Well, it is already
, so to say. These days conversions between SVN/Git/Mercurial are as cheap as it can get.
But I think I see what you mean, continuing this work-mode with SVN might not make much sense if your SVN-server has to host 100GB of file data for each checkout instead of each clone, regardless of whether or not SVN can actually deal with large file sizes. And of course it would be a single point of failure, as with every centralized versioning system.
Still I'd suggest to not dismiss versioning too quickly, given that decentralized systems exist, and file sharing in this size category today uses even a very similar paradigma (torrenting).