Every mainstream Minecraft server, vanilla, Paper, Purpur, all of it, runs its core game logic on a single thread, no matter how many CPU cores the hardware has. Folia, a fork of Paper built by the PaperMC team, is a direct attempt to break that limitation, and it works in a genuinely different way from a normal performance optimization.
The problem Folia solves
Adding more CPU cores to a normal Paper server does not make the world simulation itself faster, because that work is fundamentally single-threaded. Async chunk loading and other Paper optimizations help around the edges, but the core tick loop, entities, redstone, block updates, is one thread doing one thing at a time. On a small server this is rarely the bottleneck. On a server spread across a huge, heavily explored world with hundreds of concurrent players, it eventually becomes the hard ceiling no amount of extra hardware can push past.
How Folia actually works
Folia divides the world into independent regions based on player and entity activity, and gives each region its own tick loop, run concurrently across a thread pool instead of a single thread. Regions that are far apart on the map can be processed at the same time on different CPU cores. The official Folia documentation covers the region model in more depth if you want the full technical picture.
The real cost: plugin compatibility
This is the part that matters most in practice. Regionized multithreading means plugin code can now run concurrently against data that used to be safely single-threaded, which introduces real race-condition risk if a plugin was not written with that in mind. Because of this, plugins have to be explicitly marked as Folia-compatible (folia-supported: true in their plugin.yml) and, in most cases, updated by their author to work correctly. Every plugin that exists needs at least some level of review before it is safe on Folia, and a large share of the plugin ecosystem simply has not been updated for it yet.
Who actually needs Folia
Folia is built for a fairly specific and fairly large scale: servers with 200+ concurrent players spread across a wide, actively explored map, where a normal Paper server's single tick thread is the demonstrable bottleneck. If your players tend to cluster in the same general area, Folia offers little to no benefit, since regionization only helps when activity is spread out enough to actually parallelize. For the overwhelming majority of servers, including busy ones, plain Paper (or Purpur) remains the better choice today, both for raw stability and because the plugin ecosystem is built for it first.
Where to start instead
If you are not confidently in Folia's target scale, Paper gives you the vast majority of the performance benefit with none of the plugin compatibility risk. A server status tool is a simple way to keep an eye on how a growing server is actually performing before deciding you have outgrown Paper. Tick Hosting's panel supports Paper and Purpur on every plan, free tier included, from Tick Hosting's plans page.