You have to keep in mind that a company like Blizzard pre-dates cloud tech by decades. I can’t speak much on how they run things, and I assume it’s different on each game, but legacy code is an albatross that consumes all comers. There are so many factors involved in getting anyone to even consider transitioning that just getting a project approved can take years.
As an example: A company I worked at had been talking about a seriously small Kubernetes/AWS based pipeline for six whole webservices, most of which were tiny, for two years before I got there, another year went by with still trying to get it approved, it took another year to get it up and running with enough confidence from management that payment processing wasn’t going to explode, and then another year after that of refinement before I left.
Just off the top of my head: Contracts with existing providers can get in the way. The bean counters want to know that the spend is going to pay off. Depending on how you’re running things, you likely have to tolerate a double-spend period for testing or as a failsafe, and that always looks gross on the books. Sometimes the cloud infrastructure costs more but saves a ton of money in ways that are hard to quantify in $$$. You need people who know how to do it correctly so you can actually leverage the cloud features instead of it just being “Our static servers are in AWS now!” You can’t grind development of the game to a halt while moving it over, so you need more heads - not just engineering, but also QA, PMs, and so on. You also have to content with finding people who are interested in doing it or in learning new tech, which many are not. Again as an example, nobody else on my team had any interest in learning how to do even basic management of our K8S cluster even as I was on my way out the door and was the only one who did.
You might think that doesn’t matter for something like D4 which is an entirely new game, but it totally does. Engineers, in my experience, like to do what they know far more than they like to try new stuff, no matter how goofy it is. That includes straight up ripping code/methodology from past projects. If you’re pulling internal talent from around the company - which as I understand it Blizzard does all the time - you might be getting people who have NFI what they’re doing with more modern cloud tech/features and they default to not that because their confidence in it is higher. Or you might get a massive discount from your current provider if you keep using them for new projects. Or, as the games industry has an awful reputation for, you aren’t willing to pay people enough and you can’t attract the experienced professionals you need to build it out.
Or, perhaps most importantly, you might hit a wall just trying to answer “Who cares?” to the people that make decisions. As in, “If a beta weekend goes kablooie for some people because the infrastructure isn’t there, who cares?” or “If there are huge server queues for the first week or two that we know will disappear, who cares?”
.