Purely hypothetically, if many of these AI datacenters were unused in the future is there some use that games (MMOs?) could put them to?
The ability to crunch a lot of numbers quickly essentially translates to handling those kind of tasks. Extra computing power suggests significantly increased server capacity - more players able to do more things, limited primarily by the power of the clients handling the other side. In any client/server game, there are server limitations and client limitations. The server limitations are on the developer side, and the client limitations are on the player side. Let's use an existing MMOG like World of Warcraft as an example. World of Warcraft had historical problems with world events like the opening of the gates of Ahn'Qiraj, where everybody on the server wanted to see the big event happen in real time. When it originally happened, there were server crashes due to hitting and going over capacity, players getting dropped connections, and players unable to see things play out. Having a lot more computational power on the back end would solve some of these problems, but not all.
Any given wow server is only able to handle some maximum number of players in the same space within a given time step, limited by the server hardware. When there are too many players on the server, the updates get slower because the single server needs to update a much larger load within the same amount of time. This is why there are usually server caps - if you go past that number, the server may not being able to finish all of its calculations within the given time frame and players start getting incorrect data or even find their connections dropped completely. A future MMOG could (with sufficient network infrastructure) utilize these data centers to handle massive game servers allowing more players to play in the same space concurrently.
Remember that there are client limitations are on the player side as well - even though the majority of the game's logic, rules, and gameplay processing are handled on the server, the client is what displays the game to the player. This means that the client is the one drawing the environments, the mobs, the UI, the VFX, the characters, and all of the other visuals and sounds of the game. As such, the player's computer also has a maximum number of characters, environments, mobs, UI, VFX, etc. that it can draw and display at any given time, completely independent of the server specs. If the player's GPU can only render 1000 characters, it doesn't matter if there are 10,000 or even 100,000 players in the area - the player will only get to see 1000 of them because that's all her GPU is capable of. Dealing with more than the maximum basically entails picking and choosing the higher priority characters to draw, e.g. party members, enemies the player is engaging with, etc. while choosing not to draw less-important characters like the creeps and such. Offloading that kind of computation is possible but laggy - this is the technology behind Google Stadia and OnLive where remote computers do the rendering and then the video is streamed to our clients.
This isn't the only thing we can do with more computing power on the server side - it's just an example of one thing we can do with it. We could use it for other things - better or more elaborate enemy and allied behaviors, trainable pets with memories and personalities, and so on. More computing power is useful for overall quantity of decision-making work, and that overlaps with the in-game AI behavior (enemies, allies, pets, pathfinding, etc.) as well.
[Join us on Discord] and/or [Support us on Patreon]
Got a burning question you want answered?
Short questions: Ask a Game Dev on Twitter
Short questions: Ask a Game Dev on BlueSky
Long questions: Ask a Game Dev on Tumblr
Frequent Questions: The FAQ


















