Just use an external RAID/JBOD/NAS connected with Thunderbolt 3 or Ethernet. I have a 16TB Drobo connected via Thunderbolt to hold all my media (music, videos, etc).
The really serious work of that type is being done on the "Big Iron" servers like Aidenshaw supports with up to 8 CPUs with hundreds of total cores, terabytes of memory and petabytes of cloud storage. The PC is just the front end terminal to send the jobs and look at the returned data so you won't be using a Mac Pro (or a Z8) to do that. More like a MacBook Pro or Dell 5000 series.
The Mac Pro will be used for machine learning, but it will be based on Apple Core ML.