The key piece of information needed to make such a recommendation is the workload. In particular the peak number of IO ops that you expect to see and the response time SLA (don't forget unusual, but critical, events like year end reporting, restoring and rolling forward, rebuilding indexes or dumping and loading....). Without that information we'd just be guessing.
Assuming that this is a live system -- if you do not have that information you need to gather it.
If the system is not yet live then you need to somehow estimate the workload. Then, once you are live, compare your estimate to reality and determine if changes are needed. (You will have a much stronger case if you take this approach -- management will see that you are not just pulling numbers out of thin air...)
The number of IO ops is also sensitive to tuning. If your system has not been well tuned then you probably have significant opportunities to greatly reduce to IO profile.