It's difficult to give a general answer to this kind of questions, as it depends a lot on the kinds of workloads that are envisioned. There can be an order of magnitude in requirements between a team of data analysts only performing visual data preparation on Spark (thus, offloading all work to the Spark server) and a team of data scientists doing heavy in-memory Python/R code and/or machine learning.
We would advise that you ask your Dataiku sales contact who will be able to give a better answer knowing your context