I'm not even sure that's close to true anymore. Like, the biggest companies in the world, when they need to use a relational database, typically go for something "not oracle".

Wikipedia gets more traffic than most enterprises and they use mariadb. They don't even really do it in the weird ways that AWS does, where they use it as a bizarre caching layer.

It's kinda like how for the longest time "mainframes" were the go-to computer for enterprise applications with extreme IO needs. You probably wouldn't recommend someone go talk to IBM if they need to process a lot of records in this day and age unless they were already neck deep in cobol, the go-to language for enterprise applications.

processing extreme amounts of data at extreme io rates

For anybody trying to take that seriously, keep in mind that the Oracle's query planner is absolutely shit, the write performance can go to shit depending on the contents of your transactions, any advanced transaction insulation also destroys performance, and single-machine performance is completely shit, the extreme numbers are all on scalability and when compared to other (unpatched on the case of Postgres) ACID databases.

Also, it's not clear if those claims are still real, because you can only run Oracle databases if you sign a contract saying you won't compare it to other systems.