Often asked - never answered directly. The correct answer is "it depends" :smileywink:
Apart from the fact that VDR is limited to 100 vm's it really depends on how much redundant data is stored (and empty blocks) and how similar the vm's are. If they are all created from the same template you will have the chance that the deduplication rate is higher than if all of them were very unique. The retention policy is another important factor in respect of the dedupe capacity you need.
I have to protect 70 vm's with about 4TB total capacity. I am able to backup all of them to a single dedupe store of 1TB size, with 350GB currently not used. So all my vm's are shrinked to 650GB total. But this is just an example. As every environment is somewhat unique you may find different results.
Please note that VDR is tested with a maximum of 2 dedupe stores with a 1TB max each. So this is another limitation that you have to keep in mind.
In large environments it is recommended to use another solution than VDR.