This typically will come down to a latency and cost analysis.
Consider the following from a raw IOPS/disk perspective:
7.2k SATA = 80 IOPS
10k SAS = 160 IOPS
15k SAS = 180 IOPS
SSD = 8000 IOPS
The break point would be at the point where available capacity required meets available performance required at a particular latency number.
If I need hundreds of TB of capacity and reasonable latency numbers (sub 10ms) then havin many SAS spindles will accomodate.
If my required capacity is not extremely high but I need sub ms latecy then a handful of SSD's will do the trick.
In hybrid systems you get the balance of both -- I get the desired capacity of SAS or SATA with a performance enhancement of caching hot data within an SSD cache.
So what is the breaking point? Whatever it is that you need that meets the capacity and latency requirements.
Hoe this helps,