Those were physical servers -- essentially nothing on a machine except MySQL server. For large systems, that is the only way to go. For small ones, multiple virtual servers should be more cost effective.
Only a few years ago, Virtual machines chewed up something like 30% of the horsepower. Today it is closer to zero. So, even for a 'dedicated' server, "virtual" is more attractive, at least for consistency.
Note: Master-Slave was a requirement for BCP, plus a benefit for read scaling.
Galera gives you some write scaling.
After that, you have to go with sharding.
The problem with write scaling is that every replica (traditional 'Slave' or Galera 'node') has to perform every write.
Feel free to start another discussion (in, say, the Performance forum) to go into more details about scaling options. Please provide more details about your application; there are often different techniques for diff apps, especially in Data Warehousing. See
http://mysql.rjweb.org/doc.php/datawarehouse and a couple of companion blogs on Summary Tables and High Speed Ingestion.