Expert here, are these made for general purpose workloads or do you expect them to be fast for AI only?