The people who use IBM cloud are the same people who used them for data centers in the past.

I think part of why we use IBM cloud services is we also use their consulting heavily. Everything runs on rhel and is supported by ibm.

Though, I haven't personally needed a compelling feature which AWS or GCP was missing, so I have never had to use any other product for work.

Azure is a 'best effort' attempt at a seriously important strategic vertical by a very smart company with deep pockets.

MS would throw zillions of dollars and blood at Azure to make it work.

MS has very deep enterprise relationships, way more so than Amazon or Google and in the long run, that may be the advantage.

A lot of 'fancy services' that we see are not 'the thing' that corps want: they want to spin up servers, basic networking and security, do things at scale. That's the bread and butter and Azure can surely provide that.

Being on the buying side of some of these things, even 100% aware that 'other offers may be better' it's extremely hard to fend off the professionalism of a well managed sales effort. It's even rational: entities that are so operationally effective at managing the relationship 'surely must have tech that works'. And it's a 'safe choice'.

I always assumed Azure would carve something out, IBM, not so much. Oracle, somewhere in between.

I was just pointing out how easily we can misjudge things based on our echo chambers.

> MS has very deep enterprise relationships, way more so than Amazon or Google and in the long run, that may be the advantage.

I think this is something even their competitors realize. That's why the current and previous heads of Google Cloud were from enterprise software companies, Oracle and VmWare.

MS will do very well with Azure, but note that their cloud revenue is also Office 365 revenue (which I suspect is a lot higher).

I'd imagine it could be popular at banks that are used to working with them for 30 years

Given that banks were among the earliest commercial customers of computing machinery, it's more like 70 years:

https://www.ibm.com/ibm/history/ibm100/us/en/icons/bankauto/....

Same as Oracle cloud, their users are organisations locked into long-term support contracts, who are offered sweeteners to use their clouds. Which of course only makes them more locked in.

That said, as a person who's coming from AWS, their demo and documentation have thoroughly piqued my interest. There's a decent amount stuff to like. It's got a slick UI, and the APIs appear to be comprehensive enough to cover our use case.

But I do know IBM has some interesting hardware available in the cloud servers. The Power architecture allows, for example, "1st-class citizen" status to GPUs w.r.t. main memory access: They can access main memory using NVlink. NVIDIA has therefore had some on-mainboard, non-card GPUs available mostly (solely?) in IBM power machines.

Obviously this is not enough for an entire cloud business but they're not just an "also-ran" cloud offering.

Softlayer was a very nice place to get bare metal hosting. You get a whole machine to do whatever on, they provide a solid network, and replace components when they fail (and you open a ticket). We didn't use any ancillary services for the most part, we had been using their included DNS, but switched off because update delays were getting too large and we didn't like the support response on that; we used their loadbalancers for one service for a few months, but it had worse uptime than the hosts behind it. I have heard their other ancillerary services were (are?) bad too. From what I saw, support in general was getting worse, but that could have been because of IBM takeover or because we were going from a top ten customer to an average customer.

I would consider them an option in cases where you want to run all the services you consume, but you don't want to run your own datacenters. They had a decent worldwide footprint, and post-IBM they got a lot more datacenters outside the US; that could be a big deal, because sometimes you just need a couple machines in another continent, and it's easier if you already are on board. Also, at least when we were there, private network traffic was free, worldwide, which is nice to manage replication for disasters.

If you're more cloudy though, and make use of integrated services, and small instance types with per minute billing, I wouldn't consider it. Also, they had that global routing issue earlier this year, which would have been hard (impossible?) to mitigate as a single vendor customer.

I have considerably more doubt about them as an AI player. And their efforts to brand themselves as a blockchain player don't even seem to have merited a mention. Have they given up on that already?