high-quality product. Read one of their post-mortems. If you've learned SRE at Google, you'll want to be sitting down with a stiff drink.
Not that Google is in practice better. Their policy of deleting accounts without recourse with YouTube & GMail seems to have snuk into GCP, so you can't really use it for anything business critical without some very serious SRE on your end.
As for Oracle, their corporate arrogance means that their system is likely as bad as Azure, it's just too small to notice. (I've some first-hand knowledge.)
I've 0 exposure to the IBM cloud, but it would take a lot to convince me that it's being built & maintained by first-rate engineers & not some offshored devops monkeys.
Salesforce? Don't make me laugh. They cannot document their on-prem offerings, and don't offer the control needed to do serious work.
---
All of which is NOT so say that the cloud is never the right answer. For a lot of startups, is certainly is. But the business risk is a lot higher than you might expect, and so you're going to need to bring in a real SWE with ops knowledge (the Google idea of SRE) fairly early on.