[Neil44]

Almost all decent datacentres and compute providers have these features, and it's far from unusual to have downtime in an EC2 reigion to the point where the standard advice is to spread across multiple reigions to avoid being caught out. EC2 does not have DDOS protection as standard its an additional cost service (AWS Shield). They will just charge you an insane ammount for the DDOS bandwidth as standard.

Linode, DO etc all have these features at a fraction of the price.

[StreamBright]

>> Almost all decent datacentres and compute providers have these features

Multiple tier1 networking routes? I seriously doubt that. Those vendors won't deal with small players. AWS used to have 160Gbps in NA 10 years ago. EC2 has DDOS protection, for one simple reason, if somebody DDOSing it it will impact many customers. And you can also use Shield. The point here is that you have options. What sort of DDOS protection Linode or DO have? How much bandwidth did they get? How many tier1 routes? There are the sort of questions you need to ask for a fair comparison, which absolutely not happing on HN. Many people think that building a data center is like creating a hello world application. I have seen this repeadetly, several times.

[Wintereise]

Multiple tier1 networking "routes" are really, really easy to do these days. It's within the reach of pretty much anyone with a semi-sizable engineering budget. This includes carriers like CenturyLink/GTT/NTT and most of the global tier ones. The carriers are hungrier than ever for business, you can do this at a fraction of what AWS egress costs per GB.

So yes, really. Nothing on your list is unattainable at much, much lower cost if you're willing to put in the work at maintaining it yourself (no cloudformation at most traditional providers of this stuff, and while you can somewhat roll your own with k8s for /some/ stuff, the offering traditionally is nowhere near as cohesive.)

The non-open source but complementary services (think Dynamo/EventBridge etc) are imo AWS' real strength. If you need to be able to just throw cash at the problem and never worry about capacity management yourself, AWS (and GCP, and Azure, and <insert PaaS provider here>) is usually an excellent fit.