“David Friedberg predicts that enterprises will shift their AI inference resource allocation to a model of 70% in a major cloud, 20% on-premise, and 10% in other clouds.”