For EvE, which was designed before 2003 (IIRC, that's the release date, design is probably pre-2000) it's the main performance bottleneck. They are not using cloud but just a huge cluster, which is a product of the technology level from 20 years ago. Their RDBMS used to be SQL Server, which is the slowest offering, doubt they changed it in the mean time. As a result, the game pace is very slow, actions take seconds, and massive instances require preparation from CCP so they can allocate stronger cluster nodes to it, and even so massive instances are a slowwww crawl and even counting the time warp trick they use to try and preserver their database from melting.
You cannot afford that in a real time MMO, and technology moved quite a lot during 20 years. That's why CiG are using fancy names for in-memory caches and object persistence. The cloud allow to use distributed in-memory caches that deliver sub-millisecond response times (plus the network lag to the client of course) with O(log

) scaling, using binary tree indexing. You can have excellent performance for writes too if needed, and the eventual consistency is golden for that use case (only resolves when players actually interact, which is comparatively infrequent).
Are CiG using all of that ? Currently it's obviously a big no. From what i can see they use an SQL storage for the backend, as 30+ seconds response times (just to get a single player inventory !) were observed during fleet week. Again, transactions and rollbacks are a no-no in this environment, the game engine has to do without - there are proper ways to do it, but CiG are not known for "proper"...
I'll also point out that performance is a complete chain, and their network code is terribly inadequate, and is the first bottleneck we encounter. I highly suspect that their servers have a single thread dedicated to sorting client networking (then connected to the main game loop server-side), which doesnt matter in a low-player-count FPS, but definitely cannot do more than 50 players at once.