Raid 0 multiplies the speed of each drive by the number of the drives total,
Most of the time, no. For naïve disk benchmarks, yes.
RAID-0 puts chunks of data on different disks. Chunk 0 is on disk 0, chunk 1 on disk 1,... Chunks are typically 64 KiB to 256 KiB, depending on the RAID controller or software.
For the sake of the discussion, let's say that you have four disks with 128 KiB chunks. That put's 512 KiB per stripe.
If you read or write 32 KiB, you'll only use one disk, and you'll get the speed of one disk - not four. (Actually, if the alignment of the 32 KiB I/O splits across two drives, it might be a bit slower since you have the overhead of two physical I/O transfers to the disks.)
Only if you read or write a full, aligned 512 KiB stripe (or multiples of full aligned stripes), then you can get more or less linear scaling. (Those naïve disk benchmarks are doing large transfers.) Having very deep queues can also help - especially with NVMe drives.
RAID-0 can be a performance boost - but it's much more complicated than "Raid 0 multiplies the speed of each drive by the number of the drives".