The same as for Hivemind Core. Quoting this from the README (approx.):
HARDWARE:
- 2.5GB of memory for hive sync process
- 250GB storage for database
And then there's the bandwidth for reading every new block that's added to the chain in realtime (after the initial sync).
RE: Announcing Hivemind X: Who wants to run a node?